A phylogenetic analysis of the arachnid orders based on morphological characters

Morphological evidence for resolving relationships among arachnid orders was surveyed and assembled in a matrix comprising 59 euchelicerate genera (41 extant, 18 fossil) and 202 binary and unordered multistate characters. Parsimony analysis of extant genera recovered a monophyletic Arachnida with the topology (Palpigradi (Acaromorpha (Tetrapulmonata (Haplocnemata, Stomothecata nom. nov. )))), with Acaromorpha containing Ricinulei and Acari, Tetrapulmonata containing Araneae and Pedipalpi (Amblypygi, Uropygi), Haplocnemata (Pseudoscorpiones, Solifugae) and Stomothecata (Scorpiones, Opiliones). However, nodal support and results from exploratory implied weights analysis indicated that relationships among the ﬁve clades were effectively unresolved. Analysis of extant and fossil genera recovered a clade, Pantetrapulmonata nom nov. , with the topology (Trigonotarbida (Araneae (Hap-topoda (Pedipalpi)))). Arachnida was recovered as monophyletic with the internal relationships (Stomothecata (Pal-pigradi, Acaromorpha (Haplocnemata, Pantetrapulmonata))). Nodal support and exploratory implied weights indicated that relationships among these ﬁve clades were effectively unresolved. Thus, some interordinal relationships were strongly and/or consistently supported by morphology, but arachnid phylogeny is unresolved at its deepest levels. Alternative hypotheses proposed in the recent literature were evaluated by constraining analyses to recover hypothesized clades, an exercise that often resulted in the collapse of otherwise well-supported clades. These results suggest that attempts to resolve speciﬁc nodes based on individual characters, lists of similarities, evolutionary sce-narios, etc., are problematic, as they ignore broader impacts on homoplasy and analytical effects on non-target nodes. © 2007 The Linnean Society of London, Zoological Journal of the Linnean Society , 2007, 150 , 221–265. Results from analyses of node-speciﬁc hypotheses


INTRODUCTION
Despite an ever-increasing reliance on molecular sequence data for phylogeny reconstruction and evolutionary inference, morphological characters remain an important source of phylogenetic signal (both alone and in combination with molecular data) and are essential for reconstructing and exploring patterns in organismal evolution. Construction and maintenance of digital databases of structural information are essential if morphology and the results of morphological analyses are to remain useful. In the present work, I define, homologize and code morphological and other non-molecular characters that vary among orders and major intraordinal groups of arachnids. The results are summarized in a taxon-by-character matrix, and the phylogenetic signal within the matrix is explored using parsimony-based analyses. This research clarifies the strengths and weaknesses of current understanding of arachnid phylogeny and highlights several aspects of phylogenetic practice that may impede progress in the evolutionary morphology and phylogeny of Arachnida.
Results from the present analysis ( Fig. 1) essentially affirm this characterization of our current understanding of arachnid phylogeny based on morphology and other non-molecular characters but also offer new proposals and evaluate alternative interpretations that have emerged in the last 17 years, i.e. since my previous attempt to resolve arachnid phylogeny (Shultz, 1990). Specifically, parsimony-based analyses corroborate the monophyly of Arachnida as well as Uropygi, Pedipalpi, Haplocnemata and Acaromorpha. The data also support more recent proposals, including the monophyly of Opiliones and Scorpiones ( = Stomothecata nom. nov. ) (Shultz, 2000) and Pantetrapulmonata nom. nov. with an internal structure anticipated by Dunlop (1999Dunlop ( , 2002c; i.e. (Trigonotarbida (Araneae (Haptopoda, Pedipalpi))). However, relationships among Palpigradi, Acaromorpha, Haplocnemata, Stomothecata and Pantetrapulmonata are effectively unresolved. These results indicate that morphology offers important phylogenetic information, but it is not yet sufficient to resolve relationships at the deepest levels within Arachnida.

METHODS T ERMINAL TAXA
The study was based on 59 euchelicerate genera (41 extant, 18 fossil), with most represented by one species (detailed below), coded for 202 binary and unordered multistate characters (Table 1, Appendix).
Xiphosura (horseshoe crabs) : The xiphosurans are an ancient (Silurian-Recent) aquatic lineage with its greatest diversity occurring in the fossil record. It comprises two main groups, Synziphosurida (Silurian-Devonian: ~ ten genera) and Xiphosurida (Carboniferous-Recent: ~14 genera). Synziphosurids are probably paraphyletic and retain plesiomorphic features, such as a ten-segmented opisthosoma with three segmented metasoma (Anderson & Selden, 1997). They were represented in the matrix by Weinbergina opitzi , one of the few fossil xiphosurans with preserved appendages (Moore, Briggs & Bartels, 2005), and a more typically preserved synziphosurid, Limuloides limuloides . Extant xiphosurans (three genera, four spp.) were represented by the intensively studied Atlantic horseshoe crab, Limulus polyphemus , and an Asian horseshoe crab, Carcinoscorpius rotundicauda , with supplemental information drawn from another Asian species, Tachypleus tridentatus.
Eurypterida (sea scorpions) : A diverse ( > 60 genera) aquatic group of fossil euchelicerates that ranged from Only those relationships that were well supported by bootstrap analysis or consistently recovered in sensitivity analysis are depicted; deepest relationships within Arachnida are effectively unresolved.
Haptopoda: This is a fossil terrestrial group (Carboniferous) containing one known species, Plesiosiro madeleyi. The known specimens have been reexamined by Dunlop (1999).

Araneae (spiders):
This is a very large order (110 families, ~3600 genera, ~39 000 extant spp.) (Platnick, 2005) represented here by one genus from each of the three principal extant lineages (Platnick & Gertsch, 1976;Coddington & Levi, 1991;Coddington, 2005): Liphistius (Mesothelae), Aphonopelma (Mygalomorphae) and Hypochilus (Araneomorphae). The sample is small with respect to known diversity, but the basal phylogeny and relevant groundplan states of the order are well established and states derived from the represented taxa are consistent with them.

