The evolution of arthropod limbs

253 Biol. Rev. (2004), 79, pp. 253–300. f Cambridge Philosophical Society DOI : 10.1017/S1464793103006274 Printed in the United Kingdom The evolution of arthropod limbs Geoff A. Boxshall Department of Zoology, The Natural History Museum, Cromwell Road, London SW 7 5BD, UK (Email : [email protected]) (Received 24 May 2002 ; revised 2 April 2003) ABSTRACT Limb morphology across the arthropods is reviewed using external morphological and internal anatomical data from both recent and fossil arthropods. Evolutionary trends in limb structure are identified primarily by reference to the more rigorous of the many existing phylogenetic schemes, but no major new phylogenetic inferences are presented. Tagmosis patterns are not considered, although the origins and patterns of heteronomy within the postantennulary limb series are analysed. The phenomenon of annulation is examined and two basic types of annuli are recognised : terminal and intercalary. The annulation of the apical segment of a limb results in the formation of terminal flagella, and is typical of primarily sensory appendages such as insect and malacostracan antennules and maxillary palps of some hexapods. Intercalary annulation, arising by subdivision of existing subterminal segments, is common, particularly in the tarsal region of arthropodan walking limbs. Differentiating between segments and annuli is discussed and is recognised as a limiting factor in the interpretation of fossils, which usually lack information on intrinsic musculature, and in the construction of groundplans. Rare examples of secondary segmentation, where the criteria for distinguishing between segments and annuli fail, are also highlighted. The basic crown-group arthropodan limb is identified as tripartite, comprising protopodite, telopodite and exopodite, and the basic segmentation patterns of each of these parts are hypothesised. Possible criteria are discussed that can be used for establishing the boundary between protopodite and telopodite in limbs that are uniramous through loss of the exopodite. The subdivision of the protopodite, which is typical of the postantennulary limbs of mandibulates, is examined. The difficulties resulting from the partial or complete failure of expression of articulations within the mandibulate protopodite and subsequent incorporation of partial protopodal segments into the body wall, are also discussed. The development and homology between the various exites, including gills, on the postantennulary limbs of arthropods are considered in some detail, and the question of the possible homology between crustacean gills and insect wings is critically addressed. The hypothesis that there are only two basic limb types in arthropods, antennules and postantennulary limbs, is proposed and its apparent contradiction by the transformation of antennules into walking limbs by homeotic mutation is discussed with respect to the appropriate level of serial homology between these limbs. Key words : Arthropoda, skeletomusculature, limb segmentation, annulation, arthropod evolution. CONTENTS I. Introduction ................................................................................................................................................. II. Material examined ...................................................................................................................................... III. The antennule .............................................................................................................................................. (1) The uniramous antennule of the arthropodan groundplan ............................................................ (2) Biramous or polyramous antennules .................................................................................................. (3) Loss of antennules ................................................................................................................................. (4) Antennulary development and the origin of flagella ........................................................................ (5) Function ................................................................................................................................................. (6) The antennule of the arthropodan groundplan ................................................................................ IV. The postantennulary limbs – homonomous or heteronomous series ................................................... V. The postantennulary limb of the arthropodan groundplan .................................................................. (1) Protopodite ............................................................................................................................................. (a) Defining the protopodite ................................................................................................................ 254 255 255 257 261 263 263 265 266 266 267 268 268 Geoff A. Boxshall 254 VI. VII. VIII. IX. X. XI. XII. XIII. (b) Evidence of protopodal segmentation .......................................................................................... (c) The origin of coxa-basis differentiation ........................................................................................ (d ) Outer lobes – exites, epipodites and the question of insect wings ............................................ (2) Telopodite .............................................................................................................................................. (a) Primary segmentation ..................................................................................................................... (b) Intercalary and terminal annulation ............................................................................................. (c) Secondary segmentation ................................................................................................................. (3) Exopodite ............................................................................................................................................... Origin of the biramous limb ...................................................................................................................... Origin of uniramous limbs ......................................................................................................................... Differentiation within the sequence of postantennulary limbs .............................................................. (1) The first postantennulary limb ............................................................................................................ (2) The second postantennulary limb ...................................................................................................... (3) The third postantennulary limb .......................................................................................................... (4) The fourth postantennulary limb ........................................................................................................ (5) The fifth postantennulary limb ........................................................................................................... (6) The sixth postantennulary limb .......................................................................................................... (7) The seventh and subsequent postantennulary limbs ........................................................................ Heteronomy ................................................................................................................................................. Segmentation and musculature ................................................................................................................. Conclusions .................................................................................................................................................. Acknowledgements ...................................................................................................................................... References .................................................................................................................................................... I. INTRODUCTION The debate on arthropod origins and relationships has traditionally raged around key questions relating to limb structure and body tagmosis : is the mandible gnathobasic or whole limb in origin ? How many segments comprise the euarthropodan head ? Are the thorax and abdomen of hexapods homologous with the thorax and abdomen of crustaceans, and how do these tagmata relate to the prosoma and opisthosoma of chelicerates ? In recent years, this debate has been transformed by the advent of new methodologies : the methods of phylogenetic systematics have injected rigour into the process of estimating relationships, the use of DNA sequence data has provided a novel set of characters for use in phylogenetic analysis, and the techniques of developmental genetics have, particularly through the analysis of Hox genes, granted us insight into the basic genetic mechanisms regulating the expression of morphology. The impact of these methods at all levels has been profound. There is, for example, some evidence that the nearest relatives of the Arthropoda may be found within a group of moulting organisms, the Ecdysozoa (Aguinaldo et al., 1997 ; SchmidtRhaesa et al., 1998), rather than within the Articulata [ but see Wägele et al. (1999) for a contradictory view]. Similarly, acceptance of the long-standing relationship between Hexapoda and the Myriapoda (Pocock, 1893; Tiegs & Manton, 1958; Manton, 1973 ; Fryer, 1997) is seriously challenged by an accumulating body of molecular and morphological data supporting a close relationship between Hexapoda and Crustacea (Averof & Akam, 1995; Edgecombe et al., 2000; Giribet et al., 2001). These data include analysis of nuclear ribosomal sequences (Friedrich & Tautz, 1995), nuclear and mitochondrial protein-encoding gene sequences (Regier & 271 273 275 279 279 282 283 283 285 286 286 286 287 288 288 289 289 289 290 291 292 293 293 Shultz, 1997 ; Shultz & Regier, 2000 ; Hwang et al., 2001), mitochondrial gene order (Boore, Labvrov & Brown, 1998) and Hox gene sequences (Cook et al., 2001). Neither these results nor the continuing revelations from the study of arthropod developmental genetics are reviewed here. Excellent reviews of the evolution of developmental patterns in arthropods can be found in Akam (2000) and Scholtz (2001). The discovery or restudy of fossil arthropods has also played a major role in the renewed debate on arthropod origins and relationships. In particular, the meticulous studies of the Burgess Shale fauna by Whittington and colleagues (Whittington, 1971, 1974, 1977, 1980, 1981 ; Hughes, 1975; Briggs, 1976, 1978 ; Bruton, 1981 ; Bruton & Whittington, 1983 ; Gould, 1989; Briggs & Collins, 1999), and the continuing discoveries from the Orsten (Müller, 1982 ; Müller & Walossek, 1985, 1986a, b, 1987, 1988 ; Walossek, 1993; Walossek & Müller, 1990; Waloszek & Dunlop, 2002) and Chengjiang faunas (Hou & Bergström, 1997; Ramsköld et al., 1997; Shu et al., 1999) have demonstrated the amazing diversity of Cambrian non-trilobitic arthropods. The description of Ercaia minuscula Chen, Vannier & Huang, interpreted as a possible derivative of the stem-line Crustacea, from the Maotianshan Shale in China (Chen, Vannier & Huang, 2001), and of phosphatocopines from the Lower Cambrian (Siveter, Williams & Waloszek, 2001), indicates that the divergence of major arthropodan groups had taken place by the early Cambrian. Further discoveries from the Hunsrück Slate have also highlighted the survival of archaic arthropod morphologies into the Devonian (Stürmer & Bergström, 1976, 1981 ; Briggs & Bartels, 2001). The inclusion of fossils into comprehensive phylogenetic analyses of the Arthropoda has resulted in the production of a veritable forest of phylogenetic trees (e.g. Wills et al., 1995, 1998 ; The evolution of arthropod limbs Wills, 1997; Emerson & Schram, 1997 ; Schram & Hof, 1998). Data sets including a large proportion of fossil arthropod taxa tend to emphasise characters relating to gross morphology and tagmosis, partly because the state of preservation, even in many of these Lagerstätten, does not permit the elucidation of important characters relating to structures at the proximal end of the limb and to limb segment homology (Boxshall, 1997). I believe that this focus on tagmosis has diverted much attention from the comparative analysis of limb structure across the major arthropodan taxa, a vital key to understanding arthropodan phylogenetic relationships. This review seeks to identify patterns and trends in the evolution of arthropodan limbs. While I recognise that any identified evolutionary pattern or trend constitutes a phylogenetic statement, new phylogenetic inferences are avoided wherever possible in favour of reference to rigorous analyses available in the literature (e.g. Shultz, 1990; Dunlop & Selden, 1997 ; Wills, 1997; Edgecombe & Ramsköld, 1999 ; Bitsch & Bitsch, 2000 ; Edgecombe et al., 2000; Dunlop & Braddy, 2001). Attention is instead directed at analysis of phenomena such as annulation, the process of subdivision of true segments by the formation of annuli, and oligomerization, the widespread evolutionary trend within groups towards decreasing numbers of limb segments through fusion or, more commonly, through failure to differentiate during development. In addition, various currently illdefined structures, such as endites, gnathobases and epipodites, are closely examined with the intention of improving the quality of inferences made concerning such structures in phylogenetic analyses of arthropods. The term groundplan is used to refer to the putative set of plesiomorphic character states attributed to the ancestral stock of a given lineage. The distinction made here between limb segments and annuli is based on characteristics of their musculature. True segments are characterised by the presence of intrinsic muscles that originate, insert, or attach within each segment. By contrast, annuli lack intrinsic muscle origins, intermediate attachments or insertions, although intrinsic muscles or their tendinous extensions may pass through annuli to a more distal insertion site. Articulations between true segments may fail to be expressed. In such cases the plane of the original articulation can be marked externally by an integumental suture line, and internally by a muscle insertion or by the retention of a transverse tendinous section within a muscle (Boxshall, 1985). Although linked, the processes of limb specialization and tagmosis are separated here in order to facilitate the comparison of limbs between taxa exhibiting different patterns of tagmosis. The consideration of limb specialization independent of tagmosis, which relies in large part on limb specialization for characterization of the different tagmata, can also be justified by the obvious decoupling of body somite-forming processes from those involving limbs. For example, the first three postantennulary limbs of the Cambrian arthropod Sidneyia inexpectans Walcott were interpreted as ‘ cephalic ’ by virtue of the absence of the outer gill branch (Bruton, 1981), even though all these segments are free and each is provided with a tergite which is not incorporated into the cephalon. Thus, limbs can be cephalized 255 without their somites being incorporated into a cephalon. The alternative scenario, in which postantennulary segments are incorporated into a cephalon covered by a unitary cephalic shield but the posteriormost limbs carried on the incorporated segments remain uncephalized, is exemplified by the extant cephalocaridan crustaceans (Sanders, 1963; Hessler, 1964) and by the Cambrian Rehbachiella kinnekullensis Müller (Walossek, 1993) and Dala peilertae Müller (Walossek, 1999). The decoupling of the dorsal expression of segmentation, as indicated by the arrangement of tergites, from the number and arrangement of limbs, has been documented in extant taxa such as the Diplopoda (Blower, 1978 ; Enghoff, Dohle & Blower, 1994) and Notostraca (Fryer, 1988), as well as in fossils such as the xandarellid Cindarella eucalla Chen, Ramsköld, Edgecombe & Zhou according to Ramsköld et al. (1997). II. MATERIAL EXAMINED Anaspides tasmaniae Thomson: unregistered material from Tasmania, in the zoology collections of The Natural History Museum, London. Caenestheriella australis (Loven) : Reg. Nos. 1932.2.25. 182–191, in the zoology collections of The Natural History Museum, London. #Canadaspis perfecta (Walcott): Reg. Nos. USNM 189017, 189018, 213860, 207290 in the collections of U.S. National Museum of Natural History, Washington. #Emeraldella brocki Walcott : Reg. Nos. USNM 136441 in the collections of U.S. National Museum of Natural History, Washington. #Leanchoilia superlata Walcott : Reg. Nos. USNM 250221 in the collections of U.S. National Museum of Natural History, Washington. Mysidopsis slabberi (Van Beneden) : adult males, Reg. Nos. 1923.3.14.1–20, in the zoology collections of The Natural History Museum, London. Pauropus furcifer Silvestri: adult females from the R. S. Bagnall collection in the entomology collections of The Natural History Museum, London. #Rhyniella praecursor Hirst & Maulik : Reg. Nos. In38228, In38230, In38232 in the palaeontology collections of The Natural History Museum, London. #Sidneyia inexpectans Walcott : Reg. Nos. USNM 57494, 139708, 139680, 139720, 250207 in the collections of U.S. National Museum of Natural History, Washington. Speleonectes tulumensis Yager : Reg. Nos. 1992. 296–305, in the zoology collections of The Natural History Museum, London. Tachypleus tridentatus (Leach) : unregistered eggs containing trilobite larvae in the zoology collections of The Natural History Museum, London. III. THE ANTENNULE The anteriormost limb in arthropods goes by a variety of names. In Hexapoda, Trilobita, Chilopoda and Diplopoda Geoff A. Boxshall 256 A B E C F D G Fig. 1. See opposite page for legend. The evolution of arthropod limbs it is most commonly referred to as the antenna, whereas in Crustacea either antennule or first antenna is used according to group or personal preference. The anteriormost limbs of modern Chelicerata are the paired chelicerae which, on the basis of Hox gene expression data (Telford & Thomas, 1998 ; Damen et al., 1998 ; Abzhanov, Popadić & Kaufman, 1999), can be interpreted as positional homologues of the antennules/antennae in these other major taxa (Akam, 2000). I will use antennules for the first (anteriormost) limb pair in all arthropods except chelicerates. Other authors have previously adopted antennules for the anteriormost flagelliform limb in arachnates (e.g. Ramsköld et al., 1997, for the Xandarellida). In some Palaeozoic fossil arthropods of uncertain affinity, such as Leanchoilia superlata, the first limb is simply termed ‘the great appendage ’ (Bruton & Whittington, 1983). Budd (2002) suggested that this ‘ great appendage ’ is derived from a pre-antennulary head segment (i.e. that it is not homologous with the antennules of other Arthropoda) and that it is lost or incorporated into the labrum in all crown-group Arthropoda (his Euarthropoda). This suggestion is based partly on evidence that the antenna of an onychophoran develops anterior to its eye (Eriksson, Tait & Budd, 2003) but this extrapolation is weak given that the relative positioning of these structures in extant Onychophora may be affected by the convergent ventral positioning of the mouth in the Onychophora and Arthropoda. (1 ) The uniramous antennule of the arthropodan groundplan The single axis, or uniramous, antennule composed of true segments, as identified by the presence of intrinsic musculature, is widely distributed amongst major arthropod clades. The antennules of most crustacean groups comprise a single proximo-distal axis only. The biramous antennules of malacostracan crustaceans and of the Remipedia are both exceptions (see Section III.2 below). In crustacean taxa, such as the Cephalocarida, Copepoda and Ostracoda, antennules have a single axis, consist of multiple segments, and have segmentally arranged intrinsic musculature occurring along the entire length of the antennule (Sars, 1922; Hessler, 1964). In copepods a maximum of 27 antennulary segments is found in any extant species, although Huys & Boxshall (1991) demonstrated that 28 different segments are expressed within the Copepoda as a whole. Although multisegmented, copepod antennules are not flagellar in structure ; all segments are provided with intrinsic musculature (e.g. Boxshall, 1985) and should be considered as segments not annuli (Fig. 1 A). All extant copepods exhibit some reduction 257 in antennulary segmentation from the groundplan number of 28, and these reductions are primarily a result of failure of expression of inter-segmental articulations during development (Boxshall & Huys, 1998). A maximum of eight antennulary segments is expressed in podocope ostracods although most recent taxa have fewer (Maddox, 2000). Analysis of developmental patterns indicates that reductions in segmentation in adult ostracods are, as in copepods, primarily a result of failures of expression of inter-segmental articulations rather than secondary fusions (Maddox, 2000). In Hexapoda, Chilopoda, Symphyla and Diplopoda the antennule is uniramous. The number of segments varies enormously. The common pattern of the antennule in pterygote hexapods (=Insecta) is a two-segmented proximal part comprising the scape (the first segment) and pedicel (the second segment), and a distal flagellar part. The scape and pedicel both contain intrinsic muscles, but the entire distal flagellar part is devoid of intrinsic musculature (Imms, 1939). The same type of antennule (lacking intrinsic musculature in the distal flagellar portion) is also found in two groups of apterygote hexapods, the Archaeognatha (e.g. Machiliidae) and the Zygentoma (e.g. Lepismatidae) (Imms, 1939). In insects the pedicel typically houses Johnston’s organ, which can be used as a reference point when establishing segmental homologies in the proximal portion. In the Diplura (non-insectan hexapods) the antennules are composed of true segments, each provided with a complement of intrinsic muscles (Imms, 1939). The four segments of the typical antennule of the Collembola are also provided with intrinsic muscles (Fig. 1 D). In lithobiomorph centipedes the antennule contains welldeveloped intrinsic muscles : both first and second segments contain extensor, flexor, rotator and retractor muscles. The third segment and each distal segment up to the subapical segment houses four intrinsic muscles (Imms, 1939). The Middle Devonian centipede Devonobius delta Shear & Bonamo has antennules comprising at least 14 telescopically articulating segments (Shear & Bonamo, 1988). The antennules of the millipede Cylindroiulus punctatus (Leach) (Diplopoda: Iulidae) are relatively short, comprising only seven or eight segments, but each segment is provided with intrinsic muscles. In the symphylan Scutigerella immaculata (Newport) the distal segments up to and including the subapical segment each house four intrinsic muscles (Imms, 1939). In all these cases it appears that the antennulary units are true segments, defined by the possession of segmentally arranged, intrinsic muscles. Scutigeromorph centipedes can have very high numbers, up to 400 or more, of antennulomeres (Minelli et al., 2000). Fig. 1. (A) Multisegmented, uniramous antennule of a copepod crustacean (Calanoida : Euaugaptilus placitus), with insets showing segmentally arranged intrinsic musculature both proximally and distally. (B) Antennule of Cambrian Sidneyia inexpectans, composite drawn from USNM 139720 and USNM 250207, showing proximal articulations apparently provided with arthrodial membrane (arrowheads). (C) Flagellate antennule of Spinicaudata crustacean (Caenestheriella australis), showing distal annulate flagellum lacking intrinsic musculature. (D) Antennule of collembolan (Collembola : Isotomurus palustris) ; basic four-segmented type showing intrinsic musculature. (E) Flagellate antennule of sminthurid collembolan (Sminthurus viridis) showing modular increase in distal sensory array resulting from annulation. (F ) Antennule of Sminthurus viridis, showing intrinsic musculature. (G) Antennule of collembolan (Orchesella villosa) showing intercalary annuli (arrowed) divided off proximally from first and second true segments. Sources : A adapted from Boxshall (1985) ; B–C original ; D, F–G adapted from Imms (1939) ; E adapted from Nayrolles (1991). Scale bars, B=6.3 mm, C=0.5 mm. Geoff A. Boxshall 258 A B C D F E G H J I Fig. 2. See opposite page for legend. The evolution of arthropod limbs Imms (1939) demonstrated that, in Scutigera longicornis (Fabricius) the antennule comprises a two-segmented peduncle bearing a distal flagellum of 300 or more annuli. This flagellum consists of three regions, each being defined by an intersegmental articulation (referred to as a nodus) at which intrinsic muscles insert, and composed of progressively smaller annuli. Where it differs from the insectan type of flagellum, as described above, is in the musculature : the scutigeromorphan flagellum houses intrinsic muscles which extend almost the entire length of the limb (Imms, 1939) but they are arranged regionally. The proximal region up to the first nodus comprises 54–64 annuli and is traversed by intrinsic levator muscles that originate in the peduncle, and by flexor and extensor muscles which originate in the distal half of the proximal region of the antennule. All insert at the first nodus, on the basal rim of the second region (Fig. 2J). Each of the mid and distal regions is traversed by a similar set of intrinsic muscles. This type of flagellum, with intercalary annulation, is characterised by the presence of intrinsic muscles, which extend to the apex of the limb but do not attach/insert on every annulus. Most of the arthropod taxa represented in the early Palaeozoic, such as the trilobites and the Cambrian arthropods of the Orsten of Sweden, and the Burgess Shale and Chengjiang faunas, have uniramous antennules. The antennules of these forms are often described as annulated and some comprise a large number of ‘ annulations’. Over 120 annuli were shown in Emeraldella brocki (Bruton & Whittington, 1983). Most other species are shown with fewer annulations, for example Branchiocaris pretiosa (Resser) (Briggs, 1976) and Marrella splendens Walcott (Whittington, 1971). Despite the spectacular preservation of these fossils, we still lack detailed information on internal anatomy, such as musculature, and are unable to characterise most of these antennules as comprising either true segments or annuli. The uniramous antennules of the trilobite Triarthrus eatoni (Hall) were described as annulated and Cisne (1975) showed approximately 70 annuli in his reconstruction. Cisne (1975) reported strands of pyrite from the proximal third of this antennule, and interpreted them as representing traces of intrinsic muscles. This interpretation may not be correct, but if it were, I would infer from the presence of muscles in the proximal third (presumably 20+ segments) that the entire antennule of this trilobite was segmented, rather than annulated. A faint thread-like structure preserved within one of the proximal podomeres of Mimetaster hexagonalis (Gürich) may represent a muscle according to Stürmer & 259 Bergström (1976) and may also be interpreted as evidence that at least the proximal part of the antennule of this marrellomorphan arthropod is composed of segments. The proximal segments of the antennule of Sidneyia inexpectans appear to show zones of arthrodial membrane (Fig. 1B) consistent with the interpretation of this antennule as comprising telescopic segments separated by true articulations, as inferred above for the Devonian chilopod Devonobius delta. The spectacularly preserved and informative Cambrian arthropods of the Orsten-type of preservation are preserved in three-dimensions, and can be extracted and observed using scanning electron microscopy (SEM) almost as though they were recent organisms. Some Orsten forms, such as Agnostus pisiformis (Linnaeus) and Cambrocaris baltica Walossek & Szaniawski, have antennules with well-defined podomeres (used here as a neutral term for a subdivision that might be either a segment or an annulus) both proximally and distally, whereas others, such as the branchiopod crustaceans Rehbachiella kinnekullensis, Skara anulata and Bredocaris admirabilis Müller, have a proximal portion described as comprising ‘incomplete annuli’ and a distal portion consisting of cylindrical segments (Fig. 2 A) (Müller & Walossek, 1988; Walossek, 1993). The proximal part of the Agnostus pisiformis antennule is not a peduncle, as demonstrated by its apparent subdivision between meraspid stage 1b and stage 2a (cf. Fig. 11 A, B in Müller & Walossek, 1987). The presence of proximal annuli is possibly indicative of intercalary annulation but the absence of data on musculature makes it impossible unequivocally to interpret these fossils. Proximal annulation is expressed transiently during the naupliar phase of development of some harpacticoid and cyclopoid copepods but is lost by the first copepodid stage (e.g. Dahms, 1992). The protozoeal larvae of penaeid decapods similarly exhibit transient annulations in the proximal part of the antennule (Gurney, 1942), which are lost by the end of the zoeal phase (Fig. 2 B–D). This may be interpreted as additional evidence of the larval status of Rehbachiella kinnekullensis and Bredocaris admirabilis (for other evidence see Walossek, 1993), but may also indicate that a proximal annulated zone is plesiomorphic for the Crustacea. The Devonian lipostracan branchiopod Lepidocaris rhyniensis has reduced, three-segmented antennules (Scourfield, 1926, 1940), but information on musculature is lacking. However, representatives of the fossil Kazacharthra, sister group to the Notostraca (Schram, 1986 ; Walossek, 1993; Wills, 1997), such as Almatium gusevi (Chernyshev), retained up to 15 segments in their antennules at least into the Lower Fig. 2. (A) Antennule of probable developmental stage of Cambrian crustacean (Skara anulata), showing proximal annulate zone. (B) Antennule of first protozoeal stage of decapod crustacean (Penaeoida : Penaeus paulensis Pérez-Farfante), showing proximal annulate zone. (C) Antennule of third protozoeal stage of P. paulensis, showing entire, non-annulate first segment. (D) Antennule of first mysis stage of P. paulensis, showing first appearance of second flagellum. (E) Antennule of generalised adult penaeid decapod. (F) Chelicera of eurypterid (Baltoeurypterus tetragonophthalmus). (G) Biramous antennule of pauropod (Pauropus furcifer), showing intrinsic musculature and annulated setae, which are cut off at level of arrowheads. (H) Biramous antennule of remipede crustacean (Speleonectes tulumensis) with inset showing intrinsic musculature in distal part of dorsal ramus. (I) Same, showing musculature in proximal part of antennule. ( J) Node of scutigeromorph chilopod antennule (Scutigera longicornis), showing intrinsic musculature insertions and origins around node but lack of musculature associated with annuli. Sources : A adapted from Müller & Walossek (1985) ; B–D adapted from de Calazans (1992) ; E adapted from Pérez-Farfante & Kensley (1997) ; F adapted from Selden (1981) ; G–I original ; J adapted from Imms (1939). Scale bars, G=25 mm, I=1 mm. 260 Jurassic (Chen & Zhou, 1985; McKenzie, Chen & Majoran, 1991; Walossek, 1993). Living branchiopods all have reduced, unsegmented or two-segmented antennules in the adult (Martin, 1992), although apparently segmented antennules have been reported from some larval branchiopods (Heath, 1924). The long antennules of the spinicaudatan conchostracan Caenestheriella australis have a distal flagellar part provided with a series of lobes along one margin which gives the limb a segmented appearance but intrinsic musculature extends only into the proximal part (Fig. 1C). I consider that reports of muscles extending the length of the antennule, such as that of Shakoori (1968), are erroneous, Shakoori presumably having misidentified the large nerves present in Caenestheriella as muscles. This type of flagellate antennule is of independent origin within the conchostracan lineage. The presence of multisegmented antennules in the Kazacharthra suggests that antennulary reduction proceeded independently in the Notostraca, the Anostraca and the conchostracan lineage, by inference from the scheme of phylogenetic relationships proposed by Walossek (1993). There is confusion surrounding the identity and form of the antennule in the Burgess Shale arthropod Canadaspis perfecta. Briggs (1978) described a pair of small, narrow projections located between the eyes and either side of the cephalic rostral spine. He carefully stated ‘ The nature of this projection is unknown, but it is assumed to have been a sensory appendage, and is therefore referred to as the ‘ first antenna’.’ The grounds for this assumption were not given. Briggs also stated that there is ‘ no unequivocal evidence that the first antenna was either segmented or spinose ’. The typical uniramous and multisegmented appendage was identified by Briggs as the second antenna. Dahl (1984) challenged this interpretation, identifying the second antenna of Briggs as the real antennule and the ‘first antenna ’ as a possible eye papilla. After examination of C. perfecta type material (USNM 20790, USNM 213860, USNM 189017 and USNM 189018), I agree with Dahl (1984) that Canadaspis has a single pair of antennules, previously referred to as the second antennae by Briggs (1978). Dahl’s (1984) interpretation was validated by Hou & Bergström’s (1997) reconstruction of Canadaspis laevigata Hou & Bergström, which shows a single pair of uniramous and multisegmented antennules. The bivalved arthropod Kunmingella maotianshanensis Hou & Shu has short, four-segmented uniramous antennules (Hou et al., 1996 ; Shu et al., 1999). Kunmingella has been classified as a member of the Bradoriida, but the affinities of this taxon, formerly treated as ostracods, are uncertain. The form of the antennule with its limited number of rod-shaped segments bearing setae, was interpreted as evidence by Shu et al. (1999) for the view of bradoriids as stem-line crustaceans, based on character states postulated as apomorphies of the Crustacea stem-lineage by Walossek (1999). Acceptance of the new Hox gene-based system of homologies of anterior limbs in arthropods requires that the chelicerae are considered as profoundly modified antennules. Chelicerae are short uniramous limbs typically comprising only two or three segments, the distal one of which is the movable distal claw or fang. They are so different from the sensory antennules of mandibulates that few authors Geoff A. Boxshall have ever compared them. In non-arachnid chelicerates, such as the Xiphosura, Eurypterida and Pycnogonida, the three-segmented chelicerae are chelate, with the distal segment forming a movable claw, which typically opposes a fixed finger-like process on the middle segment (Fig. 2 F). This was regarded as the plesiomorphic state for the Arachnida by Shultz (1990), and three-segmented chelicerae are also retained in many arachnids, such as Scorpionida (scorpions), Opiliones (harvestmen) and some Acari (mites). In Araneae (spiders), Amblypygi (tailless whipscorpions) and Uropygi (whipscorpions) the chelicerae are two-segmented and are not chelate. Instead there is a movable distal fang which, in spiders, serves to inject poison into the prey. The segments are each provided with intrinsic muscles ; those of the movable finger (=claw) typically have pinnate fibres inserting on an apodeme (Palmgren, 1978), an arrangement which permits contraction within a confined space (Alexander, 1968). The state of the anteriormost limbs in possible stem-group chelicerates is of considerable phylogenetic interest. The Aglaspida, in particular the Upper Cambrian Aglaspis spinifer Raasch, have been treated as basal chelicerates (Størmer, 1944; Eldredge, 1974), or as sister group to the other chelicerates (Bergström, 1979 ; Dunlop & Selden, 1997). The first pair of limbs, originally described as chelate by Raasch (1939), was re-interpreted by Briggs, Bruton & Whittington (1979) who found the first limbs to be uniramous, ‘probably cylindrical and possibly jointed ’. They could not confirm that the limb was chelate. The short uniramous antennule, as found in Aglaspis spinifer, represents a possible intermediate condition between the long uniramous limb that was widespread in Cambrian arthropods and the true chelicera comprising only about three segments. Another important fossil from the chelicerate lineage which offers insight into the state of the anteriormost limbs at this early stage in the origin of the chelicerate clade, is the Silurian Offacolus kingi Orr, Siveter, Briggs, Siveter & Sutton (Orr et al., 2000 ; Sutton et al., 2002). The anteroventrally projecting first limbs of Offacolus kingi are described as uniramous, possibly three-segmented and distally chelate (Sutton et al., 2002). They are small and not easily visible even in three-dimensional reconstructions. The Middle Cambrian Sanctacaris uncata Briggs & Collins was originally treated as the sister group to all chelicerates by Briggs & Collins (1988), but in subsequent analyses Sanctacaris uncata has been identified as an arachnomorphan rather than as sister group to the chelicerates (Briggs, Fortey & Wills, 1992; Wills, 1996; Selden & Dunlop, 1998 ; Sutton et al., 2002). The first limb of Sanctacaris uncata, as identified by Briggs & Collins (1988), is biramous, consisting of a short, raptorial inner ramus comprising perhaps four segments, plus an outer ramus of cylindrical segments. An alternative interpretation of the head limbs of Sanctacaris uncata is possible. Comparison with the prosoma of Offacolus kingi, in which the chelicerae are easily overlooked, suggests to me that the chelicerae may not be discernible in the available Sanctacaris uncata material. Based on their photographs, I consider that the raptorial limbs numbered 1–5 by Briggs & Collins (1988), are the telopodites of the second to sixth limbs, respectively. Each of these limbs carries a slender The evolution of arthropod limbs exopodite, referred to as the antenna-like ramus by Briggs & Collins (1988). No early Palaeozoic arthropods have a biramous antennule/chelicera combining a short telopodite with a flagelliform exopodite. The structure labelled limb ‘ 6’ in their Text-Fig. 1 A, B is here re-interpreted as the seventh and last prosomal limb. The limb identified by Briggs & Collins (1988) as a biramous first prosomal limb of Sanctacaris uncata is here re-interpreted as the second limb (i.e. the first postantennulary limb). Another Burgess Shale arthropod, Yohoia tenuis Walcott, has antennules referred to as ‘ great appendages ’ by Whittington (1974). Whittington found no evidence for the three-segmented antennule shown in an earlier reconstruction of Yohoia tenuis by Simonetta (1970). The antennules of Yohoia tenuis are short and carry four movable ‘ spines ’ apically. These short, uniramous antennules can be compared with other antennules as there are similarities to the great appendages of several other Cambrian species, such as Leanchoilia (Bergström & Hou, 1998; and see below). Cambropycnon klausmuelleri Waloszek & Dunlop from the Orsten of Sweden, described as Larva D by Müller & Walossek (1986 a), offers a tantalising glimpse of the variety of anterior limb structure already apparent by the Upper Cambrian. The paired structures interpreted as appendage I by Müller & Walossek (1986a) and as ‘ possible antenna ’ (a1 ?) by Waloszek & Dunlop (2002) lack an articulation with the cephalon, are unsegmented, and are so tiny (3 mm at their widest point) that there is no likelihood of any associated musculature having been present. In the absence of any evidence I do not accept the assertion of Waloszek & Dunlop (2002) that this frontal structure represents the antennule. The chelate limb (labelled appendage II or chelicerae, respectively) represents the anteriormost limb, or antennule, of this animal. I agree with the inference made by Walossek & Müller (1997), that C. klausmuelleri is a crowngroup chelicerate, as diagnosed by the possession of chelicerae. The classical assumption that antennules were lost in the chelicerate lineage was widely accepted before the latest findings on Hox gene expression patterns (Telford & Thomas, 1998; Damen et al., 1998 ; Abzhanov & Kaufman, 2000) suggested that chelicerae are modified antennules. (2 ) Biramous or polyramous antennules Biramous antennules are typical of adult malacostracan crustaceans. The antennule of malacostracans has a robust muscular basal part, the peduncle, which may be two, three or four segmented (Fig. 2 D, E) and usually bears two long, multi-annulate flagellar branches. The malacostracan antennules thus far investigated are all characterised by the lack of intrinsic musculature in the flagellar parts. The peduncular segments are provided with powerful intrinsic muscles, the most distal of which are extensors and flexors, which insert around the basal regions of both flagella and serve to move each flagellum as a whole. Three flagellar branches are present in the Stomatopoda and in certain caridean decapods (Calman, 1909). Current phylogenetic schemes lead to the inference that the triflagellate state evolved independently in stomatopods and in some caridean decapods. In the male of the mysid Mesopodopsis slabberi the 261 antennulary peduncle carries four distal structures : the two typical flagella, an appendix masculina and an accessory appendix (see Fig. 108 A in Tattersall & Tattersall, 1951). The discovery of the Remipedia (Yager, 1981) generated much discussion concerning the uniramous or biramous status of the antennule in the crustacean groundplan. The antennule of Speleonectes tulumensis was described as comprising a two-segmented peduncle, a dorsal ramus of 11 slender segments and a ventral ramus of nine to ten segments (Yager, 1987 ; Emerson & Schram, 1991). Examination of the musculature of the antennule of this species reveals an important difference between the dorsal and ventral branches (Fig. 2 H, I). The segments of the main, dorsal branch are each provided with antagonistic intrinsic muscles, even though the articulations between segments are neither well defined, nor provided with extensive arthrodial membrane, as is typical for antennulary segments in copepods, for example. By contrast, the ventral branch lacks intrinsic musculature. It is moved as a whole by muscles originating within the peduncle and inserting on its basal rim. Its units represent poorly defined annuli, not separated by arthrodial membrane, and it is here interpreted as flagellar in structure. Based on these new data, the remipede antennule can be regarded as fundamentally uniramous and carrying a secondary ventral flagellum. The flagellum can be reduced, as in the remipede family Godzilliidae (Schram, Yager & Emerson, 1986). The antennules of the Pauropoda are also biramous. Silvestri (1902) first elucidated the musculature of the pauropod antennule and his observations are here confirmed for Pauropus furcifer (Fig. 2 G). The antennule comprises a four-segmented peduncle and two unisegmented branches, referred to as the superior and inferior rami. Each peduncular segment is characterised by the insertion of muscles and there are intrinsic muscles, antagonistic pairs of which insert on the basal rim of each ramus. However, in all pauropods the rami are only unisegmented. It is interesting to note that the rami carry a total of three long, annulated, sensory elements [often referred to as flagella, e.g. by Bagnall (1911)]. These probably serve as passively flexible (by virtue of their annulation) sensory arrays equivalent to flagella and, given the physical scale at which these minute arthropods operate, the pauropod antennules can be considered as effectively triflagellate. In stomatopods with triflagellate antennules, only one of the three flagella carries aesthetascs (Mead & Koehl, 2000). The antennule (or ‘great appendage ’) of the Burgess Shale arthropod Leanchoilia superlata is described as comprising four podomeres. Podomeres 2 and 3 are each produced into a rigid shaft which extends distally into a longer, annulated, flexible portion ; podomere 4 consists of a tapering rigid shaft bearing three apical claws and a similar flexible distal portion (Bruton & Whittington, 1983). Examination of new material of Alalcomenaeus cambricus Simonetta from the Burgess Shale (Briggs & Collins, 1999) has revealed that the antennule of A. cambricus is very similar in basic structure to that of Leanchoilia superlata. This antennule is effectively triflagellate, as in the pauropod P. furcifer, but it is difficult to categorise the ‘ flagellae’. Bruton & Whittington (1983) and Briggs & Collins (1999) described them as annulated 262 I V X XV XX XXV Col ColI ColII CoIV CoV F Fig. 3. See opposite page for legend. Geoff A. Boxshall M The evolution of arthropod limbs or flagellum-like in structure and their interpretations are adopted here. The flagellar structures may be modified setal elements, as in pauropods, despite the very much larger body size and aquatic habitat of Leanchoilia superlata. The antennule morphology exhibited by the rare Burgess Shale arthropod Actaeus armatus Simonetta is also similar to that of Leanchoilia superlata. The limb is short, comprising few segments, bears two subapical flagellum-like structures, and is armed with distal claws (Whittington, 1981). In the Chengjiang arthropod Fortiforceps protensa Hou & Bergström the identification of the antennule is somewhat equivocal (Hou & Bergström, 1997). In most specimens the large grasping appendages were the anteriormost limbs found. Given other similarities to Actaeus armatus and the Leanchoilia-Alalcomenaeus lineage, it seems possible that the socalled antennule of Fortiforceps protensa is an artefact and that the true antennule is the great appendage interpreted as the first postantennulary limb by Hou & Bergström (1997). It is well established that the biramous state is apomorphic for the malacostracan antennule. In part this inference is based upon developmental data : in malacostracans such as mysids (Manton, 1928) and leptostracans (Manton, 1934 ; Olesen & Walossek, 2000) the antennule first appears as a simple lobe. The second ramus does not appear until later in development. In addition, biramous antennules of the malacostracan type do not appear in the fossil record until the Late Palaeozoic (Carboniferous). The vast majority of early Palaeozoic arthropods and the set of stem-lineage crustaceans identified by Walossek & Müller (1990) also exhibit a uniramous antennule. The recently discovered Cambronatus brasseli Briggs & Bartels from the Devonian is described as having biramous antennules (Briggs & Bartels, 2001). However, the bases of these two branches were not well preserved and both branches are shown with well-developed proximal peduncles, rather than arising from a common peduncle as in extant malacostracans. I interpret this as evidence that these branches belong to different appendages. I consider that the longer, more anteriorly located branch, with four peduncular segments, is a uniramous antennule and that the shorter branch with two peduncular segments may be the exopodite of a biramous first postantennulary limb, the telopodite of which was labelled lh2 in their Text-Fig. 2 by Briggs & Bartels (2001). The head limbs of Cambronatus brasseli are reinterpreted herein (see Fig. 4D ). The crustacean affinities of this fossil are equivocal. A biramous or polyramous condition of the antennule has arisen independently at least three times : twice within the Crustacea (in malacostracans and in remipedes), and once 263 in the Pauropoda. Given current estimates of phylogenetic relationships within the Malacostraca (e.g. Wills, 1997; Richter & Scholtz, 2001), the presence of three flagellae in the Stomatopoda and within the caridean Decapoda is indicative of an independent origin of the third flagellum in these taxa. Experimentally, it is possible to induce a supernumerary proximal-distal axis in arthropod limbs by producing an ectopic intersection point between the cells expressing wingless (Wg) and decapentaplegic (dpp) (Lecuit & Cohen, 1997). The evolution of supernumerary (second and third) flagellae in stem-line malacostracans presumably involved such genetic mechanisms. (3 ) Loss of antennules Members of the Protura lack any trace of antennules and it is widely inferred that the lack of antennules is derived. The detailed study of cephalic anatomy in proturans by François (1969) found no evidence to support the suggestion, originally made by Berlese (1908), that the enigmatic pseudoculus was derived from the antennule. (4 ) Antennulary development and the origin of flagella The most complex uniramous antennule for which good developmental data are available is that of copepod crustaceans. The groundplan development pattern from nauplius to adult female was modelled by Boxshall & Huys (1998), for a hypothetical ancestral copepod (Fig. 3). The metamorphic moult from the sixth and final nauplius stage to the first copepodid stage was marked by the subdivision of the third (=apical) segment of the nauplius to form the distal eight segments of the adult antennule (segments XXI–XXVIII). No further divisions occur in this distal section of the antennule throughout the five copepodid-phase moults to definitive adult. During the copepodid phase the two proximal antennulary segments of the nauplius undergo a sequence of subdivisions to form segments I–XX of the adult. In the adult, muscle bundles originating in the proximal segment extend the entire length of the antennule, inserting on the apical segment but attaching to every segment in between (Boxshall, 1985). I assume that the intermediate muscle attachments are formed at the time of first expression of the particular articulation, but this requires confirmation. The earliest phase of development is the formation of a simple unsegmented lobe in the embryo. Subsequently there is a process of subdivision giving rise to segments that are expressed both externally and internally, by the formation of Fig. 3. Schematic diagram showing development of antennulary segmentation and setation through the copepodid stages of calanoid copepods (from Boxshall & Huys, 1998). The scale at the top indicates the presumed 28 (I–XXVIII) segments of the ancestral copepod (see Huys & Boxshall, 1991). Elements are shown as setae or aesthetascs. A segment carrying a seta not present at the preceding stage is shown in red if the newly added seta is the anterodistal member of a trithek, in green if it is the anteroproximal member. The basal segment (I) is shown in yellow when its second and third setae first appear. Aesthetascs are shown in yellow when they first appear, thereafter in black. Compound segments derived by secondary fusion in adult males are in lilac. The plane of the XX to XXI articulation is indicated by a vertical dotted line. The neocopepodan geniculation of the adult male is indicated by the arrow. Modified setae either side of the geniculation are shown as asterisks. Abbreviations : CoI to CoV=copepodid stages I–V, F=adult female, M=adult male. Geoff A. Boxshall 264 A B ex t ? rh4 rh3 ex rh2t rh5 t rh2ex C Ih2ex Ih2t ex Ih5 Ih3 Ih4 t Fig. 4. See opposite page for legend. The evolution of arthropod limbs articulations provided with arthrodial membrane and by the presence of intermediate muscle attachments. This generalised pattern is the likely ancestral arthropod pattern. Interestingly, what drives segmentation is the sequential appearance of articulations and, in terms of the ontogenetic events, the segments themselves are essentially a by-product of this process. Antennules with fewer expressed segments can readily be generated by restriction or by early cessation of the process of subdivision (Boxshall & Huys, 1998; Schutze, Da Rocha & Boxshall, 2000). Indeed heterochronic change in segmentforming processes has been interpreted as a potential evolutionary mechanism in arthropods in general (e.g. Minelli & Fusco, 1995 ; Dunlop, 1998 ; Schutze et al., 2000). Antennules with a larger number of true segments can be generated by extending this process of subdivision. The formation of antennulary flagella by annulation, is a different process. An interesting model taxon for studying the formation of distal flagellar portions is the Collembola. Collembolans are non-insectan hexapods and typically have four-segmented antennules. The oldest fossil hexapod, Rhyniella praecursor from the Lower Middle Devonian, is a collembolan which has subsequently been placed in the family Isotomidae (Greenslade & Whalley, 1986). It also has four-segmented antennules. The antennulary musculature of another isotomid, Isotomurus palustris (Müller), was figured by Imms (1939) who showed intrinsic muscles inserting on segments 2–4 (Fig. 1 D). Imms (1939) also showed that the intrinsic musculature is unaffected by the annulation of the distal segment exhibited in sminthurid collembolans (Fig. 1 E, F) (cf. Nayrolles, 1991). This illustrates what is regarded here as the typical formation of a distal flagellar portion – where annular subdivision of the terminal segment leads to the formation of a long flagellum that does not affect the intrinsic musculature that inserts on the proximal rim of the flagellum. In some insects, the Dermaptera for example, flagellum development proceeds by the sequential formation of additional annuli during the larval phase (Davies, 1966). The scape and pedicel do not divide during the larval phase. The distal flagellar portion comprises an apical region of five annuli, which remain unchanged through the larval phase, and a middle region. During the larval phase, annuli (referred to as meristal annuli) are added within the middle region of the antennule between the pedicel and the region of apical annuli. This is interpreted as a secondary elaboration of the process of flagellum formation in which, as a result of heterochronic timing changes, some annuli (the apical annuli) appear early in a pre-formed block and others (the meristal annuli) appear later at successive moults. The proximal-most meristal annulus (i.e. that immediately distal to the pedicel) presumably retains, unchanged, the 265 insertions of the flagellar muscles, originating in the pedicel, on its basal rim throughout this process. Within the Collembola, the first and second segments of the basic four-segmented limb can each subdivide to form two annuli (as in Orchesella villosa Geoffroy). No muscles insert or attach around the plane of these annulations (Imms, 1939) and they are treated here as derived annuli, rather than segments (Fig. 1G). They serve to illustrate the minimal or baseline condition for non-terminal or intercalary annulation, since in each case only a single annulus is added. The long filiform antennules of scutigeromorph centipedes may represent an extreme case of intercalary annulation, in which nodal flagellar regions contain intrinsic muscles but these originate and insert only around each nodus (Fig. 2 J). It is inferred here that the nodal points represent original segmental articulations and that the subdivisions within each nodal region are annuli, which are not reflected in the muscle signature. Based on the developmental data presented by Andersson (1979), it is apparent that this type of antennule can be derived from a five-segmented precursor by annulation of the middle three segments (see Minelli et al., 2000). The musculature of this type of antennule can be easily derived from the basic segmental arrangement. (5 ) Function Consideration of function should inform interpretation of morphology but its value in aiding the determination of homology is not always obvious. Many arthropods possess elaborate compound eyes and vision is obviously an important sensory mode in these taxa, but for the majority of crustaceans, hexapods and myriapods it is the antennules that form the primary sensory interface with the environment. These antennules can be provided with rich arrays of mechanosensory, chemosensory or bimodal receptors that play a role in almost every aspect of behaviour of the animals. There are numerous examples of enhancement of the sensory array, by increasing antennulary length or density of receptors (Boxshall & Huys, 1998), or by the addition of mutliple, modular sets of sensory elements via annulation. Given its importance as a sensory interface, the secondary loss of the antennules in the proturans is counter-intuitive, especially as it appears that the first walking legs are functionally co-opted to serve as ‘ replacement ’ sensory limbs in this group. Similarly, the functional role of the extremely reduced antennules in extant notostracan crustaceans appears to be performed by the elongate, flagelliform distal endites of the first postcephalic trunk limb. The hallmark of the arthropodan limb is its multifunctional potential. Other uses for the primarily sensory antennules include : grasping the female during mating (in most male copepod crustaceans) ; forming the cementing Fig. 4. (A) Trunk limb of Sidneyia inexpectans. (B) Trunk limb of Rehbachiella kinnekullensis. (C) Second opisthosomal limb of trilobite larva of xiphosuran Tachypleus tridentatus. (D) Re-interpretation of cephalic limbs of Cambronatus brasseli based on tracing from photograph in Briggs & Bartels (2001). Sources : A original ; B adapted from Walossek (1993) ; C original ; D original drawing. Abbreviations : ex=exopodite, t=telopodite, rh2–5=right side cephalic limbs 2–5, lh2–5=left side cephalic limbs 2–5, ? indicates uncertainty over precise origin of exopodite. Scale bars, A=5 mm, C=1 mm. 266 apparatus for attaching the animal to the substratum during settlement (in the cypris larvae of cirripedes) ; and forming the clawed attachment mechanism for holding onto the host (in parasitic branchiuran fish lice). In the unclassified Cambrian arthropod Yohoia tenuis Walcott the short antennules also appear adapted for grasping. In chelicerates the antennules are modified as chelicerae which are primarily involved in feeding, although they may also play an important role in grooming. Their sensory role is much reduced. In various chelicerate groups other limbs have been co-opted to serve a sensory role, such as the pedipalps in scorpions (which are richly supplied with trichobothria) and the ‘antenniform’ first walking legs of Amblypygi. Any functional analysis within the Arthropoda must be interpreted with caution. If nothing else, the arthropods testify to the possibility of alternative solutions to biological problems. (6 ) The antennule of the arthropodan groundplan The antennule of the arthropodan stem species is here inferred to be uniramous and segmented, i.e. composed of a small number of segments, each of which is characterised by the insertion or attachment of intrinsic muscles. Flagellate antennules consisting of annuli which lack intrinsic musculature, are derived. At least two different types of flagellate antennules are found within recent arthropods : those with terminal flagella formed by annulation of the distal segment, as for example in some collembolans, ectognathan insects and malacostracan crustaceans, and those with nonterminal or intercalary annulations, such as scutigeromorph centipedes and some collembolans. Terminal flagella are by far the most common. It is possible that an antennule may be doubly flagellate, with both intercalary and terminal annulations. The anteriormost limb of modern chelicerates is composed of at most three segments. Except in chelicerates and possibly Actaeus, where they are primarily involved in feeding, the anteriormost limbs of arthropods are primarily sensory although they can fulfil a number of other functions in particular arthropodan taxa. Consideration of pivotal fossil taxa with chelicerate affinities, such as Aglaspis spinifer, suggests that chelicerae were derived from short, uniramous limbs consisting of a few (perhaps four) cylindrical segments, which were primarily sensory in function. IV. THE POSTANTENNULARY LIMBS – HOMONOMOUS OR HETERONOMOUS SERIES Trilobites, as exemplified by taxa with soft part preservation such as Phacops ferdinandi (Kayser) from the Devonian Hunsrück Slate, typically possess uniramous antennules followed by an homonomous series of postantennulary limbs, that does not reflect body tagmatization into head, thorax and pygidium (Bruton & Haas, 1999). Limbs are unknown for the vast majority of trilobites but a similar pattern can be found in trilobites from Lagerstätten of other periods, for example in Triarthrus eatoni from the Ordovician (Cisne, 1975; Whittington & Almond, 1987) and in Olenoides serratus Geoff A. Boxshall (Rominger) from the Middle Cambrian (Whittington, 1980). Extrapolating from the trilobite condition, I here present evidence in support of the long-established hypothesis that all postantennulary limbs in the arthropodan groundplan were similar in basic structure and formed an homonomous series. This hypothesis implies that the arthropodan