Amblypygi (whipspiders):
This is a small extant terrestrial order (five families, 17 genera, 136 spp.) (Harvey, 2002) and was represented in the matrix by two fairly well-studied genera, Charinus and Phrynus. Charinus retains several features that appear to be plesiomorphic for the order (e.g. adult pedal pulvilli, coxal glands associated with leg 3, eversible vesicles). Weygoldt (1996) has examined the phylogeny of the order and has provided an important summary of morphology and general biology (Weygoldt, 2000).
Thelyphonida or Uropygi s.s. (whipscorpions, vinegaroons): Extant whipscorpions (16 genera, 106 spp.) (Harvey, 2002) were represented here by one wellstudied species, Mastigoproctus giganteus, although additional information was drawn from Typopeltis and Thelyphonus. The morphology of the group is highly conserved; there have been no modern studies of intraordinal phylogeny. A controversial fossil species from the late Carboniferous, Proschizomus (Dunlop & Horrocks, 1995/1996, was also included but less wellpreserved Carboniferous forms were not. Schizomida: This small order (two extant families, 34 genera, ~200 spp.) (Harvey, 2002) is widely regarded as the sister group of Thelyphonida. The taxon sample included one representative from each extant family, Protoschizomus (Protoschizomidae) and Stenochrus (Hubbardiidae), especially S. portoricensis. There are three recent (Pliocene?) fossil species, all from the same locality; these were not included here.
Ricinulei: The extant ricinuleids (three genera, 55 spp.) (Harvey, 2002) were represented by two genera, Cryptocellus and Ricinoides. Detailed studies of gross cuticular anatomy and post-embryonic development are available for a representative of each genus (Cryptocellus: Pittard & Mitchell, 1972;Ricinoides: Legg, 1976) and some information on internal anatomy is available for Ricinoides (Millot, 1945), and this has been extrapolated to Cryptocellus in the matrix. There are two basic fossil types (Selden, 1992), one resembling modern taxa and another with a unique opisthosoma that superficially resembles the closed elytra of a beetle (i.e. curculioids). Two fossils from the former group were included, Terpsicroton and Poliochera, as they are reasonably well preserved and show important characters (e.g. two pairs of eyes or evidence of opisthosomal diplosegmentation) not expressed in extant forms. No curculioid ricinuleids were included, as they appear to offer no additional information relevant to resolving ordinal relationships.

Opilioacariformes (= Opilioacarida, Notostigmata):
The opilioacariform mites (nine genera, 20 spp.) (Harvey, 2002) are generally regarded as plesiomorphic Acari and are fairly conserved in their morphology. The group was represented by two species, Neocarus texanus and Siamacarus withi. N. texanus has a typical opilioacariform morphology, and its external anatomy has been particularly well studied (e.g. Van der Hammen, 1989;Klompen, 2000). S. withi differs from most other opilioacariforms in having trichobothria and three rather than two pairs of lateral eyes (the latter also in Paracarus), features that are potentially significant for assessing ordinal relationships. Some information on internal morphology of Opilioacarus was taken from With (1904) and extrapolated to Neocarus and Siamacarus.

Solifugae (= Solpugida) (sun spiders):
The order is small (~1000 spp.), but the absence of modern phylogenetic treatments precludes meaningful estimates of genera and families (Harvey, 2002). The taxon sample included two genera, one from the New World, Eremocosta, and one from the Old World, Galeodes.

Scorpiones (scorpions):
The extant scorpions (16 families, 155 genera, 1279 spp.) (Fet et al., 2000) were represented by three taxa, Centruroides vittatus (Buthidae), Hadrurus arizonensis (Iuridae) and Heterometrus spinifer (Scorpionidae). Buthidae is widely regarded as the sister group to other extant lineages, and similarities among the represented terminals are likely to be ground plan features of extant scorpions generally. The morphology of fossil scorpions is substantially more diverse (Kjellesvig-Waering, 1986), but their phylogenetic relationships are unclear (but see Jeram, 1998). The fossil taxa used in this study (Proscorpius, Stoermeroscorpio, Palaeoscorpius, Prearcturus) were chosen for quality of preservation and/ or presence of a phylogenetically significant constellation of characters.

CHARACTER CODING
Most character states were determined from direct observation, the primary literature and authoritative reviews. In some cases, states were assigned to termi-nals based on observations from related species, as noted in the Appendix. In the matrix, state '-' indicates that the character is inapplicable because the taxon lacks a more general character. For example, if a taxon lacks eyes, then special features of the eyes (number, position, retinal configuration, etc.) are inapplicable to that taxon. A '?' indicates that the state is unknown or uncertain. An entry with multiple states (e.g. 0/1, 3/4/ 5) should be treated as an ambiguity code, not a polymorphism; it indicates that two or more interpretations of homology are applicable and that assignment was established analytically by character concordance. Although '-', '?' and '0/1' are analytically identical when applied to binary characters, they provide information about the empirical status of the character state in specific taxa. Characters are cited throughout the text as italicized numbers in parentheses and are discussed in the Appendix.

PHYLOGENETIC ANALYSIS
Phylogenetic analysis of the full data matrix was performed using the program Tree Analysis using New Technologies (TNT), ver. 1 (Goloboff, Farris & Nixon, 2000) using 'traditional' search based on 1000 replicates using TBR branch swapping. Results were compared to those obtaining using the ratchet algorithm (Nixon, 1999) to determine any difference due to analytical approach. Nodal support for the minimal-length topology was evaluated by bootstrap (Felsenstein, 1985) and Bremer support (Bremer, 1994). Bootstrap analysis was conducted in TNT and based on 1000 pseudoreplicates each analysed by ten random-addition replicates using TBR branch swapping. A nexus file containing the resulting 1000 trees was imported into PAUP* ver. 10 (Swofford, 2002) to obtain bootstrap frequencies.
Bremer support was determined in TNT by constraining specific nodes in the minimal-length topology and then determining the shortest tree that did not recover the specified clade. The difference in length between the unconstrained and constrained minimal-length trees is the Bremer support. The effect of homoplasy on results was explored by conducting implied weights analysis (Goloboff, 1993) in TNT. Six analyses were conducted, each with constant of concavity (k) set to a different integer value of 1-6, where 1 is weighted most severely against homoplasious characters. Each implied weights analysis was conducted using 'traditional' search based on 1000 replicates using TBR branch swapping. The same procedures were used in analysing a matrix that included only extant taxa. However, characters rendered uninformative by removal of fossil taxa (i.e. 49, 98, 99, 100, 113, 114, 125, 161) were excluded prior to analysis.

COMPARISONS OF ALTERNATIVE HYPOTHESES
Phylogenetic hypotheses proposed in the recent literature were also evaluated, including those that attempted to resolve arachnid phylogeny completely (Fig. 2), their hypothesized subclades (Fig. 3) and hypotheses that proposed only specific nodes (Figs 3,  4). The fully resolved topologies of Weygoldt & Paulus (1979), Van der Hammen (1989), Shultz (1990), Wheeler & Hayashi (1998 and Giribet et al. (2002) were compared with the optimal topology using the Templeton test (Templeton, 1983) as implemented in PAUP*. Internal relationships of multisampled orders were constrained to match those of the optimal topology, unless the original authors explicitly favoured an alternative. Node-specific hypotheses were evaluated by determining the frequency with which the node was recovered in bootstrap analysis. They were also evaluated by constraining parsimony analysis in TNT to recover the shortest tree containing each specific node and then assessing the effect on relative tree length and overall resolution, taking note of effects on otherwise stable or well-supported clades. The entire matrix was used to assess the two cases where node-specific hypotheses involved fossil taxa (Fig. 4).