groundplan includes only two limb types : uniramous multisegmented antennules and a basic biramous postantennulary limb. Where is the evidence for this ? Several Cambrian non-trilobitan arthropods possess a long series of undifferentiated, homonomous, postantennulary limbs. The xandarellid Cindarella eucalla Chen, Ramsköld, Edgecome & Zhou, for example, shows no evidence of significant differentiation (=one facet of tagmosis) in postantennulary limbs (Ramsköld et al., 1997). Outside of the trilobites and their possible arachnate relatives, such as the Xandarellida, the Cambrian bivalved arthropod Isoxys Walcott has been reconstructed as possessing short uniramous antennules followed by a series of 14 similar biramous limbs (Vannier & Chen, 2000). These postantennulary limbs were specialised, having a flattened exopodite and a reduced telopodite, but all 14 pairs exhibited the same basic pattern. Even more specialised, in the apparent absence of any telopodite, is the enigmatic Cambrian arthropod Sarotrocercus oblita Whittington which also presents an homonomous postantennulary series of lobed appendages with marginal lamellae (Whittington, 1981). Burgessia bella Walcott shows incipient differentiation at both anterior and posterior ends of the postantennulary limb series (Hughes, 1975). In the anterior three pairs of postantennulary limbs, the outer branch is flagelliform rather than gill filament-like, but this is easily derived from a biramous trunk limb-type. The posteriormost limb is uniramous and spiniform, and is here interpreted as an autapomorphy. The postantennulary limbs of Sidneyia inexpectans show further differentiation into cephalic and trunk limbs, in the form of the missing gill-branch in the cephalic limbs (Bruton, 1981). However, apart from this loss, the postantennulary limbs form a basically homonomous series in both Burgessia bella and Sidneyia inexpectans. Crustaceans do not appear to conform to this scheme of having uniramous antennules followed by an homonomous series of postantennulary limbs, since crown-group crustaceans exhibit heteronomy of the postantennulary limbs. The basic pattern in the crown-group is a differentiation between the naupliar limbs and the postnaupliar limbs. The naupliar limb series comprises the uniramous antennules plus the biramous antennae and mandibles. The possession of just these three pairs of functional limbs defines the nauplius larva of Crustacea but is not unique within the Arthropoda since other aquatic arthropodan larvae, such as the protonymphon larva of pycnogonids and possibly the phaselus larva of trilobites (Fortey & Morris, 1978), also have three pairs of limbs. Antennae and mandibles exhibit an amazing diversity of form within the Crustacea but their basic similarity is attested to by the close resemblance between antennae and mandibles during the naupliar phase in copepods and thecostracans (e.g. Boxshall & Böttger-Schnack, 1988), in adult mystacocarids and in larval cephalocarids (Sanders, 1963). Both these limbs have multisegmented exopodites The evolution of arthropod limbs provided with segmentally-arranged setae along inner and distal margins. Both have coxal gnathobases, although that of the antenna is transient, being lost by the later naupliar stages, and that of the mandible might not yet be present in the earliest stages. There is a substantial body of evidence showing that the two pairs of postantennulary naupliar limbs in crustaceans have a common basic structure ( Walossek, 1993). All postnaupliar limbs of crustaceans are plesiomorphically biramous and similar in structure. The Cephalocarida provides a robust example, in which maxillules, maxillae and trunk limbs all share a common biramous plan, which differs from that of the postantennulary naupliar limb type in its exopodal segmentation and in the disposition of setae around the outer and distal margins of the exopodite. There is no differentiation in limb structure to coincide with the head-thorax tagma boundary in cephalocarids, or in the Palaeozoic branchiopods Rehbachiella kinnekullensis, Bredocaris admirabilis and Lepidocaris rhyniensis, or in the possible maxillopodan Dala peilertae (Walossek & Müller, 1998). Crown-group Crustacea are, therefore, characterised by a heteronomous postantennulary limb series, with differentiation between naupliar and postnaupliar limbs (Walossek, 1993, 1999), but what of the stem-group crustaceans? The hypothetical urcrustacean of Hessler & Newman (1975) is considered here as equivalent to a stem-group crustacean. Hessler & Newman (1975) attributed a high degree of serial homology in postantennal limb morphology to their urcrustacean, but refrained from conjecture concerning the degree of specialisation of the mandibles and first maxillae. Their figured urcrustacean displays little heteronomy in the postantennulary limb series. The crustacean affinities of Martinssonia elongata Müller & Walossek were recognised by Müller & Walossek (1986 b) and this taxon has now been attributed to the stem-lineage of the Crustacea (Walossek & Müller, 1990 ; Walossek, 1999). The largest specimens known might still be larval and the condition of the two posterior limb pairs might not be definitive, but the first four pairs of postantennulary limbs are homonomous. Ercaia minuscula, a possible crustacean stem-line derivative according to Chen et al. (2001), also shows a modified first postantennulary limb, but the remaining cephalic and trunk limbs are homonomous. Available data on Cambrocaris baltica Walossek & Szaniawski, another stem-lineage crustacean, point to a similar pattern for at least the first four pairs of postantennulary limbs, and possibly more since the exopodites are unknown on the more posteriorly placed limbs (Walossek & Szaniawski, 1991). The limbs are biramous and form an undifferentiated postantennulary limb series. The acquisition of naupliar/postnaupliar limb differentiation is interpreted as a synapomorphy of crown-group Crustacea (Walossek, 1993), and is derived with respect to the homonomous series here attributed to both the arthropodan groundplan, and to stem-lineage crustaceans. The characteristics of the Crustacea sensu lato (stem-lineage plus crown group) do not therefore contradict the hypothesis of an homonomous postantennulary limb series in the arthropodan groundplan. At least two of the great lineages of Arthropoda present in the Cambrian (trilobites and crustaceans) exhibit a basically 267 homonomous postantennulary limb series. The presence of an homonomous postantennulary limb series in the arthropodan groundplan is in accord with Hessler & Newman’s (1975) statement that ‘a high degree of serial homology in metameric animals is a primitive trait ’. Only two basic limb types existed in crown-group arthropods : the uniramous multisegmented antennule and the biramous, gnathobasic postantennulary limb. All of the different limb types found across the spectacular diversity of the Arthropoda would, therefore, be derived from one or other of these two basic limbs. This distinction between antennules and postantennulary limbs seems to be contradicted by the existence of homoeotic mutations, such as Antennapedia (Antp) in Drosophila, in which antennules are transformed into thoracic walking legs (cf. Struhl, 1981). Our understanding of such homoeotic mutations is that they transform the identities of segments and their associated structures into those of other segments (Carroll, Grenier & Weatherbee, 2001). By such means Antp mutations in Drosophila cause the transformation of antennules into mesothoracic (second) legs, which are thus inferred to be serial homologues (e.g. Carroll et al., 2001). Casares & Mann (1998) showed that the dominant Antp acts to suppress the function of the antennule-determining genes, extradenticle (exd ) and homothorax (hth). Casares & Mann (1998) also found that hthx mutants show the same transformation. They concluded that the leg pathway is the ground state for the limbs (ventral appendages) : Antp simply represses hth in the limb primordia, thereby blocking the development of an antennule and allowing the development of the ground-state leg by default. The key question is, can antennules and thoracic (or any postantennulary) legs be serial homologues when they are derived from different ancestral limb types ? The answer is dependent upon the level of inclusiveness. On the basis of the evidence presented in this review, I do not consider antennules and postantennulary limbs to be serial homologues as limb types at the level of the phylum Arthropoda. At a wider (pan-arthropodan) level, including the lobopodians such as the Onychophora, characterised by the possession of an homonomous series of lobopod limbs, antennules and postantennulary limbs can be considered as serial homologues. The involvement of these genes in limb specification must, therefore, pre-date the origin of the basic arthropod pattern proposed here. This inference is in accord with the conclusion of Grenier et al. (1997) who postulated that the arthropodan Hox gene set predated the origin of the entire onychophoran/arthropodan clade. I regard the conclusion of Casares & Mann (1998), that their results ‘ support the view that the ancestral insect possessed leg-like appendages in head and thoracic segments ’ as incorrect. V. THE POSTANTENNULARY LIMB OF THE ARTHROPODAN GROUNDPLAN The basic postantennulary limb of arthropods is widely regarded to be biramous (cf. Walossek, 1993 ; Hou & Bergström, 1997), comprising a proximal part bearing inner and outer branches. The need for a consistent terminology Geoff A. Boxshall 268 to facilitate comparison between major arthropodan taxa necessitates a limited foray into the minefield of arthropodan limb terminology. In developmental literature inner is referred to as ventral, and outer as dorsal. Irrespective of any subdivision into segments the proximal part of the postantennulary limb bearing the rami is collectively referred to in crustaceans as the protopod or protopodite (e.g. Huxley, 1877, 1880), the sympod or sympodite (Calman, 1909; Hansen, 1925; Heegaard, 1945), or more recently, the corm (Fryer, 1988), the basipodite or basis (Milne Edwards, 1851; Walossek, 1993; Walossek & Müller, 1997). In insects it has been called the coxopodite (Snodgrass, 1935). In trilobites, it has traditionally been called the coxopodite, although Ramsköld & Edgecombe (1996) and Ramsköld et al. (1997) adopted the term basis, accepting the identification of the crustacean coxa as a novel structure derived from the mobile proximal endite (Walossek, 1993; Walossek & Müller, 1997). This view is rejected here and the term protopodite is used for the proximal stem of the limb that bears the rami. Milne Edwards (1851) introduced the terms coxopodite and basipodite for the segments of the subdivided protopodite within the Crustacea. It is widely accepted that the inner branch, walking leg, or telopodite of trilobites is the homologue of the endopodite of the typical biramous limb of crustaceans and agnostids, and of the part distal to the protopodite of the single remaining limb axis in hexapods (Snodgrass, 1935), myriapods and chelicerates, including pycnogonids. The trunk limb of Cambrian and Devonian arthropods of uncertain affinity, such as Burgessia bella, Sidneyia inexpectans, Leanchoilia superlata, Mimetaster hexagonalis (Gürich) and Emeraldella brocki, also has an inner, walking branch which can be referred to as the telopodite or endopodite. To facilitate comparison, the term telopodite is adopted here, since it is widely used in developmental literature. The terminology applied to the outer branch of the basic biramous arthropodan limb tends to be more explicitly hypothesis-dependent. Terms such as gill branch are best avoided as too function-dependent, given the propensity to multifunctionality of arthropodan limbs. The outer branch is here referred to as the exopodite, based partly on the recognition (see below) that epipodites were not a feature of the arthropodan groundplan and that the outer branch of the trilobite limb is the homologue of the crustacean exopodite. The two rami are the major limb branches or axes, carried distally on the protopodite. They are plesiomorphically supplied with intrinsic muscles originating within the protopodite and are commonly segmented whereas other leg structures, such as epipodites, are never segmented. The terms protopodite, telopodite and exopodite are adopted here for the proximal stem and distal rami of the biramous postantennulary limb hypothesised here for the arthropodan groundplan. A spinose or setose inner (=ventral) lobe carried on a limb segment is referred to as an endite and any outer (=dorsal) lobe, irrespective of its segment of origin, is referred to as an exite and may be articulated at the base. The exopodite is not an exite – it is a ramus. A powerfully developed endite occupying the entire medial margin of the protopodite, or the entire margin of the proximal protopodal segment if the protopodite is segmented, is referred to as a gnathobase. These two types of gnathobases, found in arachnomorphs and mandibulates respectively, are not homologous. The term epipodite is used specifically for an outer lobe or process originating on the precoxal or coxal segments of the protopodite, in accord with current practice in the Crustacea. The term palp is used for a variety of non-homologous structures across the Arthropoda. ( 1) Protopodite ( a ) Defining the protopodite The protopodite is easy to recognise in biramous limbs, as found in trilobites, the chelicerate Offacolus kingi, crustaceans and some other fossils of uncertain affinity, such as Agnostus pisiformis, because, by definition, it is that proximal part of the biramous limb which carries the rami. The ambiguities arise in taxa with uniramous limbs, since in order to establish homologies, it is necessary to identify the boundary between the protopodite and the remaining ramus, usually, but not always, the telopodite. A uniramous limb such as the maxilliped of copepods may comprise up to nine segments (Huys & Boxshall, 1991), three of which (precoxa, coxa and basis) are traditionally regarded as protopodal in origin, carrying a six-segmented telopodite. But how do we objectively distinguish between protopodal and telopodal segments and what are the criteria that will facilitate this process? In addition to making direct or indirect anatomical comparisons with related taxa retaining both rami, possible criteria might include the presence of endites, the pattern of expression of particular genes along the proximo-distal axis of the limb, the timing of appearance of features during development, or the musculature signature pattern. The presence of endites is characteristic of protopodal segments. In trilobites the inner margin of the undivided protopodite is convex and bears spines. The two features, convexity and presence of spines, appear to characterise a typical endite and the occupation of the entire medial margin of the protopodite defines it as a gnathobase. The functional involvement of the gnathobase or proximal protopodal endite of the postantennulary limb in feeding is almost certainly plesiomorphic to the Arthropoda (e.g. Manton, 1964 ; Whittington, 1971; Bruton, 1981 ; Walossek, 1993), but it is hard to include function within the definition since structures are frequently conscripted to serve other functions. In addition, endites may be transient features of particular limbs. For example, the enditic process on the coxa of the crustacean antenna (see Fig. 8 A) is lost during the naupliar phase of development (e.g. Fryer, 1988), and there are numerous examples of secondary loss of endites from other limbs across the arthropodan groups. Large, undivided endites (gnathobases) are typically retained on the protopodite in all postantennulary limbs of trilobites, in the six prosomal limbs of recent xiphosurans, in the pedipalps of spiders and other arachnids, in the first and second walking legs of scorpions and some harvestmen, and in the third and fourth walking legs of some ischyropsalidoid harvestmen (Gruber, 1978). The retention of protopodal endites in these taxa was a plesiomorphic condition in the The evolution of arthropod limbs analysis of Shultz (1990). No traces of any endites are retained on any limbs in extant pycnogonids, but in the Cambrian Cambropycnon klausmuelleri the two post-cheliceral appendages retain gnathobases on undivided protopodites (Waloszek & Dunlop, 2002). Endites or gnathobases are also retained on protopodal segments in many mandibulate limbs, for example in the naupliar antenna of crustaceans, in the mandibles, in the postmandibular limbs of crustaceans such as the cephalocarids and branchiopods, and in the maxillule of hexapods, chilopods and symphylans. The number of endites on the protopodite of postmandibular limbs is variable with taxon. In Cambrian crustaceans such as Rehbachiella kinnekullensis, Dala peilertae and possibly Ercaia minuscula (see reinterpretation in section V.1.b below), the number of endites can be as high as eight or nine (Walossek, 1993 ; Walossek & Müller, 1998 ; Chen et al., 2001). They form a linear array along the inner margin of the undifferentiated protopodite of the limb (Fig. 4 B). The proximal endite is slightly larger and has modified setation, as would be expected for the food groove end of the endite row, but basically does not differ from the others. A maximum of only five or six endites is retained in the trunk limbs of Lepidocaris rhyniensis and extant branchiopods. In cephalocaridan trunk limbs the enditic margin of the protopodite is not strongly divided into defined endites. Outside of the crown-group crustaceans, one endite per limb segment appears to be usual, as in stem-line forms such as the phosphatocopines (Walossek, 1999). The lacinia and galea of the basic hexapod maxilla are interpreted (Chaudonneret, 1977) as representing the endites of two protopodal segments. The presence of endites is highly indicative of protopodal origin, but endite-like structures can also be found on telopodal segments. For example, the typical limb of Rehbachiella kinnekullensis as illustrated by Fig. 47 C in Walossek (1993) has a convex setose margin on the proximal telopodal segment and the setal pattern closely resembles that of the distal protopodal endite. According to the re-interpretations of Ramsköld & Edgecombe (1996), fossil arthropods such as the trilobites Olenoides serratus, Eoredlichia intermedia (Lu) and Misszhouia longicaudata (Zhang & Hou) (as Naraoia longicaudata) carry well-defined endites on two proximal telopodal segments (Fig. 5 A). Conversely, endites are often lost from protopodal segments as, for example, in most hexapod, myriapod and arachnid walking legs (see Fig. 7). Taken alone the presence of endites is insufficient to distinguish unequivocally between components of the protopodite and the telopodite. Some types of eversible vesicles may be modified endites, in particular those found on certain postcephalic limbs in Hexapoda and progoneate Myriapoda. There is uncertainty concerning their segment of origin and, hence, their homology across these taxa (Matsuda, 1976 ; Klass & Kristensen, 2001). However, similarities in vesicle musculature are indicative of homology between the eversible vesicles of progoneates (Symphyla, Diplopoda and Pauropoda). [In pauropods they are present only on the otherwise limbless collum (corresponding to the third postmandibular segment according to Kraus & Kraus, 1994).] This homology is sometimes extended to include also the Archaeognatha and Diplura, within the Hexapoda. The evidence points to their 269 origin from the medial enditic surfaces of the protopodal region of the limb but Klass & Kristensen (2001) concluded that the supporting evidence for homology is not yet very strong. The pattern of expression of particular genes along the proximo-distal axis of the limb provides some evidence of differences in genetic control pathways between protopodite and ramus (or rami). González-Crespo & Morata (1996) demonstrated that function of the gene extradenticle (exd ) was restricted to the body wall and protopodal area of the insect and crustacean limb, and was not affected by the absence of expression of Distal-less (Dll ). The distal telopodal segments are under the control of Dll expression and the hedgehog (hh) signalling pathway : exd is not expressed distally. They inferred a fundamental distinction between the proximal domain of exd expression and the distal domain of Dll expression, and interpreted it as supporting evidence for the hypothesis (Snodgrass, 1935) that ancestral arthropod appendages were originally formed by two segments, a proximal coxopodite derived from the trunk and, a distal telopodite. González-Crespo & Morata (1996) appear to locate the protopodite-telopodite boundary between the coxa and the trochanter in Drosophila. However, they did note overlap in the zones of expression of exd and Dll from the second larval instar onwards. There is currently no unequivocal evidence from genetic expression products that can be used to differentiate precisely the boundary between protopodite and telopodite in arthropod limbs. Given the rapid progress in understanding the genetic control of proximo-distal differentiation in arthropod limb development, it is likely that improved resolution of zones of gene expression (cf. Kojima, Sato & Saigo, 2000 ; Tsuji et al., 2000) will provide new reference points of use in inferring homology between limb segments across arthropod taxa. There is patchy evidence to support the hypothesis that the timing of appearance of features during development differentiates between protopodite and rami. In the collembolan hexapod Orchesella villosa, Bretfeld (1963) found that the first articulation expressed in walking limb buds is that separating the protopodite from the telopodite (cf. Bitsch, 2001). In conchostracan branchiopods, Olesen (1999) showed that the articulations separating the protopodite from both rami of the antenna were the first to appear during development. However, the picture is not uniform across the arthropods. In the symphylan Hanseniella agilis Tiegs, for example, Tiegs (1940) showed the first articulation expressed (see Text Fig. 17C–D – the transverse groove in Tiegs, 1940) separated the ‘ limb base ’ from the body wall, and the second separated femur and tibia. The phylogenetic information within musculature signature patterns has not been fully explored. In extant crustaceans Boxshall (1997) showed that the plane of the coxabasis joint subdividing the protopodite could be identified by its characteristic musculature. However, it is apparent that the profound modifications of the musculature of uniramous limbs for terrestrial life prevent the recognition of such a signature pattern in hexapods and myriapods. Protopodal musculature is considered in more detail below. Shultz (1989) found numerous characters of phylogenetic significance in his comparative analysis of walking leg musculature Geoff A. Boxshall 270 ar A B pl ar po ep ex ex D t ex t C ex t ep1 t ep2 ex F ex t E t Fig. 5. See opposite page for legend. The evolution of arthropod limbs across the arachnid orders, but comparison between major arthropod classes remains problematic. The unequivocal identification of a protopodal vs. a telopodal segment in uniramous limbs is not yet possible. The best criterion is the presence of the outer ramus (=exopodite), or vestige of the outer ramus (Walossek, 1993), as in most trilobites, most Cambrian fossil arthropods, most crustaceans and some chelicerates. ( b ) Evidence of protopodal segmentation The inferred segmentation pattern of the protopodite is pivotal to various phylogenetic schemes. It is not surprising, therefore, that its interpretation has been, and continues to be, keenly debated. The protopodal part of biramous arthropod limbs can only be identified unequivocally when both rami are retained. All trilobites for which the limbs are known have an entire and undivided protopodite, and the same applies to most other Cambrian arachnomorphans, such as Cindarella eucalla and Burgessia bella, and marrellomorphans such as Marrella splendens Walcott and the Devonian Mimetaster hexagonalis. The arachnomorphan Sidneyia inexpectans is the exception here, as it has a remarkable protopodite, providing clear evidence of subdivision (Fig. 4 A). The transverse elongation of the protopodal part of the postantennulary limb in Sidneyia inexpectans has resulted in a marked resemblance to the walking limb of xiphosurans, as previously noted by Bruton (1981). However, in Sidneyia inexpectans the second limb segment (telopodal segment 1 in Bruton’s terminology) bears a well-developed endite and, according to Bruton, it is this segment that bears the exopodite (as gill branch). I could not confirm this site of origin, but if correct, this would provide evidence of a differentiated coxa-basis articulation (Fig. 4A). In addition, one specimen of S. inexpectans (USNM 57494) shows a clear indication of a separate proximal endite (Bruton, 1981). A second exception within the arachnomorphan lineage is the distinct mobile endite carried proximally on the large gnathobases of walking legs 1–5 of Xiphosura and Eurypterida (see Fig. 8B) (Selden, 1981). The mobile endite is provided with its own musculature in xiphosurans (see Fig. 9 E) (Manton, 1964). In crustaceans, the protopodite may be undivided, or may comprise two well-defined segments (typically referred to as coxa and basis), or three segments (precoxa, coxa and basis). The paper by Hansen (1925) details the arguments for the existence of a precoxa in crustaceans but his arguments have failed to gain acceptance. In my opinion, Hansen (1925) incorrectly interpreted limb segmentation in most of the crustacean taxa he considered. There is a distinct precoxal 271 segment in most postmandibular limbs of copepods (Huys & Boxshall, 1991), although it is often reduced and incomplete, as for example in the swimming legs. Using evidence from the musculature, the presence of a precoxal segment was also demonstrated for certain postnaupliar limbs of ostracods, mystacocarids and remipedes within the Crustacea (Boxshall, 1997). There is no evidence for a differentiated precoxa in trunk limbs of cephalocarids (Hessler, 1964) or anostracan and notostracan branchiopods (Fryer, 1983, 1988). There is no evidence for the existence of a precoxa in the naupliar limbs of copepods, remipedes or other maxillopodans. It appears that the precoxa evolved within the Crustacea by subdivision of the coxa of postmandibular limbs, rather than being part of the groundplan. In the Crustacea, it is necessary to consider the naupliar limbs (i.e. second antenna and mandible), separately from the maxillule, and from the postmaxillulary limbs. The antenna and mandible of both stem-lineage and crown-group crustaceans all have a protopodite clearly divided into coxa and basis only. This division of the protopodite was apparent even in Lower Cambrian stem-lineage forms such as the Phosphatocopina (Siveter et al., 2001). The division of the protopodite of the maxillule into coxa and basis may be a feature of the crustacean groundplan. The maxillules of cephalocarids and remipedes both express a well-developed articulation separating coxa and basis (Hessler, 1964; Boxshall, 1997), but the fossil branchiopod Rehbachiella kinnekullensis lacks protopodal subdivision in the maxillule (Walossek, 1993). The possibility that absence of a coxa/basis articulation in the maxillulary protopodite might be secondary rather than plesiomorphic gains some support from the presence of a separate coxa and basis in the maxillule of the closely related Orsten crustacean Bredocaris admirabilis (Müller & Walossek, 1988). Boxshall (1997) concluded that the groundplan maxillule of crown-group Crustacea included a separate coxa and basis, and that the coxa divides to form a separate precoxa and coxal segments in this limb in the Maxillopoda and Remipedia. The postmaxillulary trunk limbs of cephalocarids and both extant and fossil branchiopods do not have an articulation dividing the elongate protopodite into coxa and basis, although cephalocarids have an incipient coxabasis articulation, according to Boxshall (1997). This undivided state is also interpreted here as plesiomorphic and the subdivided state exhibited by all other crustaceans is apomorphic. Hexapods exhibit subdivision of the protopodite, although there is considerable confusion concerning the details of the subdivision, partly because of different nomenclature systems (Deuve, 2001). [In hexapods the protopodal region is referred to as the ‘coxa ’ ; its subdivisions may be termed Fig. 5. (A) Trunk limb of Olenoides serratus. (B) Generalised trunk limb of penaeid decapod crustacean, showing protopodal exites. (C) Second peraeopod of malacostracan Spelaeogriphus lepidops Gordon, showing intrinsic musculature. (D) Trunk limb of Canadaspis perfecta. (E) Fourth postantennulary limb of Agnostus pisiformis. (F ) Second peraeopod of Anaspides tasmaniae, showing intrinsic musculature and flagellate exopodite. Sources : A adapted from Ramsköld & Edgecombe (1996) ; B adapted from Pérez Farfante & Kensley (1997) ; C adapted from Hessler (1982) ; D adapted from Briggs (1978) ; E adapted from Müller & Walossek (1987) ; F original. Abbreviations : ar=arthrobranch, ep=epipodite, ex=exopodite, pl=pleurobranch, po=podobranch, t=telopodite. Scale bar, F=0.25 mm. Geoff A. Boxshall 272 B A C pl ar po D E Fig. 6. Gill development in decapod crustaceans. (A) Developing gill buds of mysis III stage of Penaeus paulensis. (B) Detail of same. (C, D) Developing gill buds of mysis I stage of Pleoticus muelleri (Bate), showing detail of pleurobranch, arthrobranch and podobranch buds. (E) Late stage gills of Pleoticus muelleri. Sources : A–D adapted from de Calazans (1992) ; E unpublished scanning electron micrograph provided by de Calazans. Abbreviations : pl=pleurobranch, ar=arthrobranch, po=podobranch. Scale bar, E=2.6 mm. epicoxa, subcoxa and coxa.] In various hexapods the pleural region of the thorax wall surrounding the articulation with the leg coxa is composed of several small sclerites (cf. Matsuda, 1970). After meticulous analysis of the musculature Barlet (1988) considered that one of these sclerites (the catapleural arch) possibly corresponds to a vestigial protopodal segment, the subcoxa, while another (the anapleural arch) was probably a secondarily sclerotised area of body wall. In his consideration of the hexapodan groundplan, Deuve (2001) recognised the presence of a subcoxa ‘ probably primarily composed of two podomeres (the precoxa and the true subcoxa) more or less fused ’. Deuve explicitly recognised a