COMPARISONS OF ALTERNATIVE HYPOTHESES
The hypothesis of Weygoldt & Paulus (1979) (Fig. 2) was 398 steps (15 steps longer than the unconstrained minimal-length tree), Van der Hammen (1989) was 400 steps (17 steps longer), Shultz (1990) was 399 steps (six steps longer), Wheeler & Hayashi (1998) was 391 steps (eight steps longer) and Giribet et al.   Shultz (1990), Wheeler & Hayashi (1998) and Giribet et al. (2002). The Giribet et al. topology is based on neontological data (morphology and molecules) and the original 'ROOT' may be an artefact from use of the highly divergent pycnogonids as an outgroup.  . Consensus trees produced by parsimony analysis of neontological data constrained to produce relationships proposed in the recent literature. The constrained (target) node is indicated by a black dot and the taxa encompassed by the constraint are enclosed in a box. Numbers below each tree represent the number of minimal-length constrained trees, length of minimal-length trees, difference in the length of the unconstrained minimal-length tree and the constrained minimal-length tree, and percentage unconstrained bootstrap trees in which the target node was recovered, respectively. These trees indicate the effect on branch length imposed by specific hypotheses and impact of constraining target nodes on nontarget nodes.
Results from analyses of node-specific hypotheses are summarized in Figures 3, 4. In most cases, proposed relationships were recovered in fewer than 5% of bootstrap pseudoreplicates. The notable exceptions were Rostrosomata (25%), Cryptognomae (24%) and Megoperculata (28%), and the strict-consensus constrained minimal-length tree for each was only one step longer than the unconstrained optimal tree. In most cases, constrained topologies did not add many steps; most were less than five steps longer and none was greater than 14 steps longer than the unconstrained minimal-length tree. It is noteworthy, however, that even constrained nodes that impose relatively few extra steps sometimes supported improbable relationships or eliminated clades that were well supported or stable in the unconstrained analysis. Specific examples and their implications are discussed below.

RECENT ISSUES IN ARACHNID PHYLOGENY
Available morphological evidence consistently resolves some interordinal relationships and fails to resolve others (Fig. 1). Continued progress depends on the ability of arachnologists to discover new characters and to assess the evidence critically. While reviewing the recent arachnological literature, several aspects of phylogenetic practice emerged that seemed counterproductive to both the perception and the actual rate of progress toward resolving arachnid phylogeny. Some of these are summarized here, with specific examples given in the remainder of the Discussion and in the Appendix.
There is a tendency to portray arachnid ordinal phylogeny as more poorly resolved and contentious than is actually the case (Coddington et al., 2004). Phylogenetic hypotheses generated during different historical periods and using differing standards of evidence are often cited as examples of current disagreement (e.g. Selden, 1993;Dunlop, 1996;Selden & Dunlop, 1998;Wheeler & Hayashi, 1998). In fact, recent parsimonybased analyses of morphology have tended to converge on topologies with internal structures congruent with those found here (e.g. Shultz, 1990;Wheeler & Hayashi, 1998;Giribet et al., 2002) (Fig. 2). Arachnid phylogeny is not fully resolved, especially at its deepest levels ( Fig. 1), but this does not mean that all aspects of arachnid phylogeny are controversial or poorly supported by the available evidence.
Matrices are sometimes constructed by uncritical 'recycling' of erroneous or problematic characters based on diverse, secondary sources (Jenner, 2001). Conclusions derived from mixtures of valid, invalid, speculative and redundant characters are sometimes portrayed as the phylogenetic signal provided by morphology. Data recycling can perpetuate errors (see 13,20,39,52,136,140,144,152,169,[171][172][173][174], legitimize speculations (see 13,32,63,95) or create duplicate or non-independent characters (see 13,30,70,77,172). The 'lateral organ' is a notable example. Yoshikura (1975) equated the embryonic/early postembryonic 'lateral organs' of Amblypygi, Thelyphonida and Solifugae with the dissimilar 'lateral organ' of Xiphosura but failed to note the very similar Claparède organ of Acariformes (see 173,174). This coding was recycled by Wheeler & Hayashi (1998) and then by Giribet et al. (2002). Error is probably inevitable when assembling a large matrix from morphology, including the one presented here, but this can be minimized by making original observations and by consulting primary sources.
Some workers advocate weighting characters a priori on the basis of structural or functional complexity (e.g. Kraus, 1998;Dunlop & Braddy, 2001) and dismiss phylogenetic conclusions derived from equal-weights parsimony. In short, these workers criticize parsimony for emphasizing data quantity over data quality. This criticism ignores the intense debate in systematic biology that eventually led to widespread adoption of  . Results from analysis of the full data matrix. A, minimal-length topology. Numbers below internodes are bootstrap percentages/Bremer support values. B, parsimony tree produced by implied weights with k = 1. C, parsimony tree produced by implied weights with k = 2. Parsimony trees produced by implied weights with k = 3-6 are identical to topology C. For B and C, relationships within terminal clades are the same as those shown in A. parsimony and, instead, advocates a return to the speculative and subjective approaches of the late 19th and early 20th century that once threatened the scientific legitimacy of the discipline (Bowler, 1996). Furthermore, it misrepresents the properties of parsimony-based analyses. Specifically, characters with functional significance are readily encompassed by parsimony analyses and, in fact, characters derived from locomotor systems (e.g. 46-94) have played an important role in developing current ideas about arachnid phylogeny (e.g. Shultz, 1989Shultz, , 1991. In addition, 'complex characters' can be viewed as composites of several characters, such that morphological complexity is effectively weighted by the number of independent 'subcharacters' it contains. For example, the 'sucking stomach' once considered a synapomorphy of Labellata (= Araneae + Amblypygi; Fig. 3) is coded here as a composite of three characters (199)(200)(201). The Labellata hypothesis was not corroborated in the present analysis (Figs 1, 3-5) and, in fact, was highly disfavoured (Fig. 3), but this result cannot be dismissed as a failure to acknowledge the complexity of the character. Some workers support specific (target) clades with one or more similarities without exploring the impact on overall homoplasy or relationships among nontarget clades. Each character offers its own phylogenetic hypothesis, which may or may not be consistent with relationships implied by other characters. It is exceedingly rare for all characters to be perfectly compatible in the phylogenetic hypotheses they support, and criteria such as parsimony have been developed to discover those hypotheses that minimize the conflicting phylogenetic signals of different characters. Even though character conflict (homoplasy) is a virtually inescapable phenomenon in comparative biology, it is not uncommon for workers to discover one or more characters and to promote their phylogenetic implications over alternative hypotheses, even those that otherwise appear to be well supported. This approach may have value in highlighting new data and perspectives but accomplishes this by promoting the erroneous impression that all aspects of arachnid phylogeny are so tenuous that a single character can falsify even well-supported hypotheses. Several examples of this approach have appeared in the recent arachnological literature.
For example, Alberti & Peretti (2002) argued that aflagellate spermatozoa (163) are a compelling synapomorphy for a Solifugae + Acari clade and dismissed some of the characters that support Solifugae + Pseudoscorpiones as having 'debatable value.' Yet, their proposal rejects two hypotheses that are consistently recovered in recent phylogenetic analyses: that is, Haplocnemata (= Solifugae + Pseudoscorpiones) and Acaromorpha (= Acari + Ricinulei) (Figs 1, 2). Rejec-tion of Haplocnemata would require its presumed synapomorphies to be reinterpreted as homoplasies, including features of the chelicerae (18,19,20), preoral chamber (13, 32), legs (12, 48) and respiratory system (126). The same reason applies to Acaromorpha and its synapomorphies. It is noteworthy that a Solifugae + Acari clade was recovered in the present analysis in fewer than 5% of bootstrap pseudoreplicates and that analyses constrained to recover this clade were five steps longer than the minimal-length tree and favoured a problematic clade uniting Ricinulei and Pseudoscorpiones (Fig. 3).
In another example, Dunlop (1996) proposed a close relationship between Trigonotarbida and Ricinulei based on two-segmented chelicerae (18), prosomaopisthosoma coupling mechanism (96), diplotergites (100, 101) and longitudinally divided opisthosomal tergites (115). However, phylogenetic analyses constrained to recover this relationship required eight additional steps, eliminated support for Acari and necessarily rejected Acaromorpha and forced its synapomorphies to be reinterpreted as homoplasies. Alberti (2005) proposed an interesting hypothesis for the evolution of male gonads in tetrapulmonates (153) but chose to accept Labellata (= Araneae + Amblypygi) in developing his argument over the much more wellsupported Pedipalpi (= Amblypygi + Uropygi) (Shultz, 1999), a phylogenetic reconfiguration that was recovered here in fewer than 5% of bootstrap pseudoreplicates, increased tree length by a minimum of 14 steps and resulted in the collapse of Stomothecata and Haplocnemata (Fig. 3). Many other examples can be cited.
Promoting or defending a specific phylogenetic hypothesis via lists of compatible synapomorphies is a common but problematic approach. By restricting attention to the states of specific characters at one or two target nodes, one can easily overlook the unintended impact of the hypothesis on phylogenetic signal elsewhere and its effect on non-target clades. A node supported by a long list of synapomorphies may seem convincing taken in isolation but may become less acceptable when its full phylogenetic implications are explored.

Arachnida Lamarck, 1801
Analyses consistently recovered Arachnida as a monophyletic group with high bootstrap support (Figs 1,4,5). Possible synapomorphies include the loss of the carapacal pleural doublure (9), cardiac lobe (10), pedal gnathobases (52) and moveable endites (53) and the gain of aerial respiration (120) and an anteriorly or anteroventrally directed mouth (185). Some traditional synapomorphies, such as slit sensilla (142) and fluid feeding (184), may have appeared later in arachnid evolution, but this can only be decided once the internal phylogeny of Arachnida has been determined.
Some workers regard many proposed arachnid synapomorphies as adaptations to terrestrial life and thereby link the hypothesis of arachnid monophyly to the hypothesis of a single ancestral aquatic-to-terrestrial transition and arachnid polyphyly to multiple transitions (e.g. Selden & Jeram, 1989;Dunlop, 1997;Dunlop & Webster, 1999;Dunlop & Braddy, 2001). Given this line of reasoning, the existence of apparently aquatic scorpions in the fossil record (Kjellesvig-Waering, 1986;Jeram, 1998) and the inference that terrestrialization occurred late in scorpion evolution, these workers conclude that character states supporting arachnid monophyly are actually convergences and do not necessarily support arachnid monophyly.
However, this approach to assessing phylogenetic hypotheses is founded on overly simplistic assumptions, such as the ability of the investigator to discriminate unerringly between characters that exist exclusively in aquatic organisms (including fossils) from those that occur exclusively in terrestrial organisms. There also appears to be an assumption that homoplasy can be generated through parallelism (i.e. multiple aquatic-to-terrestrial events) but not through terrestrial-to-aquatic reversals. Furthermore, an assumed dichotomy between exclusively aquatic and terrestrial life histories in ancestral arachnids is simplistic, as illustrated by the amphibious life cycles of basal vertebrates and pterygote hexapods. In fact, these examples demonstrate that there is no necessary inconsistency in basing a hypothesis of arachnid monophyly on the derived terrestrial features of an amphibious ancestor whose descendants then completed terrestrialization once or several times independently or even returned to a fully aquatic existence. Workers who link the frequency and direction of aquatic-terrestrial transitions to the assessment of arachnid phylogeny do so by endowing themselves with substantially greater insight than seems prudent, by ignoring the huge gaps in our understanding of early arachnid evolution, and by denying to arachnids the evolutionary complexity known to exist in other groups.
Several palaeontologists have been particularly active during the past decade in proposing new characters with the stated goal of removing Scorpiones from Arachnida and erecting a Scorpiones + Eurypterida clade (Braddy & Dunlop, 1997;Dunlop & Braddy, 1997;Dunlop, 1998;Braddy et al., 1999;Dunlop & Webster, 1999). Dunlop & Braddy (2001) recently summarized this evidence and conducted a parsimony-based analysis of Xiphosura, Eurypterida, Scorpiones, Opiliones and Tetrapulmonata (but not Haplocnemata, Acaromorpha or Palpigradi) using 33 morphological characters. Their analysis produced a topology congruent with those generated here (Figs 1, 5, 6), including recovery of Stomothecata (= Scorpiones + Opiliones). However, they rejected this result as a product of 'empirical cladistics' because it gives the same weight to prosomal characters that support arachnid monophyly and to selected opisthosomal characters that support their favoured Eurypterida + Scorpiones clade, namely, a five-segmented metasoma (116), suppression of opisthosomal tergite 1 (95), loss of lamellate respiratory organs on the postgenital somite (122), Kiemenplatten (125), loss of respiratory lamellae on the genital segment (121) and a 'nonstaining' exocuticle (but see Grainge & Pearson, 1966 for evidence of this in Opiliones; see Appendix for comments on the other characters).
A Eurypterida + Scorpiones clade was not favoured in the present analysis (Fig. 6); the strict consensus of minimal-length trees constrained to recover this clade ( Fig. 4) is nine steps longer than the tree recovered by analysis without this constraint. It is also noteworthy that Opiliones was consistently reconstructed as the sister group to Eurypterida + Scorpiones in the minimal-length constrained trees, a provocative result that was probably unintended and unanticipated by supporters of the Eurypterida + Scorpiones. Given that Dunlop & Braddy (2001) reject equal weights parsimony as an arbiter of phylogenetic hypotheses, they would presumably dismiss these results as irrelevant to their argument, just as they dismissed their own parsimony-based results. However, if the Eurypterida + Scorpiones hypothesis is to be credible it must be open to evaluation and potential falsification using objective criteria, and the subjective or intuitive a priori weighting of characters advocated by Dunlop & Braddy clearly does not qualify. At present, it is unclear how one would objectively evaluate Dunlop & Braddy's proposal with criteria compatible with those used in its original formulation. For now, the Eurypterida + Scorpiones concept advocated by Dunlop & Braddy may persist outside the mainstream of modern systematic practice, but it is increasingly problematic within it.
Stomothecata nom.nov. Opiliones and Scorpiones were consistently recovered as a monophyletic group. The proposed name acknowledges a unique preoral chamber, the stomotheca, formed by coxapophyses of the palp and leg 1 (50), often with an auxiliary role played by the coxapophysis of leg 2 (51). In addition, the epistome appears to have been modified for adduction of the palpal coxae. The lateral walls of the epistome are fused to the medial surfaces of the palpal coxae, and the epistomal lumen is spanned by a transverse muscle (188), which apparently adducts the palpal coxae thereby constricting the stomothecal chamber. Scorpions and opilions are also unique in having a pair of large epistomal arms projecting rearward into the prosoma and attaching to the endosternite (189). The epistomal arms provide attachment sites for pharyngeal dilator muscles (196) and extrinsic muscles of anterior prosomal appendages (e.g. 37). The chelicera is equipped with a muscle that arises on the carapace and inserts on the ventral margin of the second cheliceral article (23). There is an anteriorly placed genital opening (155).
Several notable similarities were found while reviewing the literature, but information from other arachnid groups was considered too incomplete to allow them to be included in the matrix. For example, both orders have apparent haemocytopoeitic organs associated with major nerves of the anterior opisthosoma. These are termed supraneural organs in scorpions (Farley, 1999) and perineural organs in opilions (Kästner, 1935). Haemocytes develop in the cardiac wall in spiders (Seitz, 1972) and perhaps amblypygids (Weygoldt, 2000), but haemocytopoeitic organs are unknown in most other arachnid groups. Germ cells differentiate very early during embryogenesis in both scorpions and opilions relative to spiders (Moritz, 1957;Anderson, 1973). Additional research is required to determine the phylogenetic utility of these characters.