three-segmented protopodite. The proposed origin of the hexapods from within the Crustacea, perhaps with a sister-group relationship with the Malacostraca, some members of which exhibit a two- or three-segmented protopodite in at least some postantennulary limbs, indicates the possibility of a tripartite protopodite within the as yet unidentified hexapod outgroup. Interestingly the putative epicoxa and subcoxa postulated by Kukalová-Peck (1997) as components of the generalised hexapod limb, represent, according to Deuve (2001), the tergopleurite and epipleurite, respectively, both unambiguously dorsolateral components of the body wall. The evolution of arthropod limbs Chaudonneret’s (1950) elegant studies of the skeletomusculature of the third postantennulary limb (=maxilla) of Thermobia domestica (Packard), a zygentoman hexapod, provide further evidence of a tripartite protopodite. Extrinsic muscles insert on the cardo, which can be interpreted as representing a precoxa, and the intrinsic musculature in the stipes indicates its possible derivation from two segments (subcoxa and coxa). Each of these original stipal segments carries one endite, the lacinia and galea, respectively, according to Chaudonneret (1950). The classic works by Tiegs (1940, 1947) and Manton (1954, 1958, 1965, 1966) on the skeletomusculature of the limbs in the various myriapodan groups also provides evidence for subdivision of the protopodal part of the walking limbs. From the musculature pattern, it is possible to reinterpret the ‘ pleurite ’ of the symphylan walking leg (e.g. Fig. 5 a, e in Manton, 1966) as a misidentified incomplete protopodal segment. Similarly the ‘eucoxa superior ’ of chilopod limbs (Manton, 1965) can be interpreted as representing an incompletely expressed (=partially incorporated) proximal protopodal segment, although Manton (1965, 1966) categorically rejected such interpretations. The protopodite of all postantennulary limbs of trilobites and most fossil and recent arachnomorphans is entire and undivided (see Figs 8 D and 10 A, G). The separation of a small proximal endite in Sidneyia inexpectans may be interpreted as a precursor of the mobile proximal endite of xiphosurans and eurypterids. It is postulated that in mandibulates the first and second postantennulary limbs plesiomorphically displayed differentiation of the protopodite into proximal and distal segments. The first postantennulary limb (=antenna) is not expressed in non-crustacean mandibulates (the Atelocerata) (see Section VIII.1 below). In the second (=mandible) the proximal segment is the mandibular gnathobase and the distal segment (known as the basis in Crustacea), forms the first segment of the palp which is lost, together with the rest of the palp, in non-crustacean mandibulates (see Section VIII.2 below). In postmandibular limbs most crustaceans exhibit coxa-basis differentiation (see Fig. 10C) and some also exhibit the differentiation of the coxal part into precoxa and coxa (see Fig. 10 D). The two distal protopodal segments (coxa and basis) are retained as discrete segments in most extant crustaceans but the proximal one (precoxa) is often partially incorporated into the body wall (see Fig. 10E), or can secondarily fail to differentiate from the coxa. Evidence of its homology can be obtained from the insertion patterns of extrinsic muscles. In Hexapoda the diverse array of sclerites surrounding the limb bases is matched by the diversity of interpretations, but acceptance of a sister-group relationship with some or all of the Crustacea suggests that we should be seeking to derive the hexapod state from a common crustacean/hexapod groundplan. Basing my comparison on the hexapod maxilla [particularly Chaudonneret’s (1950) description of the protopodal skeletomusculature], I infer that the three protopodal parts of the more derived type of crustacean limb (precoxa, coxa and basis) are homologous with the cardo, lacinia-bearing subcoxa, and galea-bearing coxa respectively. Extending this set of proposed homologies to insect walking legs is more problematic but can be supported if we 273 accept the evidence (e.g. Deuve, 2001) that particular sclerites located to the pleural side of the limb bases represent incorporated proximal protopodal segments (see Fig. 10 F ). The telopodite-bearing coxa of insects is thus interpreted as the homologue of the rami-bearing basis of crustaceans. ( c) The origin of coxa-basis differentiation Walossek (1993, 1999) interpreted the process of coxa formation within the Crustacea as driven by the progressive differentiation and enlargement of the proximal endite of the postmandibular limbs. This model is appropriate for the antenna and mandible only, but it cannot be applied to the postmandibular limbs since neither the development of extant crustaceans nor the comparative study of skeletomusculature provides any solid supporting evidence. The maxilla of copepods, for example, carries four protopodal endites proximal to the coxa-basis articulation (two precoxal, two coxal) (Boxshall, 1982 ; Huys & Boxshall, 1991). The proximal precoxal endite is typically more setose than the distal endites (Huys & Boxshall, 1991) but given that four endites are present, it is difficult to attribute coxa formation to the enlargement of the proximal endite alone (as postulated by Walossek, 1993, 1999). The difficulty with the Walossek (1993) hypothesis is largely semantics, in that it changes the usage of the term coxa in postmandibular limbs, from its traditional crustacean definition (that part of the protopodite proximal to the basis). However, reconciling these differing hypotheses indicates that the coxa of the crustacean naupliar limbs (antenna and mandible) is not serially homologous with the coxa of postmandibular limbs. An alternative hypothesis, in which the basis is derived by a secondary process of fusion between the proximalmost segments of the telopodite and exopodite, lacks support. Originally proposed by Itô (1989), it was based on his interpretation of swimming leg structure in cephalocarids, remipedes and copepods, as well as on the apparent axial subdivision of the basis of postmandibular limbs of the Orsten crustacean Skara anulata (Müller & Walossek, 1985). One difficulty with this hypothesis is the rarity of fusion events between rami. Rami may fail to differentiate and become incorporated into the protopodite, but I know of no examples in the Arthropoda where fusion between telopodite and exopodite has been convincingly demonstrated. If neither Walossek’s proximal endite hypothesis, nor Itô’s ramal fusion model of basis formation has strong support, then how might coxa-basis differentiation in postmaxillulary limbs have occurred ? The process envisaged here is the subdivision of the enditic margin of the protopodite of a trilobite-type of limb (see Fig. 10A) to form a linear series of endites along the inner margin. Enlargement of the protopodite and increased definition of the marginal endites gave rise initially to a cephalocarid-like arrangement with up to eight poorly defined enditic lobes (see Fig. 10 B). The formation of true lobate endites provided with intrinsic muscles and able to move relative to one another is driven by trophic specialization. Some functional specialization along the proximal to distal linear array of endites is typical of branchiopods (Fryer, 1988 ; Walossek, 1993). Indeed, Geoff A. Boxshall 274 co + ba A B co ba C ti D t t ten ta ta prp F G H E tr/fe ti/ta I ti ti t ten Fig. 7. See opposite page for legend. The evolution of arthropod limbs proximo-distal functional specialization in arthropod limbs in general is the key to their multifunctionality. In the crustacean mandible, for example, while the coxal gnathobase is involved in food processing, the palp (basis plus rami) may be simultaneously performing a sensory role, or be involved in generating flow fields. Increasing length of the protopodite in the Crustacea, however, brings with it the functional requirement for increased flexibility during swimming motions. The ability to flex posteriorly during the recovery stroke is an important feature of efficient swimming. The short protopodite of cephalocarids shows a line of preferential bending across the posterior face which is strongly reflected in the musculature, as indicated by Hessler (1964) and interpreted as the incipient coxa-basis articulation (Boxshall, 1997). Most crustaceans have a defined coxa-basis articulation within the protopodite, and many have increased flexibility further by division of the proximal part into precoxa and coxa. All three protopodal segments may carry endites. The very long protopodites of Rehbachiella kinnekullensis (Fig. 4 B) and Dala peilertae retain flexibility by having a pleated proximal part to the protopodite. The Cambrian Ercaia minuscula has trunk limbs that I re-interpret here. In the illustrated material (Figs 1I and 2 A in Chen et al., 2001), these biramous limbs are preserved as a proximal series of eight separate circular features continuing as a solid structure comprising either five or six somewhat rectangular subunits, plus an associated but unattached outer fan of setae. I interpret these features as a series of eight endites along the margin of an elongate protopodite, and a telopodite of five or six true segments. The original reconstruction of the exopodite as a foliaceous plate with marginal setae is well supported, but the illustrated specimens provide no evidence of the location of the protopodite-exopodite junction. I consider that the protopodal and telopodal parts of the Ercaia limb closely resemble those of Rehbachiella. Studies of limb development in the branchiopod Leptodora kindti (Focke) have revealed that the slender stenopodial trunk limbs of this predator are secondarily derived from the typical phyllopodial limb found in the larger branchiopods (Olesen, Richter & Scholtz, 2001). Using a variety of methods, Olesen et al. (2001) demonstrated that the defined ‘ segments ’ of the first trunk limb are derived by subdivision of the protopodite of a phyllopodous limb, with each ‘segment ’ corresponding to an endite-bearing portion of the protopodite of Cyclestheria hislopi (Baird). Three novel articulations, each defined by intrinsic musculature, form within the protopodal part of the limb of Leptodora kindti and define 275 four ‘segments ’. None of these ‘segments ’ is homologous with the telopodal segments of an early branchiopod such as Rehbachiella kinnekullensis. ( d ) Outer lobes – exites, epipodites and the question of insect wings There has been much speculation over the appearance and significance of exites on arthropod limbs. Cannon & Manton (1927) regarded the ancestral crustacean limb as biramous and paddle-like, and they regarded Lepidocaris rhyniensis as transitional between the ancestral biramous limb and the epipodite-bearing, foliaceous limbs of extant branchiopods. The discovery of the Cephalocarida in 1955 changed perceptions (Sanders, 1957). Trunk limbs with exites have been a feature of virtually every hypothetical crustacean groundplan since then (Sanders, 1963; Hessler & Newman, 1975; Schram, 1978 ; McLaughlin, 1982; Fryer, 1992) yet there is no evidence for exites in the fossil record of the early Palaeozoic. Not one of the Cambrian fossils from the Burgess Shale, Chengjiang or Orsten faunas exhibits any outer lobe other than the ramus now interpreted as the exopodite. The superbly preserved Orsten crustaceans such as Rehbachiella kinnekullensis and Bredocaris admirabilis do not have exites on any limb, neither does Lepidocaris rhyniensis, the Devonian lipostracan. By the late Palaeozoic (Carboniferous and Permian) some of the Malacostraca, such as Acanthotelson stimpsoni Meek & Worthen, Palaeocaris typus Meek & Worthen, and Nectotelson krejcii (Fritsch), possessed an exite (epipodite) on the proximal protopodal segment of the thoracic walking legs (Schram, 1984). There is no evidence that exites (including epipodites) were present in any early Palaeozoic crustaceans or in any stem-line crustaceans. Exites, in particular epipodites, were not present in the Crustacean groundplan (Walossek, 1999). Cambronatus brasseli from the Devonian Hunsrück Slate has a series of ten pairs of foliaceous postcephalic trunk limbs (Briggs & Bartels, 2001). These limbs are described as apparently uniramous, through loss of the telopodite, and the exopodite apparently comprises four lobes with setose free margins. The homology of these lobes is unclear although Briggs & Bartels (2001) interpret them as exopodal in origin. Protopodal exites are present on the postcephalic trunk limbs of many extant crustaceans. Penaeid decapods, for example, have elaborate gill systems on the thoracic legs. A single biramous leg may carry four gills (pleurobranch, anterior and posterior arthrobranchs, podobranch) plus an epipodite (Fig. 5B) (Pérez Farfante & Kensley, 1997). Fig. 7. Telopodites. (A) First postantennulary limb of Mimetaster hexagonalis, showing possible annulated tarsal region. (B) Maxilla of Thermobia domestica, showing distal annuli lacking intrinsic muscles. (C) Maxilla of Petrobius maritimus Leach, showing distal annulus lacking intrinsic muscles. (D) Distal part of telopodite of walking leg of Scutigera coleoptrata (Linn.), showing annulated tarsal region. (E) Ninth walking leg of Pauropus furcifer, showing intrinsic muscles and tendon to pretarsus. (F, G) Hypothetical ancestral chelicerate walking leg showing musculature of telopodite. (H) First abdominal leg of acerentomid proturan, showing apical eversible vesicle. (I) Tarsal annuli of mite Tarsolarkus sp., arrowheads showing continuations of tendons back towards origin in tibia. Sources : A adapted from Stürmer & Bergström (1976) ; B adapted from Chaudonneret (1950) ; C adapted from Bitsch (1956) ; D adapted from Borucki (1996) ; E original ; F–G adapted from Shultz (1989) ; H adapted from Klass & Kristensen (2001) ; I adapted from Grandjean (1952). Abbreviations : ba=basis, co=coxa, fe=femur, prp=protopodite, t=telopodite, ta=tarsus, ten=tendon to pretarsus/claw, ti=tibia, tr=trochanter. Scale bar, E=50 mm. Geoff A. Boxshall 276 C B D mob prp gn prp gn t ex G t ex t E ex F A end I t H t ex t t ex ex Fig. 8. See opposite page for legend. The evolution of arthropod limbs Adult gills are elaborate branching structures (Fig. 6 E) but they first appear during the zoeal phase of development as simple lobes arising on the outer margin of the limb (podobranch), the arthrodial membrane between limb segments (arthrobranch) or on the lateral body wall (pleurobranch) (Fig. 6 A–D). In extant branchiopods and anaspidaceans there are typically two foliaceous exites on the postcephalic trunk limbs (Fig. 5 F). In copepods there is a well-developed muscular exite (the epipodite) on the coxa and a small bisetose exite on the basis of the biramous maxillule. The single seta in the appropriate position on the outer margin of the coxa of the copepod maxilla has been interpreted as the vestige of an epipodite (Huys & Boxshall, 1991). In ostracods a well-developed epipodite can be present on the maxilla and Boxshall (1997) interpreted the single outer coxal seta on the maxillule of myodocope ostracods as a possible vestigial epipodite. The so-called branchial plates on the limbs of podocope ostracods are here interpreted as exopodites (Horne, Cohen & Martens, 2002), rather than epipodites. The locations of the pseudepipodite originating at the proximal articulation of the exopodite in cephalocaridan crustaceans (Itô, 1989), and of the flagellum originating at the base of the exopodite of the first and second thoracic limbs of the branchiuran genus Argulus, indicate that these two exites are not homologues of the protopodal epipodites in other crustaceans. Across the Crustacea a minimum of nine or ten different and non-homologous exites can be identified including : precoxal epipodite, coxal epipodite (including the peracaridan oostegite), coxal plate (e.g. of amphipods), basal exite, pleurobranch, anterior and posterior arthrobranchs, podobranch, pseudepipodite on exopodal articulation and flagellum on first exopodal segment. It is clear that, within this group of aquatic arthropods, exites have arisen independently on numerous occasions since the late Palaeozoic. While lateral gills and epipodites are absent from early Palaeozoic crustaceans, the Silurian chelicerate Offacolus kingi does appear to carry a number of posterior gill lamellae on the opisthosomal limbs. Orr et al. (2000: Fig. 2k) show effectively longitudinal sections through specimens in which a thick lamella (probably the foliaceous limb ramus) is followed by several thinner lamellae (probable gills). This structure is very similar to the second opisthosomal limb of the trilobite larva of Tachypleus tridentatus, which carries just six gill lamellae on each leg when it hatches (Fig. 4 C). These lamellae originate proximally on the posterior surface of the limb protopodite, and are considered here to be novel structures rather than homologues of any exites on crustaceans. There is no evidence of exites in either the myriapod or chelicerate lineages. Kukalová-Peck’s (1983, 1987, 1992, 277 1997) interpretations of annulate exites on at least four proximal segments of an hypothetical generalized pterygote leg presumably include one on a segment homologous with the crustacean basis (and therefore a possible homologue of the exopodite) but, although widely re-used (e.g. Bitsch, 2001), the reconstruction provided by Kukalová-Peck (1983) is both over-interpreted and unduly hypothesis-dependent. It requires confirmation given that the few published photographs showing the putative exites are of poor quality. These putative exites could more readily be interpreted as annulated sensory elements, several kinds of which are already known from terrestrial arthropods (e.g. Fig. 2G). Kukalová-Peck’s (1997) statement that ‘ coxal exites called ‘swimming legs ’ in Crustacea, and all other primitive exites and endites are, indeed, annulated ’ is virtually incomprehensible due to terminological confusion, but wholly incorrect. While the exopodite can be flagellate in certain derived crustaceans (e.g. Fig. 5F), all exites, including precoxal and coxal epipodites, are unsegmented in the Crustacea. The question of whether crown-group hexapods ever had any exites has gained in significance with the discovery of shared patterns of gene expression between crustacean epipodites and insect wings (Averof & Cohen, 1997). There have long been competing hypotheses for the origin of insect wings ; the two main contenders being gill theory and paranotal theory. Both these hypotheses have nineteenth century origins but the precise formulations have changed many times. Wigglesworth (1973) and Kukalová-Peck (1983, 1987) are proponents of the hypothesis that wings evolved by modification of limb ‘ branches ’ already present in multibranched ancestral appendages, whereas Crampton (1916) and Snodgrass (1935) promoted the paranotal hypothesis that the wings arose as novel outgrowths of the body wall, distinct from the existing limbs. The sharing, by wings and legs, of a common regulatory mechanism for patterning along an antero-posterior axis, based on the contiguous domain of engrailed (en) expression (Basler & Struhl, 1994), does not provide any insight as to which hypothesis is the stronger. Averof & Cohen (1997) found that a homologue of nubbin ( pdm1 ), a Drosophila gene expressed both in the wing and in a weak set of rings in leg primordia, is also expressed specifically in the developing distal protopodal exite (the epipodite) of the branchiopod crustacean Artemia franciscana Drewes. In a second crustacean, the decapod Pacifastacus leniusculus (Dana), pdm was expressed both in the distal epipodite and in rings along the developing telopodites of thoracic legs 2–8, reminiscent of the insect expression pattern (Averof & Cohen, 1997). They also noted a similar pattern of expression for a second Drosophila gene, apterous (ap), in the wing Fig. 8. First postantennulary limb. (A) Skara annulata (Crustacea : Skaracarida). (B) First walking leg of Baltoeurypterus tetragonophthalmus (Chelicerata : Eurypterida). (C) Agnostus pisiformis. (D) Generalised pedipalp of spider (Arachnida : Araneae). (E) Antenna of Speleonectes tulumensis (Crustacea : Remipedia). (F) Antenna of Caenestheria propinqua Sars (Crustacea : Branchiopoda : Spinicaudata) showing multisegmented rami. (G) Detail of intrinsic musculature at base of rami in Caenestheria propinqua. (H) Antenna of decapod crustacean, Penaeus paulensis at Protozoea I stage. (I) Same appendage at mysis I stage. Sources : A adapted from Müller & Walossek (1985) ; B adapted from Selden (1981) ; C adapted from Müller & Walossek (1988) ; D adapted from Savory (1977) ; E adapted from Yager (1987) ; F–G adapted from Shakoori (1968) ; H–I adapted from de Calazans (1992). Abbreviations : ex=exopodite, end=enditic process, gn=gnathobase, mob=mobile proximal endite, prp=protopod, t=telopodite. Geoff A. Boxshall 278 A B C co co co ba ba s t D t mob prp prp E t t H G co ex co E co ba I co t Fig. 9. See opposite page for legend. The evolution of arthropod limbs (dorsal part only) of the insect and in the distal protopodal epipodite of the crustacean. The hypothesis that insect wings have evolved from foliaceous outer lobes, as found on crustacean thoracic limbs, was supported by Averof & Cohen (1997), who also proposed ‘that they may be homologous to specific epipodites of crustacean limbs ’. Averof & Cohen (1997) further suggest that ‘distinct structural progenitors of legs and epipodites/wings were present in the last common ancestor of crustaceans and insects.’ These inferences of Averof & Cohen (1997) deserve detailed analysis. Few people have any doubts about the legs (by which I presume that they mean the telopodites in particular), except possibly for those who argue for arthropod polyphyly. But does the suggestion of distinct structural progenitors of epipodites and wings bear scrutiny? The Averof & Cohen (1997) proposal derives support from embryological studies which demonstrate that, at least in Drosophila, the cells that form the paired wing primordia migrate out of the paired limb fields in early embryogenesis (e.g. Goto & Hayashi, 1997). These data are not incongruent with a protopodal origin of the presumptive wing bud. The distal protopodal epipodite in crustaceans (see Fig. 5F) is protopodal in origin by definition. Insect wings and crustacean epipodites may therefore be derived from the same embryological precursor. Consideration of positional homology, therefore, provides support for Averof & Cohen’s (1997) hypothesis. However, derivation from a common ancestral structure is less well supported. Before consideration of this aspect, it is necessary to eliminate the confusion wrought by Kukalová-Peck’s (1983, 1987) particular formulation of the ‘ exite-into-wings ’ genre of hypotheses. Kukalová-Peck reconstructs the wing as originating from the extreme proximal (her epicoxa) area of the limb protopodite. However, Deuve (2001), has shown that Kukalová-Peck’s epicoxa is the homologue of the tergopleurite, a component of the body wall, whereas crustacean epipodites are typically coxal in origin. Kukalová-Peck’s (1983, 1987) hypothesis is rejected here. Given the presumed homology of wings and epipodites as outlined above, can we accept the Averof & Cohen (1997) suggestion of ‘ distinct structural progenitors ’ of epipodites/ wings ‘in the last common ancestor of crustaceans and insects ’? One difficulty is the lack of epipodites/exites in the crustacean groundplan. The lack of exites in Rehbachiella kinnekullensis, identified as an early representative of the anostracan lineage by Walossek (1993) implies that epipodites are convergently derived in both the Branchiopoda and Malacostraca. Whilst this testifies to convergently co-opted gene functions between branchiopods and malacostracans it does not necessarily cause rejection of the concept of a 279 common structural epipodite/wing progenitor, if hexapods arose within the Crustacea as sister group to Malacostraca, for example. However, consideration of the state exhibited by the non-insectan hexapods is relevant here, given that these taxa (Collembola, Protura and Diplura) are presumed to form effectively the sequence of outgroups to the Insecta (Bitsch & Bitsch, 2000; Klass & Kristensen, 2001). There is no evidence of coxal exites in these hexapods. A Devonian collembolan is remarkably similar to a modern one and they have uniramous limbs without rudiments of exites. The abdominal styli found in certain Diplura and Insecta have been interpreted as exites or as exopodites, but were thought by Klass & Kristensen (2001) more likely to represent vestigial telopodites. The absence of lamellar exite/epipodite structures in these taxa and in all early Palaeozoic crustaceans, unless interpreted as due to secondary losses, must imply that there was no structural progenitor in the sense of Averof & Cohen (1997), and that, although wings and the distal coxal epipodites are homologous in the sense of being derived from a common developmental primordium (Goto & Hayashi, 1997), they are not homologous in the sense of derivation from a common ancestral structure. An alternative explanation for the genetic data, that wings have independently coopted a number of gene functions that were already used in epipodites, was briefly considered by Averof & Cohen (1997) but they regarded this as less likely. Given the rapid progress in genetics it seems possible that this perceived balance of probabilities may be changing. Carroll et al. (2001) wrote ‘ The recruitment of regulatory genes for new patterning roles is a common evolutionary mechanism underlying the convergence of several novel structures and pattern elements ’. Against this background, I prefer to interpret both epipodites and wings as results of secondary recruitment of regulatory genes. The common ancestral function of these genes is still unknown, but it was probably localised in the dorsal body wall and protopodal limb bases of the body segments. A better estimate of the point of origin of the Hexapoda and/or the Insecta, from within the Crustacea should shed further light on this unresolved problem. (2 ) Telopodite ( a ) Primary segmentation The difficulty in determining the boundary between protopodite and telopodite in uniramous limbs has been discussed above. This has added to the general confusion surrounding the number, homology and nomenclature of segments in the telopodites of arthropod walking limbs. There has been considerable debate concerning the number of Fig. 9. Second postantennulary limb. (A) Bredocaris admirabilis. (B) Conchoecia elegans Sars (Ostracoda, Myodocopa). (C) Scolopendra cingulata Latreille (Chilopoda : Scolopendromorpha). (D) Agnostus pisiformis. (E) Tachypleus tridentatus (Xiphosura), with inset showing detail of mobile endite. (F) Generalised copepod mandible (Crustacea). (G) Oniscigaster wakefieldi McLachlan (Insecta, Ephemeroptera). (H) Locusta migratoria Linnaeus (Insecta). (I) Poratophilus punctatus Attems (Diplopoda). Sources : A adapted from Müller & Walossek (1988) ; B adapted from Sars (1922) ; C adapted from Borucki (1996) ; D adapted from Müller & Walossek (1987) ; E adapted from Manton (1964) ; F adapted from Huys & Boxshall (1991) ; G adapted from Staniczek (2000) ; H–I adapted from Manton (1964). Abbreviations : ba=basis, co=coxa, ex=exopodite, mob=mobile proximal endite, prp=protopodite, s=seta representing exopod, t=telopodite. Geoff A. Boxshall 280 B C co A prp ba ex pc pc t co+ ba co+ ba pc co+ ba prp F G E D Fig. 10. Schematic showing hypothetical changes required to derive mandibulate and arachnid postmandibular limbs from arthropodan groundplan limb. (A) Hypothetical biramous limb of arthropodan groundplan, with whole protopodite gnathobase. (B) Postmaxillulary limb of mandibulate groundplan, showing elongate protopodite lacking differentiation into proximal and distal segments, as found in Rehbachiella kinnekullensis postmaxillulary limbs. (C) Postmaxillulary limb, showing elongate protopodite differentiated into proximal (coxa) and distal (basis) segments. (D) Postmaxillulary limb, showing elongate protopodite differentiated into proximal, middle and distal segments. (E) Postmaxillulary limb, showing loss of exopodite and protopodite differentiated into proximal, middle and distal segments but with proximal protopodal segment reduced to lateral sclerite partly incorporated into body wall, as found in many crustaceans. (F) Postmaxillulary limb, showing protopodite differentiated into proximal, middle and distal segments but with proximal and middle protopodal segments reduced to lateral sclerites incorporated into body wall, as found in non-crustacean mandibulates. (G) Walking leg of groundplan arachnid, showing reduced protopodal gnathobase and loss of exopodite. Abbreviations : ba=basis, co=coxa, co+=coxa (distal part of subdivided coxa), ex=exopodite, pc=precoxa (proximal part of subdivided coxa), prp=undivided protopodite, t=telopodite. telopodal segments in the groundplan of each major arthropodan taxon and numerous schemes have been proposed to establish homologous landmarks along the legs. Manton (1966) referred to the ‘ welter of assumptions ’ underpinning such schemes. Since Manton’s work, some new evidence has emerged and the widely accepted re-establishment of the Arthropoda as a monophyletic group validates the search for a groundplan telopodal segmentation pattern for arthropods. Shultz (1989), in a comparative analysis of the walking legs in arachnids, concluded that a basic system of eight segments is primitive for chelicerate legs (Table 1). He also inferred that the undivided femur found in walking legs 1 and 2 was secondary, arising by fusion or failure to develop of the basifemur/telofemur articulation in the common ancestor of Eurypterida, Arachnida and perhaps Xiphosura. Of the eight segments, only the proximal one (referred to as the coxa) can be interpreted as protopodal in origin since, 7 (Claw) Patella Tibia Basitarsus Telotarsus Pretarsus a based on Eoredlichia from Ramsköld & Edgecombe (1996) ; b based on primitive arachnid walking leg hypothesised by Shultz (1989) ; c based on posterior legs of Agnostus pisiformis described by Müller & Walossek (1987) ; d based on fourth walking leg of Limulus polyphemus from Shultz (1989) ; e based on Shultz (1989) ; f based on data from Sidneyia inexpectans in Bruton (1981) and in Fig. 4 A, but protopodal nature of segment 1 based on comments in Bruton (1981) ; g telopodite based on Cephalocarida, a precoxa is derived within Crustacea, as found in Copepoda (Huys & Boxshall, 1991) ; h reinterpreted from Manton (1954, 1958) ; i based on Borucki (1996) ; j based on Kristensen (1997) and Bitsch (2001), with possible precoxal vestige included from Deuve (2001). [ The homology of the ‘ missing’ patella in mandibulates is not addressed here.] Pretarsus Pretarsus Claw 6 (Dactylus) 7 Pretarsus Pretarsus 4 5 6 Patella Tibia Tarsus Trochanter Basifemur Telofemur Patella Tibia Tarsus Trochanter Femur 1 2 3 4 5 6 Trochanter Femur 1 2 3 4 5 6 Telopodite Coxa ?mobile endite Coxa Coxa Coxa Basis Protopodite 281 7 ? Tibia Tarsus Trochanter Prefemur Femur ? Tibia Tarsus ?Precoxa Subcoxa Coxa Trochanter Femur Coxa Trochanter Prefemur Femur Postfemur ? Tibia Tarsus [Precoxa] Coxa Basis 1 (Ischium) 2 3 (Merus) ? 