Acaromorpha Dubinin, 1957
A clade comprising Ricinulei and Acari is consistently recovered in the present analysis. It is united here by two unique and seven homoplasious synapomorphies. These include a gnathosoma (31) defined, in part, by medial fusion of the palpal coxae (30), although presence of a gnathosoma in Ricinulei is debatable, as discussed in the Appendix. Acaromorphs also have a unique post-embryonic development consisting of a hexapodal larva and up to three octopodal nymphal instars (176). The pedal patella-tibia joints are formed by a bicondylar dorsal hinge rather than a single middorsal condyle (69), all postcheliceral segmental ganglia are unified in the subesophageal ganglion (130), and a postcerebral pharynx is absent (192). The group is also tentatively united by presence of differentiated pedal basi-and telofemora (63, 64) and the absence of a ventral (sternal) pharyngeal dilator muscle (199), but these may be symplesiomorphies erroneously reconstructed as a synapomorphies.
The internal phylogenetic structure of Acaromorpha is controversial, with many recent workers favouring a monophyletic Acari (Weygoldt & Paulus, 1979;Shultz, 1990; argued most thoroughly by Lindquist, 1984) and others advocating a diphyletic Acari, with Acariformes being the sister group to Cryptognomae (= Ricinulei + Anactinotrichida) (especially Van der Hammen, 1979 (Fig. 3). Acari was recovered as monophyletic when fossils were excluded (Fig. 2), but not when they were included (Fig. 3). A 'mite-centred' survey of arachnid characters may be needed if morphology is to offer a compelling solution to the internal phylogeny of Acari and its placement within Arachnida. These issues are not resolved by the present analysis.
Pantetrapulmonata nom. nov. Pantetrapulmonata includes the extinct orders Trigonotarbida, Haptopoda and the extant orders Araneae, Amblypygi, Schizomida and Thelyphonida. The clade is united by cheliceral structure (18, 19), a megoperculum (106), booklungs on the genital and first postgenital somites (121, 122) and enlargement of the epipharyngeal sclerite (192). It is important to note, however, that most of these characters were coded as uncertain in Plesiosiro (Haptopoda). Aside from the placement of Haptopoda, the monophyly of Trigonotarbida and the extant orders was anticipated by Shear et al. (1987) and is generally regarded as a monophyletic group.
Tetrapulmonata Shultz, 1990 Tetrapulmonata was originally proposed on the basis of neontological analyses and encompassed Araneae, Amblypygi, Schizomida and Thelyphonida (Shultz, 1990) and was also recovered here in analysis of neontological data (Fig. 5). However, results from analysis of all taxa required that the Tetrapulmonata concept be expanded to include the fossil order Haptopoda. Features uniting this group are problematic, however, as many states in Haptopoda were coded as uncertain.
Schizotarsata nom.nov. Haptopoda and Pedipalpi (= Amblypygi + Schizomida + Thelyphonida) are united here in a group named for possession of divided pedal telotarsi (84). Synapomorphies include a pointed anterior carapacal margin (3) and elongation of leg 1 (46). As already noted, the placement of Haptopoda should probably be regarded as tentative because the state of many characters in this group was coded as unknown. The placement of Haptopoda as the sister group to Pedipalpi was anticipated by Dunlop (1999Dunlop ( , 2002c.

Uropygi Thorell, 1882
Uropygi has long been accepted as a monophyletic union of Thelyphonida and Schizomida. Synapomorphies include a unique mating behaviour (159), fused palpal coxae (30), 2-1-1-1 arrangement of tibial trichobothria (144), posterior defensive glands (102), elongated patella of leg 1 (68) and many others.  (Scholl, 1977;Anderson & Selden, 1997;. The fossil record suggests that this structure originated with the disappearance of the already reduced first opisthosomal tergite (= microtergite) and axial portion of the second opisthosomal tergite (Anderson & Selden, 1997 (6). The condition in Acari is controversial but is coded here as State 1. Van der Hammen (1989) considered the tergal region of the leg-bearing somites (postoral somites III-VI), or podosoma, to be reduced and replaced by posterior migration of the aspidosoma (i.e. tergal region assumed to be associated with appendages of the gnathosoma) and anterior migration of the opisthosomal tergal region to form the hysterosoma. The dorsal proterosomal and hysterosomal elements meet at a transverse sejugal furrow that continues laterally and passes ventrally between the coxae of legs 2 and 3. According to this scheme, the prosomaopisthosoma border is expressed as the disjugal furrow, which passes from the ventral posterior margin of the podosoma anterodorsally to join the dorsal part of the sejugal furrow. Based on the arrangement of setae and slit sensilla (= lyrifissures), Van der Hammen (1989) interpreted the region above coxae 3 and 4 (i.e. region C) as a fusion of the dorsal parts of the first two opisthosomal somites (= postoral somites VII and VIII) in early divergent Anactinotrichida (i.e. Opilioacariformes) and Acariformes (e.g. Alycus). These interpretations have been followed by many acarologists, although its speculative aspects are widely acknowledged (Evans, 1992;Alberti & Coons, 1999).
Scorpiones: Two principal hypotheses regarding the number of opisthosomal somites in scorpions have been advocated: a 13-somite hypothesis derived from embryological studies (Brauer, 1895;Patten, 1912;Farley, 1999Farley, , 2005 and a 12-somite hypothesis based on comparative anatomy of adults (Weygoldt & Paulus, 1979). The embryological interpretation is based on the observation of pregenital, genital and pectinal somites (each with segmental ganglia and paired limb buds) in early scorpion embryos followed by extreme reduction or loss of the pregenital somite in later embryos. According to this view, a missing pregenital somite should be added to the 12 apparent opisthosomal somites of post-embryonic scorpions to achieve a final number of 13. The anatomy-based hypothesis was introduced by Weygoldt & Paulus (1979), who advocated a literal interpretation of post-embryonic segmentation based on opisthosomal tergites. Specifically, these authors argued that the last pair of dorsal endosternal suspensor muscles of non-scorpion arachnids, especially Pedipalpi, attach to the first (= pregenital) somite, that this condition also occurs in scorpions, and that there is no reason to invoke a missing pregenital tergite. They proposed that the pectines belong to the genital somite and that functional specializations of the nervous system for pectinal function give the apparence of an extra neuromere during embryonic development.
I recently dissected the prosoma and anterior opisthosoma of the scorpions Centruroides, Hadrurus and Heterometrus and focused on the composition of the muscular diaphragm (103) that separates the haemocoelic compartments of the prosoma and opisthosoma . The diaphragm is composed of a metameric series of axial muscles from three somites; the anterior somite corresponds to the last prosomal somite and the posterior somite corresponds to the genital somite. The middle elements insert dorsally along a tranverse tendon attached to the anterior margin of the first tergite. These observations are consistent with the embryological interpretation that a pregenital somite is present but its tergite is not expressed. It appears likely that the the pregenital somite was compressed longitudinally during the evolution of the diaphragm. I code scorpions as having 13 opisthosomal somites.
Eurypterida: The eurypterid opisthosoma is widely assumed to have 12 somites (Clarke & Ruedemann, 1912;Størmer, 1944). However, in a speculative paper on the evolutionary morphology of trilobites and chelicerates, Raw (1957) proposed that both scorpions and eurypterids have 15 opisthosomal somites. Raw assumed that scorpions have 13 apparent opisthosomal somites based on the transient pregenital somite of scorpion embryology, that scorpions and eurypterids are close relatives and should have the same number of somites, and that olenellid trilobites and chelicerates always have somites in multiples of three. Raw achieved 15 somites in scorpions by assuming the last opisthosomal somite to be a diplosomite and that the telson is a postanal somite, even though there is no evidence for either of these proposals. He attributed these features to eurypterids, as well. He also noted that the connection between the prosomal carapace and first opisthosomal tergite in Eurypterida differed structurally from the connection between adjacent opisthosomal tergites and regarded this as evidence for a reduced pregenital tergite in the prosomaopisthosoma junction. Raw's speculations were largely forgotten until Dunlop & Webster (1999) resurrected the proposal that eurypterids have a reduced opisthosomal tergite and therefore share a unique similarity with scorpions. Unfortunately, Dunlop & Bullock treated Raw's conjecture as if it were based on empirical evidence rather than an attempt to force eurypterid morphology into a peculiar numerical system. In the absence of convincing evidence to the contrary, I have coded eurypterids as having 12 opisthosomal somites.