4 (Carpus) 5 (Propodus) ?endite Coxa 1 2 3 Crustaceag Sidneyiaf Pycnogonidae Xiphosurad Agnostusc Arachnidab Trilobitaa Segment Table 1. Comparative analysis of possible segmentation patterns in arthropod postantennulary (trunk) limbs Diplopodah Coxa Chilopodai Hexapoda j The evolution of arthropod limbs despite the formation of an independently mobile proximal endite in xiphosuran (see Fig. 9 E) and eurypterid (see Fig. 8 B) prosomal legs, there is no evidence of subdivision of the entire protopodal segment in chelicerate anatomy or embryology. In trilobites and in unclassified Cambrian arthropods such as Burgessia bella and Cindarella eucalla, the postantennulary limbs also have a telopodite of seven segments including the distal claw. Offacolus kingi has a xiphosuran-like telopodite comprising six segments, identified as trochanter, femur, patella, tibia, tarsus and pretarsus by Sutton et al. (2002). Walossek (1999) noted that the presence of a seven-segmented telopodite in Canadaspis perfecta suggests that it evolved early within the stem-line of the Arthropoda (his Euarthropoda). In chilopods, Borucki (1996) demonstrated a basic sixsegmented telopodite comprising trochanter, prefemur, femur, tibia, tarsus and pretarsus. The trochanter is reduced and separated from the prefemur by a non-functional articulation. It may be incompletely defined from the prefemur. In the Symphyla the walking legs have a five-segmented telopodite with an undivided tarsus (Tiegs, 1940). The telopodites of Pauropoda have a similar basic segmental composition, but they have a divided tarsus (Tiegs, 1947) except in the posteriormost legs (Fig. 7E). In both symphylans and pauropods the segmental composition was interpreted by Tiegs (1940, 1947) as trochanter, femur, tibia, tarsus and claw (=pretarsus). Diplopod legs were described by Manton (1954, 1958) as eight-segmented : comprising coxa plus a seven-segmented telopodite (trochanter, prefemur, femur, postfemur, tibia, tarsus and claw=pretarsus). Consideration of the musculature figured by Tiegs (1940, 1947) and by Manton (1954, 1958, 1965) leads me to infer that the basic seven-segmented telopodite of diplopods is probably a secondary condition, resulting from the formation of annuli: the coxa and the trochanter of Manton representing the two annuli of a subdivided segment, and the femur and the postfemur of Manton similarly representing a subdivided segment. The evidence for the former being the passage of muscles (such as levator femoris 1) through the putative coxa/trochanter articulation and through the putative trochanter/prefemur articulation (such as flexor tarsi postfemoris 1=fl.ta.po.fe.l.) without inserting or forming an intermediate attachment. The evidence for the latter is the form of the origin of muscle fl.ta.po.fe.l. which spans the femur/postfemur articulation. Both these joints have atypical muscle configurations which I interpret as evidence of their origin by secondary annulation. This reinterpretation allows for a basic six-segmented telopodite in diplopods (Table 1), as in chilopods, and provides additional morphological data in accord with the concept of a monophyletic Myriapoda (Regier & Shultz, 2001). In malacostracan crustaceans the plesiomorphic telopodite is five-segmented (Fig. 5 C), comprising the named segments : ischium, merus, carpus, propodus and dactylus. However several malacostracans, such as the Anaspidacea (Syncarida) and some mysids, have been described as having a total of six segments, with a preischium located between the basis and ischium (e.g. Hansen, 1925; McLaughlin, 1982). Re-examination of Anaspides tasmaniae peraeopods (Fig. 5 F) has failed to confirm the presence of a defined 282 preischium. However, in Cephalocarida the telopodite of the trunk limbs is six-segmented (Hessler, 1964) and the telopodite of the copepod first trunk limb (=maxilliped) is six-segmented (Huys & Boxshall, 1991). So, although most limbs in crustaceans have five or fewer telopodal segments, it is inferred here that six is the original number in crustacean postantennulary limbs (Table 1). One consequence of this is that the unsegmented state of the telopodite of extant branchiopod trunk limbs is secondary. This gains support from the presence of segmented telopodites in both Rehbachiella kinnekullensis and Bredocaris admirabilis. The existence of apparently segmented stenopodial trunk limbs in the branchiopod Leptodora kindti does not represent an example of character reversal since the ‘segments ’ are novel structures derived from the protopodite rather than the telopodite (Olesen et al., 2001). Kukalová-Peck (1997) hypothesised a generalised thoracic limb comprising 11 segments for the Insecta, of which four (epicoxa, subcoxa, coxa and trochanter) would have to be considered protopodal in origin (see Table 1 in Bitsch, 2001), and the remaining seven (prefemur, femur, patella, tibia, basitarsus, eutarsus and posttarsus) were presumably telopodal in origin. This system had the disadvantage of treating the trochanter as protopodal in origin, a treatment that is contradicted by genetic data (González-Crespo & Morata, 1996). In addition, Deuve (2001) has interpreted the epicoxa and subcoxa of Kukalová-Peck (1997) as bodywall components. As concluded by Willmann (1997), there is little evidence in support of Kukalová-Peck’s (1997) hypothesis of an eleven-segmented basic hexapod limb. The basic five-segmented telopodite of Bitsch (1994), Willmann (1997) and Kristensen (1997) is adopted here for hexapod walking legs. (b ) Intercalary and terminal annulation The recognition of oligomerization as a dominant evolutionary trend in arthropods has led to assumptions that more ‘ segments ’ must equate to ‘ more ancestral ’, and presumably provides any underlying rationale for criticisms of groundplan limbs as ‘ podomere deficient’, as for example in Kukalová-Peck’s (1997) criticism of Bitsch’s (1994) groundplan. Huys & Boxshall (1991) identified oligomerization as a dominant trend in the evolution of copepod crustaceans but noted exceptions – examples of secondary increase. The danger therefore is that groundplans could be constructed by mere compilation of maximum states of apparent segmentation within groups, without first identifying any secondary division of primary limb segments by annulation. This danger is greatest when interpreting fossils for which vital data on musculature are typically lacking [but see Stürmer & Bergström (1976) for example]. Intercalary annulation of the telopodite is widespread in arthropods, most commonly in the tarsal region. In Arachnida, for example, annulation of the tarsal region has arisen independently in the Acari (Grandjean, 1952 ; Judson, 1994), the pholcid Araneae (Lockett & Millidge, 1951), Amblypygi (Weygoldt, 1996) and Opiliones (Shultz, 1989). In its extreme form this annulation has given rise to the antenniform first walking legs of amblypygids, which may comprise as many as 28 Geoff A. Boxshall tibial and 54 tarsal annuli (Weygoldt, 1996). This form of annulation is intercalary since the opposing muscles operating the distalmost segment [the pretarsus in Shultz’s (1989) nomenclature] extend as long tendons through the annuli without forming attachments within them (Grandjean, 1952 ; Shultz, 1989), before inserting on the pretarsus (the distal claw) (Fig. 7 I). Annulation of telopodal segments in the more proximal part of the limb is also known in arachnids. Mites of the families Labidostommidae and Sphaerolochidae, for example, have a tri-annulate femur, and both the Ricinulei and the opilioacarid mites are interpreted as having a bi-annulate trochanter. The pedipalps and first walking legs of the pycnogonid Nymphonella tapetis Ohshima are distally annulate. Subterminal (=intercalary) annulation is also known in stenopodoidean and caridean crustaceans. Processid shrimps, for example, have an annulate carpus on peraeopod 4 but a normal chela (comprising propodus and dactylus) is present distally on this limb. Examples of annulation of the propodus of various mysid crustaceans were figured by Hansen (1925). The Devonian marrellomorphan Mimetaster hexagonalis is one of the earliest known examples of tarsal annulation. In the first postantennulary limb there is a distinct tarsal region consisting of four annuli at the tip of the telopodite (Fig. 7 A). The best example of intercalary annulation in Myriapoda is found in the scutigeromorph centipedes. The trunk limbs exhibit extensive annulation of the tarsal region (Fig. 7 D), interpreted by Manton (1977) as an adaptation for rapid running. As in the case of chelicerates, the distal intrinsic musculature is located primarily in the tibia and tendons pass through the annulated zone to insert on the pretarsus (Borucki, 1996). Tiegs (1947) showed that the apparently seven-segmented telopodites of pauropod limbs included an annulated tarsal region consisting of three podomeres. He showed that the telopodite comprised only five segments : trochanter, femur, tibia, tripartite tarsus and claw (pretarsus). The pretarsal flexor muscles originate in the tibia and insert on the pretarsus via a long tendon. Most hexapods exhibit some annulation in the tarsal region of the walking legs. The number of differentiated tarsal annuli may vary from none (even the tibia and tarsus are undifferentiated in Collembola) to five as in Drosophila melanogaster Meigen, Tricholepidion gertschi Wygodzinsky and many pterygote orders. Bitsch (2001) considered it possible that a pentameric tarsus is an apomorphy for the dicondylian hexapods, with secondary reductions generating the varied tarsal formulae found in Zygentoma and some pterygotes. In the machiliids (Archaeognatha) the telopodite (=palp) of the third postantennulary limb (=maxilla) appears sevensegmented. However, study of the musculature (Bitsch, 1956: Figs 1–3) revealed that the last putative segment is devoid of intrinsic muscles (Fig. 7 C). Similarly in Thermobia domestica (Zygentoma) the palp is apparently six-segmented but the two distal ‘ segments ’ are devoid of intrinsic muscles (Fig. 7B) (Chaudonneret, 1950). In both taxa these distal ‘ segments ’ should be interpreted as terminal annulations since they lack intrinsic muscles. The maxillary palp in both of these non-insectan hexapods has a flagellate tip and is The evolution of arthropod limbs clearly an inappropriate model for telopodal segmentation in the hexapodan groundplan. The maxillary and labial palps of certain ephemeropteran nymphs are described as multisegmented (e.g. Peters & Campbell, 1992) but these structures are almost certainly terminal flagella consisting of numerous annuli. It is interesting to note that these examples of terminal annulation in arthropods involve primarily sensory structures, as in the antennules. In locomotory appendages annulations are typically intercalary, although examples are known of terminal annulation, as in the flagellate exopodites of the anaspidacean peraeopods (Fig. 5 F ). ( c) Secondary segmentation Both rami of the postcephalic trunk limbs of Cirripedia (Crustacea : Thecostraca) are transformed into cirri that form the food capture apparatus of the sessile adult barnacles. Barnacle cirri extend hydraulically and the distal intrinsic musculature, comprising a single flexor which extends to the tip of each cirrus, is involved primarily in cirri furling (Cannon, 1947). The presence of intrinsic muscles which form an intermediate attachment in each segment of the cirrus, indicates that cirral divisions should not be interpreted as annuli. The cirri of barnacles have been described as a unique specialization associated with a sessile feeding habit, and they are here interpreted as secondarily multisegmented. The length of the cirri (and their segmentation) can vary with environmental factors, such as the degree of exposure to wave action on the shore (Arsenault, Marchinko & Palmer, 2001). The ancestral limb of the cirripede lineage is here inferred to resemble the biramous swimming limbs (with each ramus being three-segmented) of the cyprid larval stage and of other thecostracans, such as the Ascothoracida (by inference from the phylogenetic schemes of Høeg et al., 1999). The antenna of conchostracan crustaceans is biramous and both rami appear flagellate, however both exopodite and telopodite comprise numerous segments defined by the presence of intrinsic muscles (Fig. 8 F, G) (Shakoori, 1968). This group has exhibited antennal rami of this form at least since the Carboniferous (Orr & Briggs, 1999). Reference to the branchiopod phylogenetic scheme of Walossek (1993) indicates that the multisegmented telopodite can be interpreted as uniquely derived within the conchostracan lineage. As such it constitutes a clear example of a limb in which the number of segments has increased markedly within a lineage. Examples of secondary increases in true segmentation, such as the cirri of cirripedes and the antennae of conchostracans, are relatively rare. Some apparent examples from the fossil record are open to a different interpretation. The newly described Devonian Cambronatus brasseli was reconstructed as having four pairs of uniramous, postantennulary cephalic limbs (Briggs & Bartels, 2001). These limbs are described as comprising approximately 15 short podomeres, each bearing a pair of stout spines. I here re-interpret these limbs as having only four ( ?telopodal) segments, but the protopodal parts of the limbs were not preserved. The telopodal segments are separated by 283 well-defined articulations and with the exception of the claw, each is ornamented with pectinate cuticular scales (Fig. 4 D) which resemble spines when preserved in side view. Similarly, the telopodite of Ercaia minuscula is here reinterpreted as comprising only five or six segments (see Section V.1 above). The telopodite of the postantennulary limb in the arthropod groundplan is inferred to have been seven-segmented (see Fig. 10A) following Walossek & Müller (1990). The seven-segmented state is retained in most Trilobita, many other Cambrian arthropods, and the Chelicerata (Fig. 7 F, G) according to Shultz (1989). In Agnostus pisiformis the telopodite is six-segmented in the second and third postantennulary limbs (Fig. 9 D) but seven-segmented in the more posterior limbs (Fig. 5 E) (Müller & Walossek, 1987). I regard the six-segmented state as synapomorphic for the mandibulates. Six segments are retained in at least some representatives of all major mandibulate taxa but nearly all also exhibit some form of apparent secondary increase in telopodal segmentation. This most commonly arises as a result of intercalary annulation but a few examples of terminal annulation are noted and more may be expected. Embryological and developmental studies demonstrate that in arthropods in general, segmental reductions arise in ontogeny mainly by failure of expression of articulations rather than by secondary fusion of already separate segments. The widespread practice of referring to compound segments that arose by this means as ‘ fused ’, is misleading. They arose by failure to differentiate, specifically through failure of an articulation to be expressed during development. Describing the tibia/tarsus of collembolans as fused is less accurate than stating that the articulation separating the tibia from the tarsus in the outgroup of the Collembola is not expressed within the Collembola. The formation of articulations separating limb segments, at least in Drosophila melanogaster, requires the precise and spatially limited activation of the Notch pathway (de Celis et al., 1998; Bishop et al., 1999). Notch signalling in Drosophila limbs is modulated by selector genes which alter the expression of the Notch ligands Serrate and Delta (Casares & Mann, 2001). Controlling any aspect of the Notch signalling pathway could provide a mechanism for suppressing the process of articulation formation. (3 ) Exopodite More is known about exopodites in the Crustacea than in other arthropods. Exopodites are typically retained in the postantennulary limbs of both crown-group and stem-line crustaceans. The diverse array of specialised crustacean larvae often exhibits biramous limbs even if the adults secondarily lack exopodites on one or more limbs. Exopodites, a feature of the arthropod groundplan (Walossek, 1999) were also widely distributed in Cambrian, Silurian and Devonian arthropods including Sanctacaris uncata, Marrella splendens, Burgessia bella, Offacolus kingi, Mimetaster hexagonalis, the Trilobita and their relatives. Indeed arthropods with a series of uniramous postantennulary limbs (i.e. lacking exopodites) are the exception in the early to mid Palaeozoic. 284 Crustacean exopodites are multisegmented in the first and second postantennulary limbs (the antenna and mandible) but in postmandibular limbs the outer ramus is often one- to two-segmented, sometimes three- or more segmented, or even secondarily flagellate (Fig. 5F ). McLaughlin (1982) describes the adult malacostracan trunk limb as having a two-segmented exopodite, with the distal segment often becoming flagellate. The difference in basic exopodite structure between naupliar and postnaupliar limbs has been regarded as a diagnostic character for crowngroup Crustacea, since it is lacking in certain stem-group taxa, such as Cambrocaris baltica (Walossek & Szaniawski, 1991). However some non-crustaceans, such as Agnostus pisiformis, also share the naupliar/postnaupliar division in basic limb structure, with the postnaupliar limbs bearing only small exopodites (Fig. 5E) with a maximum of two or possibly three segments (Müller & Walossek, 1987). Exopodites are retained on the opisthosomal limbs of xiphosurans (Fig. 4 C), and in most prosomal and opisthosomal limbs of the Silurian Offacolus kingi (Orr et al., 2000; Sutton et al., 2002), but are lost in crown-group arachnids. The flabellum carried ventro-laterally on the undivided protopodite of the sixth prosomal limb of xiphosurans has been interpreted as a vestigial exopodite (e.g. Dunlop & Braddy, 2001), but also as an epipodite (e.g. Mittmann & Scholtz, 2001). It lacks musculature but carries an impressive array of sensors and is well innervated (Patten & Redenbaugh, 1899; Waterman & Travis, 1953). The interpretation of the flabellum as an exopodite is supported indirectly by the absence of epipodites in early Palaeozoic arthropods and the total absence of epipodites in crowngroup chelicerates. Interestingly, Mittmann & Scholtz (2001 : p. 238) interpret the flabellum as having ‘ evolved from an epipodite of the fourth walking leg ’ (=sixth prosomal limb). However, they provide what I consider to be the best evidence supporting the interpretation of the flabellum as a vestigial exopodite. They found that the presumptive flabellum in Limulus polyphemus Linn. expresses Distal-less (Dll ) throughout its whole development and that all anterior prosomal limbs (except the chelicerae) showed transient Dll expression in a serially homologous position. Mittmann & Scholtz (2001) state that this latter expression ‘ can be interpreted as remnants of former epipodites ’. This interpretation is rejected here. To my knowledge, epipodites never occur on the first and second postantennulary limbs in any arthropod, living or fossil. Even in the crustaceans, where numerous different types of exites are known, these limbs (antennae and mandibles) only ever have telopodites and exopodites. Serial homology in Dll expression, as found by Mittmann & Scholtz (2001) from the first to fifth postantennulary limbs is here interpreted as evidence of a vestigial exopodite. The lack of serially homologous Dll expression in the chelicerae supports the hypothesis developed herein, that the arthropod antennule is fundamentally uniramous, and never bears an exopodite. The possession, by the Silurian chelicerate Offacolus kingi, of uniramous chelicerae followed by five pairs of biramous limbs is in accord with this hypothesis. Lack of knowledge of the limbs of stem-lineage Myriapoda allows us no room to speculate on events surrounding the Geoff A. Boxshall loss of exopodites and the evolution of the particular kinds of terrestrially adapted telopodite characteristic of crowngroup Myriapoda. Uncertainty over the sister-group relationships creates a similar situation for the Hexapoda. Adopting the emerging hypothesis of hexapod origins from within a paraphyletic Crustacea, we still lack critical information on limb-structure for the group of crustaceans that served as the immediate stem-line of the Hexapoda during the time preceding their emergence on land and first appearance in the fossil record. The presence of vestiges of exopodites in crown-group Hexapoda and crown-group Myriapoda remains unconfirmed. The abdominal styli found in the Insecta and in dipluran hexapods seem more likely to represent vestiges of the telopodite (Klass & Kristensen, 2001). None of the other types of styli in hexapods or myriapods presents any evidence to support an hypothesis of exopodal derivation. The original form and segmentation pattern of the exopodite in arthropods remains difficult to estimate. It may have been unsegmented and foliaceous, as in the Cambrian crustacean Rehbachiella kinnekullensis (Fig. 4B), and in Ercaia minuscula and the opisthosomal limbs of chelicerate crowngroup taxa such as xiphosurans (Fig. 4C) and eurypterids, as well as in possible stem-lineage chelicerates such as Sanctacaris uncata. It may have been two-segmented, as in extant cephalocaridan crustaceans and some malacostracans (Fig. 5C, F), in the Cambrian crustacean Dala peilertae, in the trilobites Misszhouia longicaudata (Zhang & Hou), Olenoides serratus (Fig. 5A), Eoredlichia intermedia (cf. Ramsköld & Edgecombe, 1996) and in Sapeiron glumaceum Hou, Ramsköld & Bergström (Edgecombe & Ramsköld, 1999), in Emeraldella brocki (as reinterpreted by Edgecombe & Ramsköld, 1999) and in the second trunk limb of Agnostus pisiformis (Fig. 5 E) (Müller & Walossek, 1987). Or, it may even have been multisegmented, as in crustacean naupliar limbs (antennae and mandibles), in certain phosphatocopines, in A. pisiformis (first and second postantennulary limbs only), in Offacolus kingi (second to sixth postantennulary limbs only), and in marellomorphan fossils such as Marrella splendens and Mimetaster hexagonalis. Studies of crustacean development indicate that the exopodite typically first appears as an unsegmented lobe (as does the telopodite). In crustaceans with segmented exopodites in the adult, the lobate exopodal primordium subdivides by the sequential expression of functional articulations, resulting in a segmented condition. Copepods are a typical example, in which both rami of the biramous trunk legs appear first as lobes, then at succeeding moults become first one-segmented, then two-segmented, and finally three-segmented by the last copepodid larval stage (Karaytug & Boxshall, 1996). The opposite polarity of developmental change – from multisegmented ramus to foliaceous paddle – also occurs within the Crustacea. In the penaeid decapods, for example, the antenna of the naupliar and protozoeal phases has a multisegmented exopodite (Fig. 8 H) which gradually loses its segmentation during development, until by the megalope (=mysis I stage) phase the exopodite has formed the unsegmented antennal scale (Fig. 8 I) so characteristic of the adult caridoid facies (Hessler, 1983). A similar pattern is The evolution of arthropod limbs apparent during the ontogeny of postmandibular limbs of Hesslandona unisulcata Müller, a phosphatocopine (A. Maas, personal communication). The form of the exopodites of Offacolus kingi prosomal limbs is unique. The second to fifth postantennulary limbs each carry a robust six-segmented exopodite terminating in a setal fan (Sutton et al., 2002). The exopodite of the sixth postantennulary limb is elongated, possibly flexible and not obviously segmented. Offacolus kingi provides solid evidence that the postantennulary prosomal limbs of chelicerates were all originally biramous but lacked epipodites, supporting the re-interpretation presented here of Mittmann & Scholtz’s (2001) gene expression data. It is tentatively suggested here that the ancestral state of the exopodite in the postantennulary limbs of arthropods may have been two-segmented, or at least bipartite, and flattened. The proximal part was probably muscular, with intrinsic muscles inserting basally in the distal part (cf. Fig. 5 C). Several of the latest reconstructions of postantennulary limbs of trilobites (Fig. 5A), and some other Cambrian arthropods conform to this bipartite pattern and extant cephalocaridans have two-segmented exopodites on the postcephalic trunk limbs. Unsegmented exopodites, as found in chelicerate-lineage opisthosomal limbs, branchiopod crustacean trunk limbs, and some Cambrian forms such as Alalcomenaeus cambricus Simonetta, could easily be derived from this state simply by lack of expression of any division between proximal and distal parts. Exopodites comprising few to many segments, e.g. the three-segmented state in copepod swimming legs or the six-segmented state of Offacolus kingi, may result from the expression of functional articulations in the muscular proximal zone of the developing ramus. Flagellate exopodites (Fig. 5F ) are formed by annulation, as discussed above for the telopodite and the antennule. Distinguishing between multisegmented and flagellate exopodites in fossils remains problematic. VI. ORIGIN OF THE BIRAMOUS LIMB The basic postantennulary arthropodan limb envisaged here is a characteristic of the crown-group Arthropoda. It is, however, a complex biramous structure and its component parts probably appeared sequentially along the arthropod stem-lineage. The character states exhibited by the arthropod stem-lineage are examined by Maas & Waloszek (2001) and Budd (2001) and this issue is only briefly considered here. The lobopodan theory developed by Budd (1996) is based on re-examination and re-interpretation of the Cambrian lobopodian Opabinia regalis Walcott and on the recognition of similarities between Opabinia regalis, Kerygmachela kirkegaardi Budd and Anomalocaris canadensis Whiteaves. Budd (1996) constructed a speculative scenario in which the paired gillfolds arranged on the annulated trunk of Opabinia regalis were interpreted as homologues of the exopodites (as gill branches) of biramous arthropodan limbs [as previously suggested by Simonetta & Della Cave (1991)], and the series of triangular axial structures in the fossils was interpreted as 285 a series of ventrally-located conical limbs. In this scenario, the lobopod limb is interpreted as the homologue of the walking leg (=telopodite plus part of protopodite). These limbs were reconstructed by Budd (1996) as very lobopodlike, although the putative paired apical claws are based on weak evidence. Despite the separate origins of the gillcarrying lateral flap (putative exopodite) and the ventral lobopod-like limb (putative telopodite) on the body wall, Budd (1996) regarded this as probably the strongest contender for explaining the origin of biramous limbs. Other existing hypotheses suggested for the origin of the biramous limb are the formation of the outer lobes or exopodite from an outgrowth on a lobopod-like limb (Snodgrass, 1938), the formation of the biramous limb from paired outgrowths of the body wall (Fryer, 1992), and by the fusion of two duplosegments each originally bearing a uniramous limb (Emerson & Schram, 1990). The duplosegment hypothesis of Emerson & Schram is not tenable given the expression patterns of genes such as engrailed (cf. Akam, 2000). Budd’s (1996, 2001) scenario has not yet been fully tested but derives some support from the findings of GonzálezCrespo & Morata (1996) who demonstrated that expression of the gene extradenticle (exd) was common to the body wall and protopodal area of the insect and crustacean limb. Intermediate stages between the lobopod limb and the basic biramous limb postulated for the Arthropoda, have yet to be considered. Snodgrass (1935) hypothesised a twosegmented limb comprising protopodite and telopodite. González-Crespo & Morata (1996) found support for this from expression patterns of exd and Dll. This would also be in accord with the common developmental pattern in which the articulation separating protopodite from telopodite is the first to differentiate during development. It is contradicted by Casares & Mann’s (2001) study of the ground state of the ventral appendages of Drosophila. They concluded that a two-segmented state was indicated but that the division was between a proximal segment (comprising undifferentiated protopodite to tibia) and tarsus (including pretarsus). Their interpretation is heavily influenced by genetic similarities between thoracic legs and the highly apomorphic antennule of Drosophila. Given the distribution of intrinsic muscles (all originating within the proximal segments up to and including the tibia), this interpretation derives little support from morphology. Their so-called ground state demonstrates simply that true segments, as defined by the presence of intrinsic musculature, have a different developmental pathway from annulations. There is small group of Cambrian arthropods that I have not considered thus far. It is characterised by a unique type of trunk limb, as found in Canadaspis perfecta. The unique feature is the presence of an apparently annulate main limb axis lacking clear differentiation between protopodal and telopodal segments (Fig. 5D), other than the finer enditic setation of the latter (cf. Briggs, 1978). The units of the axis do not resemble the segments of other arthropodan limbs. If the apparent lack of true segmentation in the telopodite is original rather than secondary, then this kind of limb axis could be interpreted as plesiomorphic relative to the basic arthropodan biramous limb postulated as a synapomorphy of crown-group Arthropoda (the Euarthropoda). Indeed, 286 Hou & Bergström (1997) and Walossek (1999) concluded that the Canadaspis limb has no basis (=protopodite in my terminology) but possesses a set of seven proximal ‘articles ’. Walossek (1999) further hypothesised that the basis (my protopodite) is formed by fusion of these proximal articles. Since these taxa possess an exopodite in combination with an annulated but lobopod-like limb axis, it could be inferred that Canadaspis and Fuxianhuia protensa Hou (which has the same type of limb axis) lie close to the stem line of the Arthropoda (Maas & Waloszek, 2001). Hou & Bergström (1997) and Bergström & Hou (1998) also constructed a scenario of this kind but they failed to consider the fundamental nature of the protopodal or telopodal units. The alternative view, supported by other character states such as the presence of a carapace in Canadaspis, which is unknown in lobopodian taxa, is that the apparently annulate form of the protopodite in Canadaspis is a secondary specialization. VII. ORIGIN OF UNIRAMOUS LIMBS Although a basic biramous limb is postulated for the Arthropoda as a whole, only uniramous limbs are present in the great majority of extant arthropod species (Fig. 7A–I). Even in crustaceans, exopodites are commonly lost. Crustacean cephalic and thoracic limbs that are uniramous by loss of the exopodite include, among many others : the maxillules, maxillae and maxillipeds of remipedes, the maxillae and maxillipeds of copepods, the maxillules and maxillae of branchiurans and the maxillules and maxillae of mystacocarids. Across the Arthropoda, uniramous limbs have arisen by loss of the exopodite on numerous independent occasions. The telopodite can also be lost, leaving a uniramous limb, but this is a much rarer event than loss of the exopodite. For example, the first postantennulary limb of Agnostus pisiformis is uniramous due to loss of the telopodite (Fig. 8C) (Müller & Walossek, 1987). The small Cambrian arthropod Sarotrocercus oblita appears to have a series of foliaceous postantennulary limbs formed solely by the exopodites (Whittington, 1981). The first ten pairs of postcephalic trunk limbs in Cambronatus brasseli were also interpreted as uniramous through loss of the telopodites (Briggs & Bartels, 2001), although there is a possibility that a foliaceous telopodite is retained. These are, however, exceptions and it is safe to generalise that most uniramous limbs in arthropods have arisen through loss of the exopodite. VIII. DIFFERENTIATION WITHIN THE SEQUENCE OF POSTANTENNULARY LIMBS (1 ) The first postantennulary limb The first postantennulary limb in arthropods has two main names : in Crustacea it is the second antenna or just antenna, in Chelicerata it is the pedipalp. In Hexapoda and Myriapoda it is missing ; its absence serving as a diagnostic character of the taxon Atelocerata established by Heymons Geoff A. Boxshall (1901). However, the antennal segment is represented by the limbless intercalary segment. The intercalary segment is unequivocally marked by engrailed (en) expression in the epidermal and neural cells in hexapods and its existence is further supported by the expression pattern of the Hox gene labial (lab) in insects (Diederich et al., 1989 ; Peterson et al., 1999). The symphylan Hanseniella agilis Tiegs shows a transient pair of limb buds on the antennal segment of the sevenday old embryo (Tiegs, 1940: text-fig. 20). The buds are lost on the eighth day and are absent in the adult. Similarly transient antennal limb buds have been reported in the embryos of some hexapods (Wheeler, 1893 ; Tamarelle, 1984). The antenna of crustaceans is plesiomorphically biramous and bears a feeding endite, the enditic process (Fig. 8A), on the proximal segment (=the coxa) of the two-segmented protopodite during the early naupliar phase in copepods, cephalocarids, branchiopods and most other taxa that retain non-lecithotrophic nauplii. In crown-group Crustacea the exopodite is typically multisegmented and the telopodite typically comprises only four expressed segments (Fig. 8 A). In Agnostus pisiformis the uniramous antenna has a multisegmented exopodite (Fig. 8C) comprising a proximal region of three larger and more robust segments, and a distal region of six short segments each of which bears one or more outer setae (Müller & Walossek, 1987). A six-segmented exopodite is found in the chelicerate Offacolus kingi (Sutton et al., 2002). Exopodites may also be unsegmented, as in remipedes (Fig. 8 E). In the Conchostraca both rami are long and flagellum-like (Martin, 1992), however both rami are multi-segmented with all segments containing at least some intrinsic muscles (Fig. 8 F, G). Across the Arachnida the pedipalp is typically sixsegmented (Fig. 8D) but varies in form. In Scorpionida, Pseudoscorpiones and Thelyphonida the pedipalp forms the large chela modified for prey capture and immobilization, and for defence. In scorpions and pseudoscorpions it is also a major sensory structure, richly supplied with trichobothria. In spiders the pedipalp typically retains a welldeveloped gnathobase on the proximal segment (Fig. 8 D). It is involved in feeding and, in addition, in male spiders it can be highly modified for the storage and transfer of sperm. The retention of a protopodal gnathobase (which is, confusingly, usually referred to as the maxilla) on the pedipalp is shared by other arachnids, such as the Opiliones. In the Ricinulei and Uropygi the protopodal segments appear fused in the midline, forming a common coxosternal plate, the anterior edge of which represents the gnathobasic margin. This example of ‘ fusion ’ between members of a limb pair requires further anatomical analysis to determine the extent of involvement of the ventral sternite. In xiphosurans and eurypterids (Fig. 8B) the first postantennulary limb is one of an homonomous set of five walking limbs. In the xiphosuran Limulus polyphemus the transient expression of a focus of Dll on the first postantennulary limb (Mittmann & Scholtz, 2001) is interpreted here as representing the exopodite (see Section V.3 above). Similarly, in trilobites the first postantennulary limb exhibits no unique morphological specializations as an antenna. Some Cambrian fossils have a modified first postantennulary limb. It The evolution of arthropod limbs forms the principal appendage in Branchiocaris pretiosa (Resser), which is uniramous, seven-segmented and terminates in a spine. Briggs (1976) interpreted this limb as having a food-gathering role. In Marrella splendens these limbs, referred to as second antennae, are interpreted as swimming appendages by Garcı́a-Bellido & Collins (2001). The problematic Canadaspis perfecta, as re-interpreted by Hou & Bergström (1997), also has only one modified postantennulary limb – the limb known only from a cluster of gnathobasic spines and wrongly interpreted as the ‘mandible ’ by Briggs (1978). In other Cambrian taxa the antenna is modified by loss of the exopodite, as in Sidneyia inexpectans ( Bruton, 1981), or by loss of the telopodite, as in Agnostus pisiformis (Müller & Walossek, 1987). In the Devonian Mimetaster hexagonalis the uniramous first postantennulary limb is long and well developed (Stürmer & Bergström, 1976). It comprises a broader proximal part with three or four subdivisions and an elongate distal part consisting of four long and four short ‘ segments ’ plus a terminal claw. This limb is interpreted here as comprising a proximal protopodite with three apparent subdivisions (which are probably not true segments), and a telopodite consisting of four elongate segments, a tarsal region of four annuli, and a terminal pretarsus (=claw) (Fig. 7 A). The triangular basal segment reconstructed by Lehmann (1950) is here regarded as part of the body wall. A distal annulated flagellum is present on the tip of the pedipalp in the pycnogonid Nymphonella tapetis. (2 ) The second postantennulary limb The second postantennulary limb in mandibulate arthropods is called the mandible. The possession of a mandible has been regarded as sufficient basis for the recognition of a taxon, the Mandibulata, comprising the Crustacea, Hexapoda and Myriapoda (Snodgrass, 1938). The emerging Hox gene data on chelicerates (Telford & Thomas, 1998 ; Damen et al., 1998) now suggests that the appendage carried on the homologous segment in this taxon is a walking leg (Averof, 1998 ; Akam, 2000). This limb has, by convention, been referred to as the first walking leg in arachnids and as the second walking leg in Xiphosura (Fig. 9 E) and Eurypterida (which lack differentiated pedipalps). In those taxa, such as trilobites, with an homonomous postantennulary limb series, the second postantennulary limb looks just like the first and like all other members of the series. It exhibits no unique morphological specializations as a mandible. The second postantennulary limb simply exhibits a differently modified morphology according to group. Manton (1964) reviewed mandibular mechanisms across the Arthropoda and concluded that there were two kinds of mandibles : one in which ‘the biting structures are developed from a proximal endite or gnathobase ’ (found in Crustacea and Chelicerata), and the other in which ‘ the mandible is developed from a whole limb, the tip of which and not the base is used for gnathal purposes’ (found in Myriapoda and Hexapoda=Atelocerata, according to Manton, 1964). The latter condition was referred to as telognathy. Manton’s (1964) interpretations of the embryology of the hexapod mandible have been questioned (Panganiban, Nagy & 287 Carroll, 1994) and recent data on the expression pattern of the Hox gene Dll have demonstrated that insectan mandibles, like those of adult malacostracan and branchiopodan crustaceans, are gnathobasic and protopodal in origin (Popadić et al., 1996; Prpic et al., 2001) and lack a palp. The gnathobasic nature of the mandible in hexapods is confirmed by the lack of Dll expression. In the diplopod Oxidus gracilis (C. L. Koch), Dll is initially expressed in the distal part of the mandible, is then displaced laterally by allometric growth of the gnathobase, and is eventually lost (Popadić et al., 1996; Scholtz, 2001). The whole-limb mandible as envisaged by Manton (1964) is not found within the Arthropoda. A whole-limb jaw is found in the Onychophora but this is not a mandible. The subdivision of the mandible of diplopods into cardo, stipes and gnathal lobe is therefore interpreted to be secondary (Lauterbach, 1972; Staniczek, 2000), and not indicative of original segmentation. The presence of intrinsic muscles within the mandible of Myriapoda (Rilling, 1960 ; Fechter, 1961) is interpreted here as a plesiomorphic feature retained from the postantennulary limb of ancestral mandibulates. The gnathobasic nature of the mandible in adult branchiopods, cephalocaridans and malacostracans has never been in question because both groups express a functional distal mandibular palp in earlier stages of development. Most other crustacean taxa either retain a mandibular palp into the adult (Fig. 9F), or show a similar loss of the palp in the postnaupliar phase. Nothing is known of the early development of remipedes, which lack a palp in the adult. The mandible of some halocyprid ostracods is unusual because of the well-developed endite on the basis (Fig. 9 B). In most ostracods, as in crustaceans in general, the coxa carries a powerful gnathobase and the basis lacks an endite, forming instead the common stem of the biramous palp (Fig. 9 F). The loss of the mandibular palp is shared by Hexapoda (Fig. 9 G, H) and Myriapoda (Fig. 9 C, I) but may well be homoplasic given the propensity for loss exhibited even within the Crustacea. Within the Hexapoda the dicondylar mandible (Fig. 9 H) is synapomorphic for the Zygentoma plus the Pterygota (=the Dicondylia), although the condition is not fully developed in Zygentoma and Ephemeroptera (Fig. 9G). The mandibular gnathobase in Crustacea is formed from the proximal segment only of the two protopodal segments (i.e. the coxa). The palp, which may be lost during development, comprises the distal protopodal segment (the basis) plus the telopodite and exopodite (Fig. 9 A, F). Accepting the homology of the mandible across the Mandibulata, requires that the same gnathobasic origin is hypothesised for Hexapoda and Myriapoda, i.e. the mandibular gnathobase is formed by the proximal part only (=coxa) of a subdivided protopodite. Chelicerates, trilobites, Agnostus pisiformis (Fig. 9D) and all other Cambrian arthropods (except possibly Sidneyia inexpectans) retain an entire and undivided protopodite on the second postantennulary limb, so the gnathobase of this limb is not homologous with the gnathobase of Mandibulata (as pointed out by many authors). Machida (2000) compared the development of the mandible and third postantennulary limb (maxilla) of the machilid Pedetontus unimaculatus Machida, and homologized the molar 288 and incisor with the maxillary lacinia and galea, respectively. This proposal is rejected here since it would require molar and incisor to be derived from separate protopodal segments and there is no supporting evidence for this in hexapodan mandibular musculature. In addition, the concept of the Mandibulata as a taxon is built on the homology of the mandibular gnathobase which, as demonstrated by Crustacea, is derived from the proximal segment of a two-segmented protopodite. In insectan Hexapoda and in malacostracan and cephalocaridan Crustacea, the expanded inner margin of the mandibular gnathobase is typically divided into apical incisor and subapical molar regions. Molar and incisor regions differ in form and function. The elaboration of the molar region into so-called molar hooks was identified by Kraus (2001) as already present in the groundplan of the hexapods and myriapods (as the Tracheata) and he tentatively suggested that the presence of molar hooks was a synapomorphy of the Mandibulata as a whole. Differentiation into molar and incisor processes is absent in crustacean groups such as the Copepoda and Ostracoda, and in the early developmental stages of other taxa, including stem-lineage forms. If this is a plesiomorphic adult state, rather than a neotenic condition, then this does not support Kraus’s (2001) suggestion. The mandibular gnathobase is armed with an array of blades or teeth along the inner margin. In phosphatocopines and crustaceans the gnathobase also typically carries at least one articulated seta and may carry additional rows of articulated spines. The marginal teeth appear to form as surface outgrowths rather than as articulated setation elements. Manton’s (1928) hypothesis that the lacinia mobilis, located near the base of the incisor process on the mandible of peracarid crustaceans is derived by the secondary articulation of a marginal tooth (cusp b) was rejected by Dahl & Hessler (1982). They provided new evidence suggesting that the lacinia mobilis is derived from the distalmost member of the left spine row present in eumalacostracans. Richter, Edgecombe & Wilson (2002) rejected the homology between the lacinia mobilis of Peracarida and its alleged homologues, the prostheca of Hexapoda, internal tooth of Diplopoda and ‘lacinia mobilis ’ of Symphyla and Remipedia. (3 ) The third postantennulary limb The fourth limb – the first postmandibular limb – is referred to as the maxilla in Hexapoda and in Chilopoda, as the gnathochilarium in Diplopoda and Pauropoda, as the first maxilla or maxillule in Crustacea, and as the second walking appendage in Chelicerata. The comparative anatomy of the maxillule across the major crustacean groups was studied by Boxshall (1997). The limb is plesiomorphically biramous but may be uniramous, as in the powerful subchela of remipedes, or may even be modified as a strong muscular sucker, as in adults of the branchiuran genus Argulus Linnaeus. Boxshall (1997) concluded that the maxillule of the crown-group crustacean groundplan comprised a two-segmented protopodite, a four-segmented telopodite and an unsegmented lamellate exopodite. Each segment (coxa and basis) of the protopodite Geoff A. Boxshall originally carried two endites except in malacostracans, which exhibit the derived state of only one endite per protopodal segment. Copepods and ostracods are the only crustaceans to carry exites on the protopodite of the maxillule (Boxshall, 1997). Copepods have two maxillulary exites, one (the epipodite) on the coxa and one on the basis (Huys & Boxshall, 1991). The subsegmentation of the proximal protopodal segment of the maxillule into precoxa and coxa occurred within the Crustacea, in maxillopodans and remipedes (Boxshall, 1997). The composition and homology of the gnathochilarium in Diplopoda (and Pauropoda) is controversial. According to O. Kraus and M. Kraus (Kraus & Kraus, 1994 ; Kraus, 1997, 2001) the skeletomusculature of the gnathochilarium is in accord with the hypothesis that it comprises both the third and fourth postantennulary limbs (maxillules and maxillae). The opposing, and more widely accepted, view (cf. Klass & Kristensen, 2001) is that the gnathochilarium in these taxa is formed solely by the maxillules, as indicated by the neuromusculature (Dohle, 1964, 1997). In diplopods the sternite of the maxillulary segment grows out to form a plate connecting the paired limb buds (Dohle, 1997). The common proximal part for the gnathochilarium is therefore a coxosternite. In pauropods the paired limbs and the sternite do not coalesce. In chilopods the first maxillae comprise an undivided protopodite, often fused in the midline to form a common coxosternite, and a short telopodite. The telopodite is at most two-segmented (Borucki, 1996). The protopodite carries one or two inner lobes (coxal lobes 1 and 2) which might represent vestigial endites. The hexapod maxilla has been discussed above (see Section V.2). ( 4) The fourth postantennulary limb The fifth limb – the second postmandibular limb – is referred to as the labium in Hexapoda, as the second maxilla or labium in different groups of Myriapoda, as the second maxilla or just maxilla in Crustacea, and as the third walking leg in Arachnida. The term labium is best avoided as it is commonly used for a single structure, not reflecting its origin as a paired limb. In addition labium is also used for a diversity of non-homologous structures on arthropodan heads : for example the labium or lower lip of siphonostomatoid copepods represents the fused paragnaths (Boxshall, 1990), while the labium of spiders represents the sternite of the second prosomal segment (Savory, 1977). The fourth postantennulary limb (=second maxilla) is not expressed in Diplopoda and Pauropoda, as indicated by the absence of embryonic limb buds on the segment (Dohle, 1964, 1997). In Chilopoda the second maxillae comprise an undivided protopodite, often fused in the midline to form a common coxosternite, and a telopodite (=palp). Borucki (1996) showed, from study of the musculature, that a maximum of five segments is expressed in the chilopod palp : trochanter, prefemur/femur, tibia, tarsus and pretarsus. In insects the labium (=fourth postantennulary limbs) is formed by the paired appendages fused in the midline to form a median plate. The proximal part of the labium, the postmentum, incorporates the sternite and may represent a coxosternite. It is often subdivided into submentum and The evolution of arthropod limbs mentum. The extrinsic limb musculature typically passes through the entire postmentum to insert on the proximal rim of the prementum. The prementum is protopodal in origin. It carries the endites (paired glossae and paraglossae) and the telopodites (=palps), and contains well-developed intrinsic musculature. The homology of the postmentum is uncertain. (5 ) The fifth postantennulary limb The sixth limb – the third postmandibular limb – is identified as the first thoracic or trunk limb (thoracopod) in Hexapoda, Myriapoda and Crustacea. It is the fourth walking leg (=sixth prosomal limb) of Arachnida and the swim paddle of eurypteroid eurypterids. In stylonuroid eurypterids the sixth limb telopodite is walking leg-like according to Braddy (1996, cited in Dunlop & Braddy, 2001). In several crustacean groups it is functionally a member of the cephalic limb series, modified as an accessory feeding limb, the maxilliped. In Chilopoda these limbs are called forcipules. Averof & Patel (1997) showed that two Hox genes [Ultrabithorax (Ubx) and abdominalA (AbA)] were expressed throughout the postcephalic trunk region in crustaceans, such as anostracan branchiopods and leptostracan malacostracans, that lack differentiated maxillipeds. In different lineages of crustaceans with one or more pairs of postcephalic trunk limbs modified as maxillipeds, there is a posterior shift in the anterior boundary of embryonic expression of Ubx–abdA. Averof & Patel (1997) noted a striking correlation between the expression of Ubx–abdA and differentiation of the limbs, which they interpreted as evidence of a direct role of Hox genes in the specification of segmental identity. In Xiphosura the flabellum on the sixth prosomal limb is interpreted here as a vestige of the exopodite (see Section V.3 above). It appears relatively late in development, in embryos between 18 and 21 days old (Patten, 1896: pl. II, Fig. 6), and is located lateral to the main ramus (=telopodite) of the developing limb. Mittmann & Scholtz (2001) showed that the second to fourth postantennulary limbs of Limulus polyphemus all showed exopodal vestiges in a serially homologous position. Sutton et al. (2002) interpreted the tendril-like exopod of the sixth prosomal limb of the Silurian Offacolus kingi as a possible homologue of the flabellum. The maxillipeds or forcipules of chilopods carry the poison fangs. The protopodites (=coxae) of the limb pair may be separate but are often fused in the midline to form a common coxosternite. The antero-medial margin of each coxa in scolopendromorphs is produced into a toothed coxal plate which may represent an enditic margin. The palp comprises a maximum of six segments in Scutigeromorpha but in Pleurostigmomorpha there are only five segments, identified by Borucki (1996) as trochanter, prefemur, femur, tibia and tarsungulum. The tarsungulum is the highly sclerotised fang. In some chilopods the inner margin of the prefemur carries a toothed spinous process. The Devonian chilopod Devonobius delta has separate coxae and a fivesegmented maxillipedal palp (Shear & Bonamo, 1988; Borucki, 1996). In amphipod crustaceans the maxillipeds are carried on a common proximal plate formed by medial fusion of 289 the proximal protopodal segments and possibly also involving the sternite. Although fused, the coxae each carry an endite. (6 ) The sixth postantennulary limb The seventh limb – the fourth postmandibular limb – is identified as the second thoracic limb (thoracopod) in Hexapoda, Myriapoda and Crustacea. The seventh limbs are modified as the chilaria of Xiphosura and the metastomatic plate of Eurypterida. In the Devonian xiphosuran Weinbergina opitzi Richter & Richter the seventh limb is a walking leg whereas in the Silurian chelicerate Offacolus kingi it is flap-like (Sutton et al., 2002) and overlies the first opisthosomal limb. In extant arachnids this body segment is extremely reduced in the adult, or is represented by the pedicel uniting prosoma and opisthosoma, and is limbless. However, during embryonic development of scorpions, the seventh limbs appear transiently as laterally directed buds, which are later lost (Laurie, 1890). In most subgroups within the Malacostraca, this limb is functionally a member of the cephalic limb series, modified as an accessory feeding limb, the second maxilliped. In most crustaceans, such as copepods and branchiopods, this limb is a typical trunk limb performing functions related to locomotion, feeding or both. (7 ) The seventh and subsequent postantennulary limbs In adult Arachnida, opisthosomal segments carry a diverse array of possible limb derivatives (cf. Van der Hammen, 1989), although most are not obvious. Limbs may also make a transient appearance during development. The eighth body segment (=first opisthosomal segment) of arachnids carries the genital openings, which may be covered by genital opercula. The ninth body segment carries the pectines in scorpions and the subsequent two body segments (tenth and eleventh) carry limbs modified as spinnerets in the liphistiomorph spiders. The lung books present on the third to sixth opisthosomal segments of scorpions represent modified limbs, derived from gill book-bearing opisthosomal limbs similar to those found in Eurypterida (cf. Dunlop & Braddy, 2001), Xiphosura and Offacolus kingi. Embryological studies have shown that the lung lamellae of scorpions and spiders are formed as proximal folds on the posterior surface of the embryonic limb (e.g. Dawydoff, 1949) confirming Lankester’s (1881) original interpretations. The eversible vesicles of Collembola and Protura are not homologues of the eversible vesicles present on the protopodal parts of the limbs of progoneates, Diplura and Archaeognatha, being derived from the telopodite of the limb in both cases (Klass & Kristensen, 2001). The first abdominal limb (=eighth postantennulary limb) of Protura (Fig. 7 H) is uniramous and comprises two segments plus the eversible tip (François, 1969) which may represent a third segment (Klass & Kristensen, 2001). However, the proposed homologies of coxa, trochanter and femur plus presumptive distal segments require confirmation since they 290 are not congruent with available data on hexapod leg development. Bretfeld (1963), for example, showed that the three-segmented stage of leg development in the collembolan Orchesella villosa corresponded to protopodite (coxa), trochanter/femur (tr/fe) and tibia/tarsus (ti/ta). The fifteenth, sixteenth and possibly seventeenth postantennulary limbs can be well developed as external genitalia in hexapods. In females, the gonapophyses, which may represent modified endites, are typical components of a sometimes elaborate egg-laying apparatus. Malacostracan crustaceans can exhibit specialised limbs along the length of the postcephalic trunk. The seventh postantennulary limb is specialised as a third maxilliped in Decapoda and the eighth pair forms the large prominent claws (chelipeds) of the brachyuran crabs. The more posterior thoracic legs (peraeopods) tend to be specialised as walking legs and the abdominal limbs as swimming legs (pleopods). The pleopods can be further specialised with three posterior pairs forming posteriorly directed uropods in amphipods. These specializations all occurred within the Malacostraca and should not influence our interpretations of the crustacean groundplan. IX. HETERONOMY Full analysis of the origin of heteronomy in the postantennulary limb series in different arthropodan lineages would require detailed consideration of tagmosis, which is beyond the scope of this review. However, it is possible to identify three main theoretical patterns in limb differentiation commencing with the groundplan identified above – an antennule followed by an homonomous series of postantennulary limbs. The first is differentiation along the anterior to posterior axis, commencing with the first postantennulary limb. This is essentially cephalization and can be viewed as a progressive process involving the differentiation of increasing numbers of appendages in different lineages. The posterior shift in limb differentiation is presumed to be driven by changes in expression pattern of Hox genes, since Averof & Patel (1997) found that cephalization of anterior thoracic limbs correlated well with patterns of expression of the Hox genes Ubx and abdA. The second is differentiation along the posterior to anterior axis, commencing with the last trunk limb. This is linked to the process of abdominalization which typically involves loss of limbs from posteriorly located body segments and/or the differentiation of the caudalmost appendages, often for functions related to reproduction. It may involve adjacent limbs but is less progressive than cephalization. It is referred to here as caudalization. The third pattern is biphasic and describes non-progressive differentiation of an original postantennulary limb series into anterior and posterior blocks. Each block comprises several similar limb pairs but the limbs differ between blocks. These processes may interact, so that a biphasic pattern may be modified by subsequent cephalization within the anterior block. There are numerous examples of cephalization involving one, two, three, four, or more pairs of postantennulary Geoff A. Boxshall limbs. Indeed, cephalization is one of the dominant processes underlying the trend towards increasing diversity of arthropodan limb types in the Palaeozoic, as documented by Cisne (1974). Several Palaeozoic arthropods have just one pair of postantennulary limbs (often referred to as the second antenna) differentiated from the more posterior members of the series. These include Branchiocaris pretiosa and Marrella splendens, both of which have a uniramous first postantennulary limb (Whittington, 1971 ; Briggs, 1976), and Ercaia minuscula which has a modified but biramous antenna (Chen et al., 2001). According to the reconstruction of Hou & Bergström (1997), the Chengjiang arthropod Fortiforceps foliosa Hou & Bergström has the first pair of postantennulary limbs only, modified as uniramous grasping antennae. Arthropods with only two pairs of modified postantennulary limbs include the Labrophora, a taxon of stemline Crustacea. The marrellomorphan Mimetaster hexagonalis has two pairs of postantennulary limbs modified by loss of the exopodite. Arthropods with three pairs of modified postantennulary limbs include the Crustacea. In Burgessia bella (Hughes, 1975) the outer branch of the first three pairs of postantennulary limbs is modified and flagellum-like, rather than gill filament-like. Sidneyia inexpectans also has these three limbs modified by loss of the exopodite (outer gill branch) according to Bruton (1981), as does the Devonian Vachonisia rogeri (Lehmann) according to Stürmer & Bergström (1976). Arthropods with four pairs of modified postantennulary limbs include the majority of crown-group crustaceans, hexapods and myriapods (although one or more of these pairs is missing in the last two groups). Five pairs of modified postantennulary limbs are found in those crustaceans with a maxilliped (e.g. remipedes, copepods and most malacostracans) and in chilopods where the fifth pair forms the forcipules. These few examples of cephalization involving from one to five or more pairs of postantennulary limbs do not represent steps in a single linear process of cephalization associated with the sequential incorporation of trunk segments into the cephalon. Instead they are selected as independent examples which illustrate a widespread evolutionary process. At the opposite pole of the body, caudalization involves