Ricinulei:
The opisthosoma consists of a membranous pedicel (97) bearing the gonopore ventrally. The female gonopore is bordered by an anterior plate and a posterior plate. The remainder of the opisthosoma is composed of thick sclerites separated by less heavily sclerotized cuticle. The first dorsal sclerite is short and functions as part of a prosoma-opisthosoma coupling mechanism (96) that may or may not be a specialized component of the following tergite. This is followed by four tergites and sternites and a three-segmented metasoma (= 'pygidium'). Millot (1945 reasoned that the opisthosoma contains ten somites, with somites VII-IX incorporated into the pedicel, X-XIII expressed as tergites and sternites, and the metasoma comprising four somites. He did not regard the dorsal coupling sclerite as a separate tergite, and his interpretation of four rather than three metasomal somites has been rejected. Pittard & Mitchell (1972) also proposed ten opisthosomal somites, but achieved this number by regarding the dorsal coupling sclerite as a tergite of somite IX and by recognizing three metasomal somites. Legg (1976) adopted the system proposed by Pittard & Mitchell but did not regard the dorsal coupling sclerite as separate from the following tergite (X). Van der Hammen (1979) reconstructed 13 somites. The pedicel was regarded as having two somites (VII + VIII); the following four sets of tergites and sternites were considered diplosomites (IX + X, XI + XII, XIII + XIV, XV + XVI) and the metasoma had three somites (XVII-XIX). This interpretation may have been biased by Van der Hammen's attempt to unite Ricinulei with anactinotrichid Acari, which he also regarded as having a primitive number of 13 somites. Selden (1992) described a fossil ricinuleid, Terpsicroton, that shows two pairs of depressions on the three large premetasomal tergites, which contrasts with the single pair seen in extant species. This observation appears to corroborate the diplosomite hypothesis, but the evidence does not indicate that the tergite anterior to these is a diplosomite. Dunlop (1996) attempted to homologize the prosoma-opisthosoma coupling mechanisms of Ricinulei and Trigonotarbida, a goal that required a novel and rather forced interpretation of the dorsal sclerites. He regarded the coupling sclerite as homologous with the first opisthosomal tergite (VII) of trigonotarbids and then followed Van der Hammen's diplosomite hypothesis to achieve 12 somites in total. Dunlop's scheme differs from previous systems in suggesting that the pedicel does not contain the dorsal elements of the first and second somites and is inconsistent with Millot's (1945) observation that the pre-and postgenital plates each have dorsoventral muscles. Here I code Ricinulei as having 12 somites. There are three metasomal somites (XVI-XVIII), three diplosomites (= six somites) (X-XV), one coupling somite (IX) and two somites in the pedicel (i.e. the genital and pregenital somites) (VII, VIII).
Opiliones: Harvestmen have nine opisthosomal somites and an anal operculum that is traditionally regarded as the tergite of a tenth somite; a tenth sternite is lacking (Hansen & Sørensen, 1904;Winkler, 1957). However, the anal operculum appears to represent a persistent embryonic telson (Moritz, 1957) and is therefore likely to be a postsegmental structure comparable with the stinger of scorpions, flagella of thelyphonids, etc. (Shultz, 2000).
Acari: The number of opisthosomal somites is probematic for most mite taxa due to uncertainty about the location of the prosoma-opisthosoma boundry (5), extensive simplification or loss of metamerically arranged sclerites and muscle attachments, a paucity of developmental studies of engrailed expression and, in Acariformes, opisthosomal anamorphosis and heterochronic modification of somite number (Evans, 1992). Ixodids appear to be the exception; developmental studies indicate five somites in the opisthosoma of ticks (Evans, 1992).
Some mites retain external evidence of segmentation -metameric patterns of furrows, muscle attachments and slit sensilla -and, with certain assumptions, the number of opisthosomal somites can be estimated. Here it is assumed that the dorsal surfaces of the last two prosomal somites are present in mites and retain their primitive association with legs 3 and 4 (contra Van der Hammen, 1989) (see 5 for justification). Given this, there appear to be 11 somites in the opisthosoma of Opilioacariformes, a conclusion also reached by other workers (e.g. With, 1904;Sitnikova, 1978;Klompen, 2000). Similar reasoning suggests that Alycus has seven opisthosomal somites, not nine as advocated by Van der Hammen (1989). Unfortunately, external evidence is ambiguous for determining the number of somites in other mite lineages and these are coded as uncertain. 96. Prosoma-opisthosoma coupling mechanism: 0, absent; 1, present [GEWB ∼24] The posterior margin of the carapace and the anterior margin of the first apparent opisthosomal tergite are specialized as a coupling mechanism in Ricinulei and Trigonotarbida, and Dunlop & Horrocks (1996) proposed that these are synapomorphic for the two groups. However, there is uncertainty about the homology of the anterior opisthosomal somites in Trigonotarbida and Ricinulei (95). 97. Pedicel: 0, absent; 1, aranean type; 2, ricinuleid type [WP 30, S ∼40, WH ∼20, GEWB ∼126] The body narrows at or near the prosomaopisthosoma juncture in several arachnid lineages (i.e. Solifugae, Palpigradi, Amblypygi, Araneae, Ricinulei) and this 'waist' has often been used as character at the interordinal level (e.g. Pocock, 1893). However, this interpretation is rejected here for being subjective and for uniting non-homologous conditions. For example, Araneae and Amblypygi are often grouped on the basis of a 'pedicel', yet it is a highly specialized structure in Araneae and its parts are not readily homologized with those of Amblypygi. In contrast, the condition in Amblypygi is a slightly narrower version of the highly moveable prosoma-opisthosoma juncture in Uropygi, which is not generally considered a 'pedicel.' The pedicel in Ricinulei is also unique: a weakly sclerotized stalk containing the genital opening (Cryptocellus: Pittard & Mitchell, 1972;Ricinoides: Legg, 1976 (Hansen & Sørensen, 1905;Cokendolpher & Reddell, 1992;Reddell & Cokendolpher, 1995).