limb differentiation along the posterior to anterior axis, commencing with the last trunk limb. The last pair of trunk limbs in many crustacean taxa (e.g. anostracan Branchiopoda, Thecostraca, Tantulocarida, Copepoda and Branchiura) is often modified, at least in males, for a role in reproduction. In Burgessia bella the posterior limb is uniramous and spiniform. The posteriormost trunk limbs in Chilopoda, in pentazonian Diplopoda, and in most Hexapoda are modified as gonopods and perform a variety of functions related to mating. Whereas the apparently sequential pattern of limb differentiation in mandibulate groups is in accord with an anterior to posterior model (=cephalization) outlined above, the pattern exhibited in members of the chelicerate lineage seems best described as biphasic. Fossil arthropods recognised as having an affinity with the chelicerates tend to have only two types of postantennulary limbs : an The evolution of arthropod limbs anterior group of six limbs with walking/raptorial telopodites and a posterior group of lamellate limbs. This would correspond with a tagmosis pattern of an anterior prosoma bearing chelicerae (=antennules) plus six pairs of appendages, and an opisthosoma bearing up to a dozen or so limb pairs. Sanctacaris uncata, as reinterpreted here (see Section III.1 above), conforms to this pattern almost perfectly although there is some anterior to posterior differentiation superimposed upon the homonomous prosomal block of postantennulary limbs, in that the more anterior members of this series have fewer expressed telopodal segments than the posterior members, according to Briggs & Collins (1988). A similar pattern appears to be found in the Silurian chelicerate Offacolus kingi. Offacolus kingi has seven prosomal limbs (Sutton et al., 2002), of which the first (chelicerae) and the seventh are uniramous but the other five pairs are biramous and uniform in basic structure. The limbs of the posterior block are essentially lamellate and appear to carry gill books posteriorly as in xiphosurans. The Devonian xiphosuran Weinbergina opitzi has a welldefined prosoma which carries the chelicerae followed by six pairs of similar walking legs (Stürmer & Bergström, 1981). Extant Xiphosura conversely show only some posterior to anterior differentiation superimposed upon the otherwise homonomous postantennulary prosomal limbs, with the sixth pair modified as chilaria. Other modern and fossil chelicerates can show further specialization of the prosomal limbs : the first postantennulary limbs are modified as pedipalps in Pycnogonida and Arachnida. Some eurypterids, such as Mixopterus kiaeri Størmer, have the first and second postantennulary limbs modified for grasping while the next two pairs are typical walking legs as present in Baltoeurypterus tetragonopthalmus (Fischer). This secondary cephalization is imposed on a basic biphasic pattern. Scorpions and other extant arachnids exhibit profound modifications of the opisthosomal limbs. In scorpions, for example, there can be six pairs of opisthosomal limbs expressed in various forms (as genital operculum, pectines and lung books 1–4) (Dunlop & Braddy, 2001). The precise modifications in arachnids vary with order (Shultz, 1990). Eurypterids retain a more or less homonomous series of limbs on the opisthosoma, with five pairs of foliaceous limbs (Blattfüsse) enclosing the gill chambers, as in Baltoeurypterus tetragonopthalmus (Holm, 1898) and Tarsopterella scotica (Woodward) (Waterston, 1975). The basic biphasic arrangement of the postantennulary limbs into two homonomous series (anterior and posterior blocks) is clearly expressed in early members of the chelicerate lineage. In Palaeozoic forms it is little modified by specialization within either block and provides virtually no support for a model of either anterior to posterior, or posterior to anterior, differentiation. The biphasic model under the control of Hox genes is proposed here as the most appropriate explanation of chelicerate tagmosis. Pycnogonids appear to be an exception to this biphasic model. The possession of chelicerae and pedipalps is accepted here as synapomorphic with crown-group chelicerates. The typical pycnogonid has seven pairs of limbs: chelicerae, pedipalps, ovigers and walking legs 1–4. Both 291 pedipalps and ovigers are walking-leg-like. It has no opisthosomal limbs and virtually no opisthosoma. This pycnogonid conforms to the biphasic model proposed here, assuming that the entire posterior limb block is lost. However, the polymerous pycnogonids, the four genera with five pairs of walking legs and the two genera with six pairs, appear to constitute an exception to the biphasic model. Both groups of genera have more than seven pairs of prosomal-type limbs. There are three possible types of explanation : either the Pycnogonida separated from the chelicerate lineage prior to the evolution of the biphasic limb pattern, or the first two pairs of opisthosomal limbs have been secondarily modified to resemble prosomal limbs in these genera, or there has been segmental duplication at the posterior end of the prosoma, resulting in the secondary formation of either one or two supernumerary limb-bearing segments. Whichever explanation is correct, the classification of the five-legged pycnogonids in four different families and the six-legged ones in two families implies considerable homoplasy. The appearance of the biphasic limb pattern in Cambrian relatives of the chelicerates, such as Sanctacaris uncata, possibly prior to the evolution of chelicerae [by inference from the phylogenetic scheme of Dunlop & Selden (1997)] casts doubt on the first explanation, although it would be consistent with the scheme of phylogenetic relationships presented by Giribet et al. (2001). The reduction of the telopodite of the opisthosomal limbs in all chelicerates and their relatives casts doubt on the second explanation. The last explanation, that there has been segmental duplication at the posterior end of the pycnogonid prosoma, is favoured here. Given the degree of homoplasy involved, it seems likely that the genetic mechanisms controlling such hypothetical duplication are relatively simple. Analysis of pycnogonid segmentation genes should provide the evidence required to resolve this question. X. SEGMENTATION AND MUSCULATURE The distinction made here between limb segments and annuli is based on features of their musculature : true segments being characterised by the presence of intrinsic muscles. An analogous distinction, between eosegments and merosegments, was made by Minelli et al. (2000) in their consideration of the development of chilopod body and antennulary segmentation patterns. Their eosegments are the primary segmental units, each giving rise to merosegments by subsequent subdivision. Minelli et al. (2000) did not explicitly describe the antennulary musculature, but if the eosegments are reflected in the intrinsic musculature and the merosegments are not, then these morphological concepts could equate directly to segments and annuli as defined here. The distinction between chilopod eosegments and merosegments is reflected in ontogeny: eosegments appearing before merosegments. Is relative developmental timing important in distinguishing between segments and annuli ? There are some examples suggesting a close correlation Geoff A. Boxshall 292 between timing and development of segments/annuli. For example, Nayrolles (1991) showed that the symphypleone collembolans with a terminal flagellum on the antennule, first express the four true segments prior to the onset of annulation that generates the apical flagellum. By contrast, in copepod crustaceans the adult antennulary segments are derived by a process of subdivision of compound segments defined at earlier stages (Fig. 3). Whether appearing early or late in development, all 28 expressed copepod antennulary segments are primitively defined by intrinsic musculature. The sequence of appearance of walking leg segments during development of apterygote and hemimetabolous Hexapoda has been described as involving three steps (Bitsch, 2001). Commencing with simple lateroventral limb buds, ‘ primary’ segmentation involved the differentiation into a basal part (coxa), a middle part (trochantero-femur) and distal part (tibia-tarsus-pretarsus complex). ‘Secondary ’ segmentation separated trochanter and femur, and tibia, tarsus and pretarsus. Finally ‘ tertiary ’ segmentation resulted in subdivision of the tarsus. The first two steps both result in the formation of true segments separated by articulations and defined by intrinsic musculature. The last, tarsal subdivision, is an annulation process since tarsal segments are not muscular, although tendons effecting pretarsal flexure pass through them. The mode of limb development, characterised by the expression of intersegmental articulations along a limb axis that first appears in embryology as an undivided lobe, is retained in most arthropod groups. The notable exception is the holometabolous insects, in which the legs form as imaginal discs. The presumptive protopodite and telopodal segments first appear as concentric rings within the imaginal discs and the limb undergoes elongation and further differentiation into the imaginal limb during metamorphosis. This process is apomorphic compared to the typical mode of development and must result from heterochronic changes in the timing of expression of articulations, so that the original sequential appearance is condensed. Although data from available developmental studies are somewhat fragmentary, it is apparent that annuli tend to appear later in arthropodan development than limb segments. Whether the process of subdivision of developing limb segments results in the production of another generation of segments or in the production of annuli, may therefore depend upon precise ontogenetic timing. The critical events being the onset of formation of the intersegmental articulation complete with arthrodial membrane, and attachment of the pioneering muscle cells to the exoskeleton. Relative shifts in timing of these events might then be invoked to explain examples of secondarily segmented structures, such as barnacle cirri, in which a multisegmented state is derived from an ancestral state with fewer (i.e. three) segments. Similarly, the derivation of the slender, cylindrical (=stenopodial) trunk limbs of the branchiopod Leptodora kindti by secondary subdivision of the protopodite (Olesen et al., 2001) presumably involves heterochronic shifts in the appearance of muscles since all the limb ‘segments ’ are defined by novel articulations, but all are characterised by the presence of intrinsic musculature. XI. CONCLUSIONS (1) There are only two basic limb types in crown-group arthropods : uniramous antennules and biramous postantennulary limbs. These limbs are serial homologues at the inclusive lobopodian/arthropodan level and the genetic transformation of antennules to walking limbs by homeotic mutation reflects only this level of homology. (2) Antennules and postantennulary limbs are not serially homologous at the level of arthropodan limbs since the basic arthropodan antennule is uniramous and lacks differentiation into protopodite and telopodite, whereas the basic arthropodan postantennulary limb is biramous with a distinct protopodite. (3) Commencing with an ancestral homonomous series of postantennulary limbs, heteronomy may result from three distinct processes, cephalization, caudalization and biphasic differentiation. The biphasic model applies only to the chelicerate lineage, but available data on the Pycnogonida are equivocal and prevent the categorization of the limb series pattern expressed in this taxon. (4) Antennules are plesiomorphically uniramous and composed of segments defined by intrinsic musculature. Biramous or biflagellate antennules originated independently on at least three occasions : in Malacostraca, Remipedia and Pauropoda. (5) Flagellate antennules have evolved independently on several occasions within the Arthropoda. There are two types : one in which the terminal segment undergoes annulation to form a long terminal flagellum composed of annuli, as found in Insecta, in the symphypleone Collembola and in both malacostracan and conchostracan Crustacea, and one in which the more proximal segments form intercalary annuli, as in some isotomid Collembola and some Chilopoda. (6) The basic postantennulary limb of crown-group arthropods is hypothesised as comprising an undivided protopodite, a telopodite of seven segments and a flattened exopodite probably of two segments (Fig. 10 A). (7) The entire medial margin of the undivided protopodite was provided with spines and formed a gnathobase. An early dichotomy in the Arthropoda was the separation of a mandibulate lineage, in which the protopodite of the postmaxillulary limbs became elongate with a subdivided medial margin bearing two or more endites (Fig. 10 B, C), from the arachnomorph groups which retained the short, undivided, gnathobasic protopodite (Fig. 10G). This elongate protopodite became subdivided within the mandibulates (Fig. 10C, D). (8) Protopodal segments often fail to be fully expressed and can be more or less incorporated into the body wall at the limb base (Fig. 10E, F), as in Hexapoda, Myriapoda and Crustacea. Segments incorporated into the body wall are often incomplete, being represented by lateral sclerites and having lost the medial enditic part. (9) The gnathobase of the mandibulate mandible is not homologous with the gnathobase of the second postantennulary limb of arachnomorphs because it is derived from the medial margin of the proximal protopodal segment (=the coxa of the crustacean mandible) only, not from the The evolution of arthropod limbs entire protopodite as in arachnomorphs. There is no evidence for a precoxa in crustacean mandibles (or antennae). (10) Exites (including epipodites) on limb protopodites appeared relatively late in the Palaeozoic and were not present in the crustacean groundplan. At least nine different and non-homologous exites can be identified in the Crustacea alone. Direct homology of insect wings and the epipodites of branchiopodan and malacostracan Crustacea is supported by embryological data as well as by gene expression data, but the suggestion of a distinct structural progenitor in the last common ancestor of crustaceans and insects is rejected partly because of uncertainty concerning precise sister-group relationships, and partly because of the lack of exites in the hexapodan sister taxon of the Insecta. (11) The ancestral arthropodan form of a sevensegmented telopodite is retained in many arachnomorphs but a six-segmented telopodite probably characterised the crown-group mandibulates (Fig. 10 B–F ). (12) Secondary increases in segmentation occur infrequently. The formation of intercalary annuli within limbs, particularly in the distal tarsal portion of the limb, is common to numerous lineages in all the major arthropodan taxa. (13) Oligomerization the reduction in number of expressed limb segments, is the most common modification in limb segmentation and usually results from failure of expression of articulations during development. Heterochrony is probably an important underlying evolutionary mechanism generating the observed oligomerization trends. (14) The ancestral form of the arthropodan exopodite remains problematic because of the great diversity expressed in the form and segmentation of this ramus from the Cambrian to the present. It is tentatively proposed that a twosegmented exopodite is the likely ancestral state (Fig. 10A). XII. ACKNOWLEDGEMENTS This review has been significantly improved by the considered criticism of the first draft by Richard Fortey, Greg Edgecombe, Alessandro Minelli, Thierry Deuve and Mark Judson. I sincerely thank all these colleagues for their help. I am also grateful to Richard Fortey (for providing advice and literature on trilobites), Jørgen Olesen (for providing illustrations and preprints of work in press), Danilo de Calazans (for providing illustrations), Greg Edgecombe (for stimulating discussion of arthropod phylogeny, for sharing his doubts about the interpretation of the antennules of Fortiforceps, and for providing literature), Doug Erwin and Elisabeth Valiulis (for providing access to Burgess Shale fossils stored in USNM), Andrew Ross (for providing literature and access to fossil arthropods stored in NHM), Mark Judson (for advice on arachnid characters), Brian Kensley (for advice on penaeid gills), Louis Deharveng (for stimulating subterranean discussions on collembolans and for providing literature), Alessandro Minelli (for many interesting discussions on arthropod body form), Derek Siveter (for information on Offacolus), Roger Bamber (for our regular Friday evening discussions on pycnogonids, and for providing specimens and literature), Phil Rainbow (for advice on thecostracan musculature), Buz Wilson (for stimulating discussion of arthropod phylogeny), Dieter Waloszek (for stimulating discussions over many years and for his comments on the manuscript), Jens Høeg (for advice on thecostracan phylogeny), Jean Vannier (for providing 293 literature and preprints of work in press) and Fred Schram (for providing preprints of work in press). XIII. REFERENCES ABZHANOV, A. & KAUFMAN, T. C. (2000). Homologs of Drosophila appendage genes in the patterning of arthropod limbs. Developmental Biology 227, 637–689. ABZHANOV, A., POPADIĆ, A. & KAUFMAN, T. C. (1999). Chelicerate Hox genes and the homology of arthropod segments. Evolution and Development 1, 77–89. AGUINALDO, A. M. A., TURBEVILLE, J. M., LINFORD, L. S., RIVERA, M. C., GAREY, J. R., RAFF, R. A. & LAKE, J. A. (1997). Evidence for a clade of nematodes, arthropods and other moulting animals. Nature 387, 489– 493. AKAM, M. (2000). Arthropods : developmental diversity within a (super) phylum. Proceedings of the National Academy of Sciences 97, 4438– 4441. ALEXANDER, R. MCN (1968). Animal Mechanics. Sidgwick & Jackson, London. 346pp. ANDERSSON, G. (1979). On the use of larval characters in the classification of Lithobiomorph centipedes (Chilopoda : Lithobiomorpha). In Myriapod Biology (ed. M. Camatini), pp. 73–81. Academic Press, London and New York. ARSENAULT, D. J., MARCHINKO, K. B. & PALMER, A. R. (2001). Precise tuning of barnacle leg length to coastal wave action. Proceedings of the Royal Society of London, Series B 268, 2149–2154. AVEROF, M. (1998). Origin of the spider’s head. Nature 395, 436– 437. AVEROF, M. & AKAM, M. (1995). Insect-crustacean relationships : insights from comparative developmental and molecular studies. Philosophical Transactions of the Royal Society of London, Series B 347, 293–303. AVEROF, M. & COHEN, S. M. (1997). Evolutionary origin of insect wings from ancestral gills. Nature 385, 627–630. AVEROF, M. & PATEL, N. H. (1997). Crustacean appendage evolution associated with changes in Hox gene expression. Nature 388, 682–686. BAGNALL, R. S. (1911). A Synopsis of the British Pauropoda. Transactions of the Natural History Society of Northumberland, Durham and Newcastle-upon-Tyne III, 654 –660, pl. XIX. BARLET, J. (1988). Considérations sur le squelette thoracique des insectes aptérygotes. Bulletin et Annales de la Société royale belge d’Entomologie 124, 171–187. BASLER, K. & STRUHL, G. (1994). Compartment boundaries and the control of Drosophila limb pattern by hedgehog protein. Nature 368, 208–214. BERGSTRÖM, J. (1979). Morphology of fossil arthropods as a guide to phylogenetic relationships. In Arthropod Phylogeny (ed. A. P. Gupta), pp. 3–56. Van Nostrand Reinhold Co., New York. BERGSTRÖM, J. & HOU, X. (1998). Chengjiang Arthropods and Their Bearing on Early Arthropod Evolution. In Arthropod Fossils and Phylogeny (ed. G. D. Edgecombe), pp. 151–184. Columbia University Press, New York. BERLESE, A. (1908). Osservazioni intorno agli Acerentomidi. Redia 5, 110–122. BISHOP, S. A., KLEIN, T., MARTINEZ ARIAS, A. & COUSO, J. P. (1999). Composite signalling from Serrate and Delta establishes leg segments in Drosophila through Notch. Development 126, 2993–3003. BITSCH, C. & BITSCH, J. (2000). The phylogenetic interrelationships of the higher taxa of apterygote hexapods. Zoologica Scripta 29, 131–156. 294 BITSCH, J. (1956). Considerations anatomiques sur le palpe maxillaire des Machilides. Travaux du Laboratoire de Zoologie et de la Station Aquicole Grimaldi de la Faculté des Sciences de Dijon 15, 1–9. BITSCH, J. (1994). The morphological groundplan of Hexapoda : critical review of recent concepts. Annales de la Société Entomologique de France (N.S.) 30, 103–129. BITSCH, J. (2001). The hexapod appendage : basic structure, development and origin. In Origin of the Hexapoda (ed. T. Deuve). Annales de la Société Entomologique de France (N.S.) 37, 175–193. BLOWER, J. G. (1978). Anamorphosis in the Nematophora. Abhandlungen und Verhandlungen des Naturwissenshcaftlichen Vereins in Hamburg 21/22, 97–103. BOORE, J. L., LABVROV, D. V. & BROWN, W. M. (1998). Gene translocation links insects and crustaceans. Nature 392, 667–668. BORUCKI, H. (1996). Evolution und Phylogenetisches System der Chilopoda (Mandibulata, Tracheata). Verhandlungen des Naturwissenschaftlichen Vereins in Hamburg (NF) 35, 95–226. BOXSHALL, G. A. (1982). On the anatomy of the misophrioid copepods, with special reference to Benthomisophria palliata Sars. Philosophical Transactions of the Royal Society of London, Series B 297, 125–181. BOXSHALL, G. A. (1985). The comparative anatomy of two copepods, a predatory calanoid and a particle feeding mormonilloid. Philosophical Transactions of the Royal Society of London, Series B 311, 303–377. BOXSHALL, G. A. (1990). The skeletomusculature of siphonostomatoid copepods, with an analysis of adaptive radiation in structure of the oral cone. Philosophical Transactions of the Royal Society of London, Series B 328, 167–212. BOXSHALL, G. A. (1997). Comparative Limb Morphology in Major Crustacean Groups : the coxa-basis joint in postmandibular limbs. In Arthropod Phylogeny (eds. R. A. Fortey and R. Thomas), pp. 155–167. Chapman & Hall, London. BOXSHALL, G. A. & BÖTTGER-SCHNACK, R. (1988). Unusual ascothoracid nauplii from the Red Sea. Bulletin of the British Museum (Natural History) (Zoology series) 54, 275–283. BOXSHALL, G. A. & HUYS, R. (1998). The ontogeny and phylogeny of copepod antennules. Philosophical Transactions of the Royal Society of London, Series B 353, 765–786. BRADDY, S. J. (1996). Palaeobiology of the Eurypterida. Unpublished Ph.D. thesis, University of Manchester, UK. BRETFELD, G. (1963). Zur Anatomie und Embryologie der Rumpfmuskulatur und der abdominalen Anhäge der Collembolen. Zoologische Jahrbücher, Abteilung für Anatomie und Ontogenie der Tiere 80, 309–384. BRIGGS, D. E. G. (1976). The arthropod Branchiocaris n. gen., Middle Cambrian, Burgess Shale, British Columbia. Bulletin of the Geological Survey of Canada 264, 1–29. BRIGGS, D. E. G. (1978). The morphology, mode of life, and affinities of Canadaspis perfecta (Crustacea : Phyllocarida), Middle Cambrian, Burgess Shale, British Columbia. Philosophical Transactions of the Royal Society of London, Series B 281, 439– 487. BRIGGS, D. E. G. & BARTELS, C. (2001). New arthropods from the Lower Devonian Hunsrück Slate (Lower Emsian, Rhenish Massif, Western Germany). Palaeontology 44, 275–303. BRIGGS, D. E. G., BRUTON, D. L. & WHITTINGTON, H. B. (1979). Appendages of the arthropod Aglaspis spinifer (Upper Cambrian, Wisconsin) and their significance. Palaeontology 22, 167–180. BRIGGS, D. E. G. & COLLINS, D. (1988). A Middle Cambrian chelicerate from Mount Stephen, British Columbia. Palaeontology 31, 779–798. Geoff A. Boxshall BRIGGS, D. E. G. & COLLINS, D. (1999). The arthropod Alalcomenaeus cambricus Simonetta, from the Middle Cambrian Burgess Shale of British Columbia. Palaeontology 42, 953–977. BRIGGS, D. E. G., FORTEY, R. A. & WILLS, M. A. (1992). Morphological disparity in the Cambrian. Science 256, 1670–1673. BRUTON, D. L. (1981). The arthropod Sidneyia inexpectans, Middle Cambrian, Burgess Shale, British Columbia. Philosophical Transactions of the Royal Society of London, Series B 295, 619–653. BRUTON, D. L. & HAAS, W. (1999). The anatomy and functional morphology of Phacops (Trilobita) from the Hunsrück Slate (Devonian). Sonder-Abdruck aus Palaeontographica Beiträge zur Naturgeschihte der Vorzeit, Abt. A, 253, 29–75. BRUTON, D. L. & WHITTINGTON, H. B. (1983). Emeraldella and Leanchoilia, two arthropods from the Burgess Shale, Middle Cambrian, British Columbia. Philosophical Transactions of the Royal Society of London, Series B 300, 553–582. BUDD, G. (1996). The morphology of Opabinia regalis and the reconstruction of the arthropod stem-group. Lethaia 29, 1–14. BUDD, G. (2001). Tardigrades as ‘ Stem-Group Arthropods ’ : the evidence from the Cambrian fauna. Zoologischer Anzeiger 240, 265–279. BUDD, G. (2002). A palaeontological solution to the arthropod head problem. Nature 417, 271–275. CALMAN, W. T. (1909). Crustacea. In A Treatise on Zoology (ed. R. Lankester). Adam & Charles Black, London. 346 pp. CANNON, H. G. (1947). On the anatomy of the pedunculate barnacle Lithotrya. Philosophical Transactions of the Royal Society of London, Series B 233, 89–136. CANNON, H. G. & MANTON, S. M. (1927). On the feeding mechanism of a mysid crustacean, Hemimysis lamornae. Transactions of the Royal Society of Edinburgh 55, 219–253. CARROLL, S. B., GRENIER, J. K. & WEATHERBEE, S. D. (2001). From DNA to Diversity. Molecular Genetics and the Evolution of Animal Design. Blackwell Science, Oxford. 214pp. CASARES, F. & MANN, R. S. (1998). Control of antennal versus leg development in Drosophila. Nature 392, 723–726. CASARES, F. & MANN, R. S. (2001). The Ground State of the Ventral Appendage in Drosophila. Science 293, 1477–1480. CHAUDONNERET, J. (1950). La morphologie céphalique de Thermobia domestica (Packard) (Insecte Aptérygote Thysanoure). Annales des Sciences Naturelles. Zoologie et Biologie Animale 12, 145–300. CHAUDONNERET, J. (1977). Nécessité de l’analyse fonctionelle des structures en anatomie morphologique : trois exemples concernnant les plans fondamentaux des appendices arthropodiens. Bulletin biologique de la France et de la Belgique (N.S.) 111, 45–59. CHEN, J., VANNIER, J. & HUANG, D. (2001). The origin of crustaceans : new evidence from the Early Cambrian of China. Proceedings of the Royal Society of London, Series B 268, 2181–2187. CHEN, P. & ZHOU, H. (1985). A preliminary analysis on fossil Kazacharthra from Turpan Basin. Kexue Tongbao 3, 950–954. CISNE, J. L. (1974). Evolution of the world fauna of aquatic freeliving arthropods. Evolution 28, 337–366. CISNE, J. L. (1975). Anatomy of Triarthrus and the relationships of the Trilobita. Fossils & Strata 4, 45–63. COOK, C. E., SMITH, M. L., TELFORD, M. J., BASTIANELLO, A. & AKAM, M. (2001). Hox genes and the phylogeny of the arthropods. Current Biology 11, 759–763. CRAMPTON, G. C. (1916). The phylogenetic origin and the nature of the wings of insects according to the paranotal theory. Journal of the New York Entomological Society 24, 1–39. The evolution of arthropod limbs DAHL, E. (1984). The subclass Phyllocarida (Crustacea) and the status of some early fossils : a neontologist’s view. Videnskabelige Meddeleser fra Dansk naturhistorik Forening 145, 61–76. DAHL, E. & HESSLER, R. R. (1982). The crustacean lacinia mobilis : a reconsideration of its origin, function and phylogenetic implications. Zoological Journal of the Linnean Society 74, 133–146. DAHMS, H.-U. (1992). Metamorphosis between naupliar and copepodid phases in the Harpacticoida. Philosophical Transactions of the Royal Society of London, Series B 335, 221–236. DAMEN, W. G. M., HAUSDORF, M., SEYFARTH, E. A. & TAUTZ, D. (1998). A conserved mode of head segmentation in arthropods revealed by the expression patterns of Hox genes in a spider. Proceedings of the National Academy of Sciences 95, 10665–10670. DAVIES, R. G. (1966). The postembryonic development of Hemimerus vicinus Rehn & Rehn (Dermaptera : Hemimeridae). Proceedings of the Royal Entomological Society of London A 41, 67–77. DAWYDOFF, C. (1949). Développement embryonnaire des Arachnides. In Traite de Zoologie (ed. P. P. Grasse), Vol. 6, pp. 320–385. Masson, Paris. DE CALAZANS, D. K. (1992). Taxonomy, distribution and abundance of protozoea, mysis and megalopa stages of Penaeidean decapods from the southern Brazilian coast. Unpublished Ph.D. thesis, University of London. 435pp. DE CELIS, J. F., TYLER, D. M., DE CELIS, J. & BRAY, S. J. (1998). Notch signalling mediates segmentation of the Drosophila leg. Development 125, 4617– 4626. DEUVE, T. (2001). The epipleural field in hexapods. In Origin of the Hexapoda (ed. T. Deuve). Annales de la Société Entomologique de France (N.S.) 37, 195–231. DIEDERICH, R. J., MERRILL, V. K. L., PULTZ, M. A. & KAUFMAN, T. C. (1989). Isolation, structure and expression of labial, a homeotic gene of the Antennapedia complex involved in Drosophila head. Genes and Development 3, 399– 414. DOHLE, W. (1964). Die Embryonalentwicklung von Glomeris marginata (Villers) im Vergleich zur Entwicklung anderer Diplopoden. Zoologische Jahrbücher, Abteilung für Anatomie und Ontogenie der Tiere 81, 241–310. DOHLE, W. (1997). Myriapod-insect relationship as opposed to an insect-crustacean sister group relationship. In Arthropod Phylogeny (eds. R. A. Fortey and R. Thomas), pp. 305–315. Chapman & Hall, London. DUNLOP, J. A. (1998). The origins of tetrapulmonate book lungs and their significance for chelicerate phylogeny. Proceedings of the 17th European Colloqium of Arachnology, Edinburgh 1997, pp. 9–16. British Arachnological Society, Burnham Beeches, Bucks. DUNLOP, J. A. & BRADDY, S. J. (2001). Scorpions and their sistergroup relationships. In Scorpions 2001. In Memoriam Gary A Polis (eds. V. Fet and P. A. Selden), pp. 1–24. British Arachnological Society, Burnham Beeches, Bucks. DUNLOP, J. A. & SELDEN, P. A. (1997). The early history and phylogeny of the chelicerates. In Arthropod Phylogeny (eds. R. A. Fortey and R. Thomas), pp. 221–235. Chapman & Hall, London. EDGECOMBE, G. D. & RAMSKÖLD, L. (1999). Relationships of Cambrian Arachnata and the systematic position of Trilobita. Journal of Paleontology 73, 263–287. EDGECOMBE, G. D., WILSON, G. D. F., COLGAN, D. J., GRAY, M. R. & CASSIS, G. (2000). Arthropod Cladistics : combined analysis of histone H3 and U2 snRNA sequences and morphology. Cladistics 16, 155–203. ELDREDGE, N. (1974). Revision of the suborder Synziphosurina (Chelicerata, Merostomata), with remarks on merostome phylogeny. American Museum Novitiates 2543, 1– 41. 