RESPIRATORY SYSTEM
120. Respiratory medium: 0, water; 1, air. Among the terminal taxa included, only fossil scorpions are problematic for this character. They are coded here as uncertain. Based on a study of book lung microsculpture, Scholtz & Kamenz (2006) have argued that arachnids are primitively terrestrial and pulmonate (see also Firstman, 1973) and have questioned whether any fossil scorpions were aquatic. No position on this proposal is taken here. 121. Respiratory lamellae on opisthosomal somite 2 (= genital somite, postoral somite VIII): 0, absent; 1, present [WP ∼37, S ∼51, WH ∼22, GWEB ∼133-137] This character encompasses book gills (= lamellae that function in water) and book lungs (= lamellae that function in air). The distinction between book gills and book lungs is accommodated here by 120 in combination with 121-124. Respiratory lamellae are present on the genital somite in Trigonotarbida, Araneae, Amblypygi, Thelyphonida and Schizomida. Petrunkevitch (1949) reconstructed Plesiosiro as having book lungs, but Dunlop (1999) could not corroborate this. This character was used by Dunlop & Webster (1999) to propose that Xiphosura and Scorpiones are closely related because they both lack respiratory lamellae on the genital somite. Dunlop & Braddy (2001) also argued for the placement of Eurypterida with Xiphosura and Scorpiones based, in part, on this character. At least some eurypterids may have had respiratory lamellae (e.g. Manning & Dunlop, 1995), but the only evidence of their segmental distribution is derived from one specimen (Braddy et al., 1999 , 1912), where they take the form of ventrally projecting cones with a distinct cuticular microsculputure (Manning & Dunlop, 1995). Dunlop & Braddy (2001) inferred the existence of Kiemenplatten in all Palaeozoic scorpions based on a description and photos of one specimen of Paraisobuthus duobicarinatus by Kjellesvig-Waering (1986: pls 16-18). The plates depict dark cone-like, rearward-pointing denticles distributed within a white amorphous material. The denticles appear to lack microsculpture, even though the magnifications at which the photos were taken (×90-×330) are comparable with those illustrating the cones of Kiemenplatten (×170-×350 in Manning & Dunlop, 1995: figs 1, 2). Given the diversity of cuticular structures present in the atria of booklungs, book gills and tracheae, the denticles appear to bear no special similarity to the cones of Kiemenplatten. 126. Tracheal system: 0, absent; 1, paired ventral stigmata on postoral somite VIII (= opisthosomal somite 2); 2, paired ventral stigmata on postoral somites IX and X; 3, one pair of stigmata opening near legs 3 or 4; 4, paired stigmata associated with chelicerae; 5, four pairs of stigmata on dorsal surface of opisthosoma ; -, inapplicable, aquatic (120) [WP 40, ∼43, S ∼52 + 53, 54, WH ∼23, 45, 88, GEWB ∼138, 139] Firstman (1973) and Weygoldt & Paulus (1979) hypothesized that tracheae are homologous in all tracheate arachnids except spiders, and subsequent workers have entertained this hypothesis by including a character for the presence/absence of tracheal systems that ignores the diverse arrangement of stigmata in arachnids (e.g. Shultz, 1990;Wheeler & Hayashi, 1998;Giribet et al., 2002). However, this approach assumes that internal tracheal systems are conserved but that tracheal openings (stigmata) appear and disappear on different parts of the body with higher evolutionary frequency. Here I assume that stigmata are conserved in evolution and that differences in their anatomical placement reflect the evolution of new tracheae. State 1 occurs throughout Opiliones. State 2 occurs throughout Solifugae not Alycus), Prostigmata (Microcaeculus) and certain oribatids (Palaeacarus) (Evans, 1992;Alberti & Coons, 1999). Giribet et al. (2002) coded all representative Acari as lacking median eyes. It is not known whether the eyes of some cyphophthalmid opilions are median or lateral; evidence from tracheal branching in Cyphophthalmus (Janczyk, 1956) suggests that they are median eyes (Shultz & Pinto da Rocha, 2007) and presence of a tapetum in Stylocellus is consistent with lateral eyes (Shear, 1993). 137. Retinula cells of dorsal median eyes: 0, organized into closed rhabdoms; 1, organized into network of rhabdomeres; 2, disorganized; -, inapplicable due to absence of median eyes (136) [GEWB ∼3] State 0 is present in Scorpiones, Thelyphonida and Amblypygi, State 1 is present in Solifugae and Araneae; state 2 is present in Xiphosura (Paulus, 1979). The retinula cells of median eyes have been studied in several prostigmatid Acariformes and are organized in a network in some taxa and in an irregular pattern in others (Alberti & Coons, 1999). Retinulae in phalangid Opiliones have state 0 proximally and state 2 distally (Schliwa, 1979). 138. Ventral median eyes: 0, absent; 1, present. State 1 occurs in early instars of extant Xiphosura (Paulus, 1979). 139. Lateral eyes: 0, absent; 1, present. Lateral eyes are primitively present in Chelicerata and are absent in Palpigradi and Opiliones (Paulus, 1979). It is unclear whether the eyes of cyphophthalmid opilions are median or lateral (see 136). 140. Arrangement and number of lateral eyes: 0, compound, many; 1, five or more pairs (includes microlenses); 2, three primary pairs (excludes microlenses); 3, two pairs; 4, one pair; -, inapplicable due to absence of lateral eyes (139) [WP ∼13 + 18 + 38 + 44 + 52, S ∼49, WH ∼10, GEWB ∼4 + 5] True compound eyes are present in Xiphosura, Eurypterida, Chasmataspidida and many Palaeozoic Scorpiones. Among the anactinotrichid Acari, Opilioacariformes have two (Neocarus) or three (Siamacarus) pairs of lateral eyes (although at least one species of Siamacarus lacks eyes), the allothryrid Holothyrida and many Ixodida have a single pair of lateral lenses and the remainder apparently lack eyes (Evans, 1992). Thelyphonida was coded as '1/2' to reflect five pairs of lenses comprising three pairs of primary lenses and two pairs of small accessory lenses (not simply three pairs as coded by Giribet et al., 2002). Trigonotarbida is also coded as '1/2' to reflect three pairs of primary lenses and multiple small accessory lenses. Dunlop (1999) illustrated Haptopoda as having paired triads of lateral eyes but noted that there was actually no evidence of this in the fossils. The lateral eyes of fossil ricinuleids have two pairs of lenses (Selden, 1992). One pair of eyes or eyespots are present in extant Ricinulei and many Schizomida, although five genera of hubbardiid Schizomida have a pair of lenses (Reddell & Cokendolpher, 1995). State 1 is the groundplan for extant Scorpiones, and State 3 occurs in the pseudoscorpions Chthonius, Feaella, Neobisium (but State 4 in Chelifer) and Solifugae (Paulus, 1979). 141. Lateral eyes with closed rhabdoms: 0, absent; 1, present; -, inapplicable due to absence of lateral eyes (139) [WP 21,WH 12,GEWB 6] Presence of closed rhabdoms is probably primitive for Chelicerata and is retained in extant Xiphosura and Scorpiones. Retinula cells form a network of rhabdomeres in other extant chelicerates (Paulus, 1979), but these networks can differ substantially in detail. Weygoldt (1998) and other workers have given substantial weight to the network character in uniting non-scorpion arachnids, but these authors seem not to grant comparable phylogenetic significance to the analogous character of the median eyes shared by Solifugae and Araneae (137). 142. Slit sensilla: 0,absent;1,present [WP 19,S 47,WH 11,GEWB 209] State 1 occurs in all arachnid orders except Palpigradi (Shultz, 1990). The proposal that 'primitive' slit sensilla were present in Eurypterida (Dunlop & Braddy, 1997) appears to be based on the over-interpretation of a comparatively large notch that occurs in Baltoeurypterus at the terminus of the tibia (= podomere 7) (see also Edgecombe et al., 2000;Giribet et al., 2002). The neural construction of slit sensilla, like functionally similar campaniform sensilla of hexapods, is similar to that of trichoid sensilla (Chapman, 1998;Klompen, 2000). Thus, these cuticular stress receptors may represent modified bases of sensory setae, a view supported by the replacement of setae by slit sensilla during post-embryonic development in some Opilioacariformes (Klompen, 2000). 143. Trichobothria: 0,absent;1,present [GEWB 213] Trichobothria are present in extant Scorpiones (Jeram, 1998), Pseudoscorpiones (Chamberlin, 1931, some endeostigmatid (Alycus: Van der Hammen, 1989) and prostigmatid Acariformes, and most Oribatida (Lindquist, 1984). They occur on the ventral surface of the pedal femora in the opilioacariform Siamacarus (Leclerc, 1989). They are also present in Araneae (Foelix, 1996), Amblypygi (Weygoldt, 2000), Schizomida and Thelyphonida (Hansen & Sørensen, 1905). They are apparently absent in all non-arachnid chelicerates, Solifugae, Ricinulei, Opiliones (Reissland & Görner, 1985) and Parasitiformes (Acari) (Lindquist, 1984). 144. Tibial trichobothria with 2-1-1-1 pattern on appendages III-VI (= arachnid legs 1-4): 0, absent; 1, present; -, inapplicable due to absence of trichobothria (143) [S 48,WH 87,GEWB 88] State 1 occurs only in Thelyphonida and Schizomida (Hansen & Sørensen, 1905). Note that Shultz (1990) erroneously described this character as a 2-2-1-1 pattern and that this error was repeated by Giribet et al. (2002) and described as 2-1-1 by Wheeler & Hayashi (1998). 145. Paired trichobothria on dorsal surface of prosoma: 0, absent; 1, present; -, inapplicable due to absence of trichobothria (143) State 1 is an apparent synapomorphy of Acariformes and is present in all acariforms included in this study (Alberti & Coons, 1999 (2000) has noted that State 1 occurs in Opilioacariformes, Parasitiformes (except Mesostigmata) and Ricinulei, where it also occurs on leg 2 (Talarico et al., 2005). State 1 occurs on all legs in Araneae and, perhaps, Scorpiones (Foelix, 1985). State 0 is limited to Xiphosura (Sekiguchi, 1988); State 1 occurs throughout Arachnida (Millot, 1949a Clarke (1979) in coding gonads as reticulate (Xiphosura), ladder-like (Scorpiones, Thelyphonida, Schizomida) or 'saccular' (all remaining Arachnida). However, comparisons between Xiphosura and Arachnida are problematic given that the xiphosuran gonads are primarily prosomal and those of arachnids are primarily opisthosomal (151). Further, the reticulate pattern in Limulus (Xiphosura), but not other extant xiphosurans, contains a distinctly ladder-like component. The 'saccular' state is probably artificial, as it encompasses a wide variety of paired and unpaired structures. 153. Male gonads in two distinct parts, one producing sperm and another (tubular gland) producing a holocrine secretion similar to degenerate sperm: 0, absent; 1, present. State 1 occurs in Thelyphonida, Schizomida and Amblypygi, although the holocrine material is produced by ventral organs in Amblypygi and dorsal organs in Thelyphonida and Schizomida (Alberti, 2005). 154. Number of gonopores: 0, two; 1, one. Extant xiphosurans have two small genital openings on the base of the genital telopodite, and all extant arachnids have a single opening. The condition in Eurypterida is not known. Clarke & Ruedemann (1912) located a pair of openings near the base of the median organ (161) that are the outlets of the 'horn organs', but it is unclear whether these are genital ducts or accessory structures. Braddy & Dunlop (1997) have developed numerous speculations about these structures and extended their arguments far beyond the available evidence. 155. Genital opening (i.e. gonopore or gonostome) appearing to open in prosomal region (i.e. between leg coxae or anterior to posterior carapacal margin): 0, absent; 1, present [WP ∼50, WH 26, GEWB 166] The genital opening in Euchelicerata is located on postoral somite VIII (= opisthosomal somite 2), but it has shifted anterior to the posterior margin of the carapace or between the coxae of the last pair of legs in most Scorpiones (but not in Palaeoscorpius: Kjellesvig-Waering, 1986) and Opiliones. The genital opening occurs near or anterior to the last coxae in the opilioacariform and parasitiform Acari represented here. It is variable in Acariformes but is located posterior to the coxae in all representative taxa. 156. Ovipositor: 0, absent; 1, present [WP ∼51, S ∼60, WH ∼91, GEWB ∼172] An ovipositor with a trilobed terminus is an apparent groundplan character of Oribatida (Lindquist, 1984;Alberti & Coons, 1999). An ovipositor is also present in Opilioacariformes and Opiliones (Van der Hammen, 1989