295 EMERSON, M. J. & SCHRAM, F. R. (1990). The origin of crustacean biramous appendages and the evolution of the Arthropoda. Science 250, 667–669. EMERSON, M. J. & SCHRAM, F. R. (1991). Remipedia Part 2 Paleontology. Proceedings of the San Diego Society of Natural History 7, 1–52. EMERSON, M. J. & SCHRAM, F. R. (1997). Theories, patterns, and reality : game plan for arthropod phylogeny. In Arthropod Phylogeny (eds. R. A. Fortey and R. Thomas), pp. 67–86. Chapman & Hall, London. ENGHOFF, H., DOHLE, W. & BLOWER, J. G. (1994). Anamorphosis in millipedes (Diplopoda) – the present state of knowledge with some developmental and phylogenetic considerations. Zoological Journal of the Linnean Society 109 (1993), 103–234. ERIKSSON, B. J., TAIT, N. N. & BUDD, G. G. (2003). Head development in the onychophoran Euperipatoides kanangrensis with particular reference to the central nervous system. Journal of Morphology 255, 1–23. FECHTER, H. (1961). Anatomie und Funktion der Kopfmuskulatur von Cylindroiulus teutonicus (Pocock). Zoologische Jahrbücher, Abteilung für Anatomie und Ontogenie der Tiere 32, 233–249. FORTEY, R. A. & MORRIS, S. F. (1978). Discovery of nauplius-like trilobite larvae. Palaeontology 21, 823–833. FRANÇOIS, J. (1969). Anatomie et morphologie céphalique des Protures (Insecta Apterygota). Mémoires du Muséum National D’Histoire Naturelle, Ser. A, Zoologie LIX, 1–144. FRIEDRICH, M. & TAUTZ, D. (1995). Ribosomal DNA phylogeny of the major extant arthropod classes and the evolution of myriapods. Nature 376, 165–167. FRYER, G. (1983). Functional ontogenetic changes in Branchinecta ferox (Milne-Edwards) (Crustacea : Anostraca). Philosophical Transactions of the Royal Society of London, Series B 303, 229–343. FRYER, G. (1988). Studies on the functional morphology and biology of the Notostraca (Crustacea : Branchiopoda). Philosophical Transactions of the Royal Society of London, Series B 321, 27–124. FRYER, G. (1992). The origin of the Crustacea. Acta Zoologica 73, 273–286. FRYER, G. (1997). A defence of arthropod polyphyly. In Arthropod Phylogeny (eds. R. A. Fortey and R. Thomas), pp. 23–33. Chapman & Hall, London. GARCÍA-BELLIDO, D. C. & COLLINS, D. (2001). A new study of Marrella splendens from the Burgess Shale, British Columbia, Canada. Abstract, Third International Conference on Trilobites and their relatives, p. 13. GIRIBET, G., EDGECOMBE, G. D. & WHEELER, W. C. (2001). Arthropod phylogeny based on eight molecular loci and morphology. Nature 413, 157–161. GONZÁLEZ-CRESPO, S. & MORATA, G. (1996). Genetic evidence for the subdivision of the arthropod limb into coxopodite and telopodite. Development 122, 3921–3928. GOTO, S. & HAYASHI, S. (1997). Specification of the embryonic limb primordium by graded activity of Decapentaplegic. Development 124, 125–132. GOULD, S. J. (1989). Wonderful Life. The Burgess Shale and the Nature of History. Norton, New York. 347pp. GRANDJEAN, F. (1952). Le morcellement secondaire des tarses de Tarsolarkus sp. (Acarien). Archives de Zoologie Expérimentale et Générale 89 (3), 113–123. GREENSLADE, P. & WHALLEY, P. E. S. (1986). The systematic position of Rhyniella praecursor Hirst & Maulik (Collembola). The earliest known hexapod. In Second International Seminar on Apterygota (ed. R. Dallai), pp. 319–323. University of Sienna, Sienna. 296 GRENIER, J., GARBER, T., WARREN, R., WHITTINGTON, P. & CARROLL, S. (1997). Evolution of the entire arthropod Hox gene set predated the origin and radiation of the onychophoran/ arthropod clade. Current Biology 7, 547–553. GRUBER, J. (1978). Redescription of Ceratolasma tricantha Goodnight & Goodnight, with notes on the family Ischyropsalidae (Opiliones, Palpatores). Journal of Arachnology 6, 105–124. GURNEY, R. (1942). Larvae of Decapod Crustacea. The Ray Society, London. 306pp. HANSEN, H. J. (1925). On the comparative morphology of the appendages in the Arthropoda. A. Crustacea. Gyldendalske, Copenhagen. 176pp. HEATH, H. (1924). The external development of certain phyllopods. Journal of Morphology 38, 453–483. HEEGAARD, P. (1945). Remarks on the phylogeny of the Arthropods. Arkiv för Zoologi 37A (3), 1–15. HESSLER, R. R. (1964). The Cephalocarida. Comparative Skeletomusculature. Memoirs of the Connecticut Academy of Arts & Sciences 16, 1–97. HESSLER, R. R. (1982). The structural morphology of walking mechanisms in eumalacostracan crustaceans. Philosophical Transactions of the Royal Society of London, Series B 296, 245–298. HESSLER, R. R. (1983). A defense of the caridoid facies ; wherein the early evolution of the Eumalacostraca is discussed. In Crustacean phylogeny (ed. F. R. Schram), pp. 145–164. A. A. Balkema, Rotterdam. HESSLER, R. R. & NEWMAN, W. A. (1975). A trilobitomorph origin for the Crustacea. Fossils & Strata 4, 437– 459. HEYMONS, R. (1901). Die Entwicklungsgeschichte der Scolopender. Zoologica 33, 1–33, pls. I–VIII. HØEG, J. T., WHYTE, M. A., GLENNER, H. & SCHRAM, F. R. (1999). New evidence on the basic phylogeny of the Cirripedia Thoracica. In Crustaceans and the biodiversity crisis (eds. F. R. Schram and J. C. von Vaupel Klein), pp. 101–114. Proceedings of the 4th International Crustacean Congress, Amsterdam, The Netherlands, July. E. J. Brill, Leiden. HOLM, G. (1898) Über die Organisation des Eurypterus fischeri Eichw. Memoires de l’Academie Impériale des Sciences de St Pétersbourg, 8th Series, 8 (2), 1–57, 10 pls. HORNE, D. J., COHEN, A. & MARTENS, K. (2002). Taxonomy, Morphology and Biology of Quaternary and Living Ostracoda. In The Ostracoda : Applications in Quaternary Research, pp. 5–36. Geophysical Monograph 131. American Geophysical Union. HOU, X. & BERGSTRÖM, J. (1997). Arthropods of the Lower Cambrian Chengjiang fauna, southwest China. Fossils & Strata 45, 1–116. HOU, X., SIVETER, D. J., WILLIAMS, M., WALOSSEK, D. & BERGSTRÖM, J. (1996). Preserved appendages in the arthropod Kunmingella from the early Cambrian of China : its bearing on the systematic position of the Bradoriida and the fossil record of the Ostracoda. Philosophical Transactions of the Royal Society of London, Series B 351, 1131–1145. HUGHES, C. P. (1975). Redescription of Burgessia bella from the Middle Cambrian Burgess Shale, British Columbia. Fossils & Strata 4, 415– 435. HUXLEY, T. H. (1877). A Manual of the Anatomy of Invertebrated Animals. London. 698pp. HUXLEY, T. H. (1880). The Crayfish. An Introduction to the Study of Zoology. Keegan Paul & Co., London. 371pp. HUYS, R. & BOXSHALL, G. A. (1991). Copepod Evolution. The Ray Society, London. 468pp. Geoff A. Boxshall HWANG, U. W., FRIEDRICH, M., TAUTZ, D., PARK, C. J. & KIM, W. (2001). Mitochondrial protein phylogeny joins myriapods with chelicerates. Nature 413, 154 –157. IMMS, A. D. (1939). On the antennal musculature in Insects and other arthropods. Quarterly Journal of the Microscopical Society 81, 273–320. ITÔ , T. (1989). Origin of the basis in copepod limbs, with reference to remipedian and cephalocaridan limbs. Journal of Crustacean Biology 9, 85–103. JUDSON, M. (1994). Studies on the morphology and systematics of the Teneriffiidae (Acari, Prostigmata) I. A new species of Neoteneriffiola from Namibia. Acarologia 35, 115–134. KARAYTUG, S. & BOXSHALL, G. A. (1996). The life cycle of Paracyclops fimbriatus (Fischer, 1853) (Copepoda, Cyclopoida). Bulletin of the Natural History Museum London (Zoology) 62, 41–70. KLASS, K.-D. & KRISTENSEN, N. P. (2001). The ground plan and affinities of hexapods : recent progress and open problems. In Origin of the Hexapoda (ed. T. Deuve). Annales de la Société Entomologique de France 37, 265–298. KOJIMA, T., SATO, M. & SAIGO, K. (2000). Formation and specification of distal leg segments in Drosophila by dual Bar homeobox genes, BarH1 and BarH2. Development 127, 769–778. KRAUS, O. (1997). Phylogenetic relationships between higher taxa of tracheate arthropods. In Arthropod Phylogeny (eds. R. A. Fortey and R. Thomas), pp. 295–303. Chapman & Hall, London. KRAUS, O. (2001). ‘Myriapoda ’ and the ancestry of the Hexapoda. In Origin of the Hexapoda (ed. T. Deuve). Annales de la Société Entomologique de France 37, 105–127. KRAUS, O. & KRAUS, M. (1994). Phylogenetic system of the Tracheata (Mandibulata) : on ‘ Myriapoda ’. Verhandlungen des Naturwissenschaftlichen Vereins Hamburg (N.F.) 34, 5–31. KRISTENSEN, N. P. (1997). The groundplan and basal diversification of hexapods. In Arthropod Phylogeny (eds. R. A. Fortey and R. Thomas), pp. 281–293. Chapman & Hall, London. KUKALOVÁ-PECK, J. (1983). Origin of the insect wing and wing articulation from the arthropodan leg. Canadian Journal of Zoology 61, 1618–1669. KUKALOVÁ-PECK, J. (1987). New Carboniferous Diplura, Monura and Thysanura, the hexapod groundplan, and the role of thoracic lobe in the origin of wings (Insecta). Canadian Journal of Zoology 70, 236–255. KUKALOVÁ-PECK, J. (1992). Fossil history and the evolution of hexapod structures. In The Insects of Australia (ed. I. D. Nauman), pp. 141–179. Melbourne University Press, Melbourne. KUKALOVÁ-PECK, J. (1997). Arthropod phylogeny and ‘ basal ’ morphological structures. In Arthropod Phylogeny (eds. R. A. Fortey and R. Thomas), pp. 249–268. Chapman & Hall, London. LANKESTER, E. R. (1881). Limulus an arachnid. Quarterly Journal of Microscopical Science 21, 504 –548, 609–649. LAURIE, M. (1890). The embryology of a scorpion (Euscorpius italicus). Quarterly Journal of Microscopical Science 32, 587–597. LAUTERBACH, K. E. (1972). Über die sogenannte GanzbeinMandibel der Tracheaten, in besondere der Myriapoda. Zoologischer Anzeiger 188, 145–154. LECUIT, T. & COHEN, S. M. (1997). Proximal-distal axis formation in the Drosophila leg. Nature 388, 139–145. LEHMANN, W. M. (1950). Mimetaster hexagonalis Gürich. Neues Jahrbuch für Mineralogie, Geologie und Paläontologie 91B, 101–120. LOCKETT, G. H. & MILLIDGE, A. F. (1951). British Spiders, Vol. 1. The Ray Society, London. 310pp. MAAS, A. & WALOSZEK, D. (2001). Cambrian Derivatives of the Early Arthropod Stem Lineage, Pentastomids, Tardigrades and The evolution of arthropod limbs Lobopodians – An ‘Orsten ’ Perspective. Zoologischer Anzeiger 240, 451– 459. MACHIDA, T. (2000). Serial Homology of the Mandible and Maxilla in the Jumping Bristletail Pedetontus unimaculatus Machida, Based on External Embryology (Hexapoda : Archaeognatha, Machilidae). Journal of Morphology 245, 19–28. MADDOX, R. F. (2000). The antennule in podocopid Ostracoda : chaetotaxy, ontogeny and morphometrics. Micropaleontology 4 (Suppl. 2), 1–72. MANTON, S. M. (1928). On the embryology of a mysid crustacean Hemimysis lamnornae. Philosophical Transactions of the Royal Society of London, Series B 216, 363– 463. MANTON, S. M. (1934). On the embryology of the crustacean Nebalia bipes. Philosophical Transactions of the Royal Society of London, Series B 223, 163–238. MANTON, S. M. (1954). The evolution of arthropodan locomotory mechanisms. Part 4. The structure, habits and evolution of the Diplopoda. Journal of the Linnean Society 42, 299–368. MANTON, S. M. (1958). The evolution of arthropodan locomotory mechanisms. Part 6. Habits and evolution of the Lysiopetaloidea (Dilopoda), some principles of leg design in Diplopoda and Chilopoda, and limb structure in Diplopoda. Journal of the Linnean Society 43, 487–556. MANTON, S. M. (1964). Mandibular mechanisms and the evolution of arthropods. Philosophical Transactions of the Royal Society of London, Series B 247, 1–183. MANTON, S. M. (1965). The evolution of arthropodan locomotory mechanisms. Part 8. Functional requirements and body design in Chilopoda, together with a comparative account of their skeleto-muscular systems and an Appendix on a comparison between burrowing forces of annelids and chilopods and its bearing upon the evolution of the arthropodan haemocoel. Journal of the Linnean Society 46, 251–378. MANTON, S. M. (1966). The evolution of arthropodan locomotory mechanisms. Part 9. Functional requirements and body design in Symphyla and Pauropoda and the relationships between Myriapoda and pterygote Insects. Journal of the Linnean Society 46, 103–141. MANTON, S. M. (1973). Arthropod phylogeny – a modern synthesis. Journal of Zoology, London 171, 111–130. MANTON, S. M. (1977). The Arthropoda : habits, functional morphology, and evolution. Clarendon Press, London. 527pp. MARTIN, J. M. (1992). Branchiopoda. In Microscopic Anatomy of Invertebrates, Vol. IX. Crustacea (ed. F. W. Harrison), pp. 25–224. J. Wiley & Sons Inc., New York. MATSUDA, R. (1970). Morphology and evolution of the insect thorax. Memoirs of the Entomological Society of Canada 76, 1– 431. MATSUDA, R. (1976). Morphology and Evolution of the Insect Abdomen. Pergamon Press, Oxford. 501pp. MCKENZIE, K. G., CHEN P.-J. & MAJORAN, S. (1991). Almatium gusevi (Chernyshev 1940) : redescription, shield-shapes, and speculations on the reproductive mode (Branchiopoda, Kazacharthra). Paläontologische Zeitschrift 65, 305–317. MCLAUGHLIN, P. A. (1982). Comparative Morphology of Crustacean Appendages. In The Biology of Crustacea, Vol. 2. Embryology, Morphology and Genetics (ed. L. G. Abele), pp. 197–256. Academic Press, New York, London. MEAD, K. S. & KOEHL, M. A. R. (2000). Stomatopod antennule design : the asymmetry, sampling efficiency and ontogeny of olfactory flicking. Journal of Experimental Biology 203, 3795–3808. MILNE EDWARDS, H. (1851). Observations sur le squelette tegumentaire des Crustacés Décapods et sur la morphologie de ces 297 animaux. Annales des Sciences Naturelles, Zoologie, ser. 3, XVI, 221–291, pls. 8–11. MINELLI, A., FODDAI, D., PEREIRA, L. A. & LEWIS, J. G. E. (2000). The evolution of segmentation of centipede trunk and appendages. Journal of Zoological Systematics and Evolutionary Research 38, 103–117. MINELLI, A. & FUSCO, G. (1995). Body segmentation and segment differentiation : the scope for heterochronic change. In Evolutionary Change and Heterochrony (ed. K. J. McNamara), pp. 49–63. John Wiley & Sons, Chichester. MITTMANN, B. & SCHOLTZ, G. (2001). Distal-less expression in embryos of Limulus polyphemus (Chelicerata, Xiphosura) and Lepisma saccharina (Insecta, Zygentoma) suggests a role in development of mechanoreceptors, chemoreceptors, and the CNS. Development, Genes and Evolution 211, 232–243. MÜLLER, K. J. (1982). Hesslandona unisulcata sp. nov. with phosphatised appendages from Upper Cambrian ‘Orsten ’ of Sweden. In Fossil and Recent Ostracods (eds. R. H. Bate, E. Robinson and L. M. Sheppard), pp. 276–304. Ellis Horwood Ltd, Chichester. MÜLLER, K. J. & WALOSSEK, D. (1985). Skaracarida, a new order of Crustacea from the Upper Cambrian of Västergötland, Sweden. Fossils & Strata 17, 1–65. MÜLLER, K. J. & WALOSSEK, D. (1986 a). Arthropod larvae from the Upper Cambrian of Sweden. Transactions of the Royal Society of Edinburgh : Earth Sciences 77, 157–179. MÜLLER, K. J. & WALOSSEK, D. (1986 b). Martinssonia elongata gen. et sp. n., a crustacean-like euarthropod from the Upper Cambrian ‘ Orsten’ of Sweden. Zoologica Scripta 15, 73–92. MÜLLER, K. J. & WALOSSEK, D. (1987). Morphology, ontogeny and life habit of Agnostus pisiformis from the Upper Cambrian of Sweden. Fossils & Strata 19, 1–124. MÜLLER, K. J. & WALOSSEK, D. (1988). External morphology and larval development of the Upper Cambrian maxillopod Bredocaris admirabilis. Fossils & Strata 32, 1–202. NAYROLLES, P. (1991). La chetotaxie antennaire des Collemboles Symphypleones. Travaux du Laboratoire d’Ecobiologie des Arthropodes Edaphiques, Toulouse 6 (3), 1–94. OLESEN, J. (1999). Larval and postlarval development of the branchiopod clam shrimp Cyclestheria hislopi (Baird, 1859) (Crustacea, Branchiopoda, Conchostraca, Spinicaudata). Acta Zoologica 80, 163–184. OLESEN, J., RICHTER, S. & SCHOLTZ, G. (2001). The evolutionary transformation of phyllopodous to stenopodous limbs in the Branchiopod (Crustacea) – Is there a common mechanism for early limb development in arthropods ? International Journal of Developmental Biology 45, 869–876. OLESEN, J. & WALOSSEK, D. (2000). Limb ontogeny and trunk segmentation in Nebalia species (Crustacea, Malacostraca, Leptostraca). Zoomorphology 120, 47–64. ORR, P. J. & BRIGGS, D. E. G. (1999). Exceptionally preserved conchostracans and other crustaceans from the Upper Carboniferous of Ireland. Special Papers Palaeontology 62, 1–68. ORR, P. J., SIVETER, D. J., BRIGGS, D. E. G., SIVETER, D. J. & SUTTON, M. D. (2000). A new arthropod from the Silurian Konservat-Lagerstätte of Herefordshire, UK. Proceedings of the Royal Society of London, Series B 267, 1497–1504. PALMGREN, P. (1978). On the muscular anatomy of spiders. Acta Zoologica, Fennica 155, 1– 41. PANGANIBAN, G., NAGY, L. & CARROLL, S. M. (1994). The role of the Distal-less gene in the development and evolution of insect limbs. Current Biology 4, 671–675. 298 PATTEN, W. (1896). Variations in the development of Limulus polyphemus. Journal of Morphology 12, 17–146. PATTEN, W. & REDENBAUGH, W. A. (1899). Studies on Limulus. Journal of Morphology 16, 1–200. PÉREZ FARFANTE, I. & KENSLEY, B. (1997). Penaeoid and Sergestoid Shrimps and Prawns of the World. Keys and Diagnoses for the Families and Genera. Mémoires de Muséum National D’Histoire Naturelle 175, 233pp. PETERS, W. L. & CAMPBELL, I. C. (1992). Ephemeroptera (Mayflies). In The Insects of Australia (ed. I. D. Nauman), pp. 279–293. Melbourne University Press, Melbourne. PETERSON, M., ROGERS, B. T., POPADIĆ, A. & KAUFMAN, T. C. (1999). The embryonic expression pattern of labial, posterior homeotic complex genes and the teashirt homologue in an apterygote insect. Development, Genes and Evolution 209, 77–90. POCOCK, R. I. (1893). On the classification of the tracheate Arthropoda. Zoologischer Anzeiger 16, 1–38. POPADIĆ, A., RUSCH, D., PETERSON, M., ROGERS, B. T. & KAUFMAN, T. C. (1996). Origin of the arthropod mandible. Nature 380, 395. PRPIC, N.-M., WIGAND, B., DAMEN, W. G. M. & KLINGER, M. (2001). Expression of dachsund in wild-type and Distal-less mutant Tribolium corroborates serial homologies in insect appendages. Development, Genes and Evolution 211, 467– 477. RAASCH, G. O. (1939). Cambrian Merostomata. Special Papers. Geological Society of America 19, 1–146. RAMSKÖLD, L. & EDGECOMBE, G. D. (1996). Trilobite appendage structure – Eoredlichia reconsidered. Alcheringa 20, 269–276. RAMSKÖLD, L., CHEN, J., EDGECOMBE, G. D. & ZHOU, G. (1997). Cindarella and the arachnate clade Xandarellida (Arthropoda, Early Cambrian) from China. Transactions of the Royal Society of Edinburgh, Earth Sciences 88, 19–38. REGIER, J. C. & SHULTZ, J. W. (1997). Molecular phylogeny of the major arthropod groups indicates polyphyly of crustaceans and a new hypothesis for the origin of hexapods. Molecular Biology and Evolution 14, 902–913. REGIER, J. C. & SHULTZ, J. W. (2001). A phylogenetic analysis of Myriapoda (Arthropoda) using two nuclear protein-encoding genes. Zoological Journal of the Linnean Society 132, 469– 486. RICHTER, S., EDGECOMBE, G. D. & WILSON, G. D. F. (2002). The lacinia mobilis and similar structures – a valuable character in arthropod phylogenetics ? Zoologischer Anzeiger 241, 339–361. RICHTER, S. & SCHOLTZ, G. (2001). Phylogenetic analysis of the Malacostraca (Crustacea). Journal of Zoological Systematics and Evolutionary Research 39, 113–136. RILLING, G. (1960). Zur Anatomie des braunen Steinläufers Lithobius forficatus L. (Chilopoda). Skelettmuskelsystem, peripheres Nervensystem und Sinnesorgane des Rumpfes. Zoologische Jahrbücher, Abteilung für Anatomie und Ontogenie der Tiere 78, 39–128. SANDERS, H. L. (1957). The Cephalocarida and crustacean phylogeny. Systematic Zoology 6, 112–129. SANDERS, H. L. (1963). The Cephalocarida. Functional morphology, larval development, comparative external anatomy. Memoirs of the Connecticut Academy of Arts & Sciences 15, 1–80. SARS, G. O. (1922). Ostracoda, Parts I & II, Cypridinidae. Conchoeciidae, Polycopidae (part). An Account of the Crustacea of Norway IX, 1–32, pls. I–XVI. Bergen Museum, Bergen. SAVORY, T. (1977). Arachnida, 2nd Edn, Academic Press, London, New York and San Francisco. 340pp. SCHMIDT-RHAESA, A., BARTOLOMAEUS, T., LEMBURG, C., EHLERS, U. & GAREY, J. R. (1998). The position of the Arthropoda in the phylogenetic system. Journal of Morphology 238, 263–285. Geoff A. Boxshall SCHOLTZ, G. (2001). Evolution of developmental patterns in arthropods – the analysis of gene expression and its bearing on morphology and phylogenetics. Zoology, Analysis of Complex Systems 103, 99–111. SCHRAM, F. R. (1978). Arthropods : a convergent phenomenon. Fieldiana : Geology 29, 61–108. SCHRAM, F. R. (1984). Fossil Syncarida. Transactions of the San Diego Society of Natural History 20, 189–246. SCHRAM, F. R. (1986). Crustacea. Oxford University Press, New York and Oxford. 606pp. SCHRAM, F. R. & HOF, C. H. J. (1998). Fossils and the Interrelationships of Major Crustacean Groups. In Arthropod Fossils and Phylogeny (ed. G. E. Edgecombe), pp. 234 –302. Columbia University Press, New York. SCHRAM, F. R., YAGER, J. & EMERSON, M. J. (1986). Remipedia. Part 1. Systematics. Memoirs of the San Diego Society of Natural History 15, 1–60. SCHUTZE, M. L. M., DA ROCHA, C. E. F. & BOXSHALL, G. A. (2000). Antennular development during the copepodid phase in the family Cyclopidae (Copepoda, Cyclopoida). Zoosystema 22, 749–806. SCOURFIELD, D. J. (1926). On a new type of crustacean from the Old Red Sandstone (Rhynie Chert Bed, Aberdeenshire) – Lepidocaris rhyniensis, gen, et sp. nov. Philosophical Transactions of the Royal Society of London, Series B 214, 153–186. SCOURFIELD, D. J. (1940). Two new complete specimens of young stages of the Devonian fossil crustacean Lepidocaris rhyniensis. Proceedings of the Linnean Society 152, 290–298. SELDEN, P. A. (1981). Functional morphology of the prosoma of Baltoeurypterus tetragonophthalmus (Fischer) (Chelicerata : Eurypterida). Transactions of the Royal Society of Edinburgh, Earth Sciences 72, 9– 48. SELDEN, P. A. & DUNLOP, J. A. (1998). Fossil taxa and Relationships of Chelicerates. In Arthropod Fossils and Phylogeny (ed. G. D. Edgecombe), pp. 303–331. Columbia University Press, New York. SHAKOORI, A. R. (1968). Morphology and skeletomusculature of Caenestheria propinqua (Sars) (Conchostraca ; Branchiopoda ; Crustacea). Bulletin of the Department of Zoology University of the Panjab (N.S.) 2, 1– 48. SHEAR, W. A. & BONAMO, P. M. (1988). Devonobiomorpha, a new order of centipedes (Chilopoda) from the Middle Devonian of Gilboa, New York State, and the phylogeny of centipede orders. American Museum Novitiates 2927, 1–30. SHU, D., VANNIER, J., LUO, H., CHEN, L., ZHANG, X. & HU, S. (1999). Anatomy and lifestyle of Kunmingella (Arthropoda, Bradoriida) from the Chengjiang fossil Lagerstätte (lower Cambrian ; Southwest China). Lethaia 32, 279–298. SHULTZ, J. W. (1989). Morphology of locomotor appendages in Arachnida : evolutionary trends and phylogenetic implications. Zoological Journal of the Linnean Society 97, 1–56. SHULTZ, J. W. (1990). Evolutionary morphology and phylogeny of Arachnida. Cladistics 6, 1–38. SHULTZ, J. W. & REGIER, J. C. (2000). Phylogenetic analysis of arthropods using two nuclear protein-encoding genes supports a crustacean+hexapod clade. Proceedings of the Royal Society, London, Series B 267, 1011–1019. SILVESTRI, F. (1902). Pauropoda. In Acari, Myriapoda et Scorpiones hucusque in Italia reperta, Portici. 78 pp. SIMONETTA, A. M. (1970). Studies on the non-Trilobite Arthropoda of the Burgess Shale (Middle Cambrian). Palaeontologica Italica 66, 35– 45. The evolution of arthropod limbs SIMONETTA, A. M. & DELLA CAVE, L. (1991). Early Palaeozoic arthropods and problems of arthropod phylogeny ; with some notes on taxa of doubtful affinity. In The Early Evolution of Metazoa and the Significance of Problematic Taxa (eds. A. M. Simonetta and S. Conway Morris), pp. 189–244. Cambridge University Press, Cambridge. SIVETER, D. J., WILLIAMS, M. & WALOSZEK, D. (2001). A Phosphatocopid Crustacean with Appendages from the Lower Cambrian. Science 293, 479–481. SNODGRASS, R. E. (1935). Principles of Insect Morphology. New York, McGraw-Hill. 667pp. SNODGRASS, R. E. (1938). Evolution of the Annelida, Onychophora and Arthropoda. Smithsonian Miscellaneous Collections 97, 1–159. STANICZEK, A. H. (2000). The Mandible of Silverfish (Insecta : Zygentoma) and Mayflies (Ephemeroptera) : Its Morphology and Phylogenetic Significance. Zoologischer Anzeiger 239, 147–178. STØRMER, L. (1944). On the relationships and phylogeny of fossil and Recent Arachnomorpha. Skrifter utgitt av det Norske Videnskkaps-Akademi i Oslo, Mat.-naturv. Klasse 5, 1–158. STRUHL, G. (1981). A homoeotic mutation transforming leg to antenna in Drosophila. Nature 292, 635–638. STÜRMER, W. E. & BERGSTRÖM, J. (1976). The arthropods Mimetaster and Vachonisia from the Devonian Hunsrück Shale. Paläontologische Zeitschrift 50, 78–111. STÜRMER, W. E. & BERGSTRÖM, J. (1981). Weinbergina, a xiphosuran arthropod from the Devonian Hunsrück Slate. Paläontologische Zeitschrift 55, 237–255. SUTTON, M. D., BRIGGS, D. E. G., SIVETER, D. J., SIVETER, D. J. & ORR, P. J. (2002). The arthropod Offacolus kingi (Chelicerata) from the Silurian of Herefordshire, England : computer based morphological reconstructions and phylogenetic affinities. Proceedings of the Royal Society of London, Series B 269, 1195–1203. TAMARELLE, M. (1984). Transient rudiments of second antennae on the ‘intercalary ’ segment of embryos of Anurida maritima Guer. (Collembola : Arthropleona) and Hyphantria cunea Drury (Lepidoptera : Arctiidae). International Journal of Insect Morphology and Embryology 13, 331–336. TATTERSALL, W. M. & TATTERSALL, O. S. (1951). The British Mysidacea. The Ray Society, London. 460pp. TELFORD, M. J. & THOMAS, R. H. (1998). Expression of homeobox genes shows chelicerate arthropods retain their deutocerebral segment. Proceedings of the National Academy of Sciences 95, 10671–10675. TIEGS, O. W. (1940). The embryology and affinities of the Symphyla, based on a study of Hanseniella agilis. The Quarterly Journal of Microscopical Science 82, 1–225, pls. 1–9. TIEGS, O. W. (1947). The development and affinities of the Pauropoda, based on a study of Pauropus silvaticus. The Quarterly Journal of Microscopical Science 88, 165–267. TIEGS, O. W. & MANTON, S. M. (1958) The evolution of the Arthropoda. Biological Reviews 33, 255–337. TSUJI, T., SATO, A., HIRATANI, I., TAIRA, M., SAIGO, K. & KOJIMA, T. (2000). Requirements of Lim1, a Drosophila LIM-homeobox gene, for normal leg and antennal development. Development 127, 4315–4323. VAN DER HAMMEN, L. (1989). An introduction to comparative arachnology. SPB Academic Publishing, The Hague. 576pp. VANNIER, J. & CHEN, J.-Y. (2000). The Early Cambrian colonization of pelagic niches exemplified by Isoxys (Arthropoda). Lethaia 33, 295–311. 299 WÄGELE, J. W., ERIKSON, T., LOCKHART, P. & MISOF, B. (1999). The Ecdysozoa : Artifact or monophylum ? Journal of Zoological Systematics and Evolutionary Research 37, 211–223. WALOSSEK, D. (1993). The Upper Cambrian Rehbachiella and the phylogeny of Branchiopoda and Crustacea. Fossils & Strata 32, 1–202. WALOSSEK, D. (1999). On the Cambrian diversity of Crustacea. In Crustaceans and the Biodiversity Crisis (eds. F. R. Schram and J. C. von Vaupel Klein), pp. 3–27. Brill, Leiden. WALOSSEK, D. & MÜLLER, K. J. (1990). Upper Cambrian stemlineage crustaceans and their bearing upon the monophyletic origin of the Crustacea and the position of Agnostus. Lethaia 23, 409– 427. WALOSSEK, D. & MÜLLER, K. J. (1997). Cambrian ‘ Orsten’-type arthropods and the phylogeny of Crustacea. In Arthropod Phylogeny (eds. R. A. Fortey and R. Thomas), pp. 139–153. Chapman & Hall, London. WALOSSEK, D. & MÜLLER, K. J. (1998). Early Arthropod Phylogeny in Light of the Cambrian ‘Orsten ’ Fossils. In Arthropod Fossils and Phylogeny (ed. G. E. Edgecombe), pp. 185–231. Columbia University Press, New York. WALOSSEK, D. & SZANIAWSKI, H. (1991). Cambrocaris baltica n. gen. n. sp., a possible stem-lineage crustacean from the Upper Cambrian of Poland. Lethaia 24, 363–378. WALOSZEK, D. & DUNLOP, J. A. (2002). A larval sea spider (Arthropoda : Pycnogonida) from the Upper Cambrian ‘orsten ’ of Sweden, and the phylogenetic position of pycnogonids. Palaeontology 45, 421–446. WATERMAN, T. H. & TRAVIS, D. F. (1953). Respiratory reflexes and the flabellum of Limulus. Journal of Cellular and Comparative Physiology 41, 261–290. WATERSTON, C. D. (1975). Gill structures in the Lower Devonian eurypterid Tarsopterella scotica. Fossils & Strata 4, 241–254. WEYGOLDT, P. (1996). Evolutionary morphology of whip spiders : towards a phylogenetic system (Chelicerata : Arachnida : Amblypygi). Journal of Zoological Systematics and Evolutionary Research 34, 185–202. WHEELER, W. M. (1893). A contribution to insect embryology. Journal of Morphology 8, 1–160. WHITTINGTON, H. B. (1971). Redescription of Marrella splendens (Trilobitoidea) from the Burgess Shale, Middle Cambrian, British Columbia. Bulletin of the Geological Survey of Canada 209, 1–24. WHITTINGTON, H. B. (1974). Yohoia Walcott and Plenocaris n. gen., arthropods from the Burgess Shale, Middle Cambrian, British Columbia. Bulletin of the Geological Survey of Canada 231, 1–27, pls. I–XVIII. WHITTINGTON, H. B. (1977). The Middle Cambrian trilobite Naraoia, Burgess Shale, British Columbia. Philosophical Transactions of the Royal Society of London, Series B 280, 409– 443. WHITTINGTON, H. B. (1980). Exoskeleton, moult stage, appendage morphology, and habits of the Middle Cambrian trilobite Olenoides serratus. Palaeontology 23, 171–204. WHITTINGTON, H. B. (1981). Rare arthropods from the Burgess Shale, Middle Cambrian, British Columbia. Philosophical Transactions of the Royal Society of London, Series B 292, 329–357. WHITTINGTON, H. B. & ALMOND, J. E. (1987). Appendages and habits of the Upper Ordovician trilobite Triarthrus eatoni. Philosophical Transactions of the Royal Society of London, Series B 317, 1– 46. WIGGLESWORTH, V. B. (1973). Evolution of insect wings and flight. Nature 246, 127–129. 300 WILLMANN, R. (1997). Advances and problems in insect phylogeny. In Arthropod Phylogeny (eds. R. A. Fortey and R. Thomas), pp. 269–279. Chapman & Hall, London. WILLS, M. A. (1996). Classification of the Arthropod Fuxianhuia. Science 272, 746–747. WILLS, M. A. (1997). A phylogeny of recent and fossil Crustacea derived from morphological characters. In Arthropod Phylogeny (eds. R. A. Fortey and R. Thomas), pp. 155–167. Chapman & Hall, London. WILLS, M. A., BRIGGS, D. E. G., FORTEY, R. A. & WILKINSON, M. (1995). The significance of fossils in understanding arthropod Geoff A. Boxshall evolution. Verhandlungen Deutsches zoologisches Gesellschaft 88, 203–215. WILLS, M. A., BRIGGS, D. E. G., FORTEY, R. A., WILKINSON, M. & SNEATH, P. H. A. (1998). An arthropod phylogeny based on fossil and Recent taxa. In Arthropod Fossils and Phylogeny (ed. G. E. Edgecombe), pp. 33–105. Columbia University Press, New York. YAGER, J. (1981). Remipedia, a new class of Crustacea from a marine cave in the Bahamas. Journal of Crustacean Biology 1, 328–333. YAGER, J. (1987). Speleonectes tulumensis n.p. (Crustacea : Remipedia) from two anchialine cenotes of the Yucatan Peninsula, Mexico. Stygologia 3, 160–166.