RELATIONSHIPS

Platnick (1990) allocated those gnaphosoids with unmodified piriform gland spigots to five families (the Cithaeronidae, Gallieniellidae, Trochanteriidae, Ammoxenidae, and Lamponidae), and suggested that the other two gnaphosoid families (the Gnaphosidae and Prodidomidae) might be sister groups, united by the loss of a sclerotized ring on the anterior lateral spinnerets. The sclerotized ring apparently represents the remnants of a plesiomorphically present distal article on those spinnerets. Such a ring is found in the various "clubionoid" families that presumably include the closest relatives of gnaphosoids. For that 1990 study, only a few specimens of L. cylindrata were available for study, and none of those specimens had well-extended spinnerets. Subsequent investigation has shown that L. cylindrata and the other taxa assigned below to the Lamponidae actually differ, on one hand, from the Gnaphosidae and Prodidomidae, and on the other, from the remaining gnaphosoid families as well.

In the Gnaphosidae and Prodidomidae, the sclerotized ring is represented at most by a tiny crescent of sclerotized, seta-bearing cuticle found only around the base of the major ampullate gland spigots; the piriform gland spigots are surrounded only by soft cuticle, so that the anterior lateral spinnerets have a single article (i.e., are "unisegmented"). In the Cithaeronidae, Gallieniellidae, Trochanteriidae, and Ammoxenidae, the distal article is clearly marked by a complete ring of sclerotized, seta-bearing cuticle, so that the anterior lateral spinnerets are "bisegmented." In contrast, L. cylindrata and its relatives actually have an intermediate condition; the sclerotized ring is present but incomplete, being interrupted along its posterior edge by soft, unsclerotized cuticle that connects the area bearing the piriform gland spigots with the similarly unsclerotized areas proximal to the sclerotized ring. Hence, lamponid anterior lateral spinnerets are actually "unisegmented."

Discovery of this intermediate condition suggested that the Lamponidae, as here relimited, might constitute the sister group of the Gnaphosidae plus Prodidomidae. To test that hypothesis, a data matrix has been compiled that includes 30 taxa: the 22 genera assigned below to the Lamponidae, plus eight outgroup taxa. Two genera each represent the families Gnaphosidae (Gnaphosa Latreille and Zelotes Gistel) and Prodidomidae (Prodidomus Hentz and Lygromma Simon). The putatively more plesiomorphic gnaphosoid families are represented by two genera of the Ammoxenidae (Ammoxenus Simon and Rastellus Platnick and Griffin). Because some of the taxa here placed in the Lamponidae have been considered members of the Corinnidae, two genera of that family are used to root the tree and test lamponid monophyly: the type genus Corinna C. L. Koch, and the tracheline genus Trachelas L. Koch (chosen because the Australian taxa that have been misplaced as corinnids were originally described as members or close relatives of Trachelas). Both Corinna and Trachelas are currently heterogeneous groups, and comparisons in each case have been made with their type species (and not, for example, with the New World taxa currently misplaced in Trachelas).

An extensive search was made for new characters that might be relevant to resolving the relationships of these taxa, and several have been found. The data matrix (fig. 1) comprises 29 characters (plus an initial all-zero dummy character, no. 0).

Fig. 1. Character matrix for 30 genera; see text for character descriptions.

Spinneret Characters

As indicated above, the anterior lateral spinnerets consist of either one or two articles; in character 15, those taxa with a complete sclerotized ring surrounding the spigots of the anterior lateral spinnerets, and with that sclerotized ring separated from the proximal article by unsclerotized cuticle, are coded as having two articles (state 0). Those taxa in which the sclerotized ring is incomplete (i.e., is itself interrupted by unsclerotized cuticle) are coded as having a single article (state 1). Among the latter group of taxa, the gnaphosids and prodidomids differ from the lamponids in having the sclerotized ring reduced to a tiny crescent of cuticle, but that additional difference was not coded in the matrix, so that other characters would have a maximal chance of disputing the relationships initially suggested by these differences.

When L. Koch (1873) described the genus Centrothele, he was struck (as his choice of name indicates) by the peculiar "center" or posterior median spinnerets of these spiders; Koch even illustrated the three large and peculiar spigots found on each of those spinnerets! Unfortunately, Centrothele has remained an obscure taxon; for Simon (1897a: 186), it was merely a "genus invisum," and Simon overlooked the peculiar spigots in his own specimens of the genus, which he therefore redescribed under a different name. Koch's character is actually quite distinctive, consisting of a longitudinal row of three cylindrical gland spigots, each of which has an enormously widened base and greatly shortened shaft (figs. 410, 411). These striking spigots (character 16) are found in females of 11 of the genera included in the matrix.

Three characters in the matrix refer to modifications of the piriform gland spigots found in various gnaphosoid families. In most spiders, the piriform gland spigots are quite small, and are smaller than the major ampullate gland spigots that are also present on the anterior lateral spinnerets. However, in some gnaphosoids these spigots can be much larger than the major ampullate gland spigots (state 1 of character 23); in others, the piriform gland spigots can be lost instead (state 2 of character 23). In true gnaphosids, the piriform gland spigots have wide shafts, with broad openings (character 24), whereas in prodidomids, the piriform gland spigots have greatly elongated bases accompanied by plumose setae (character 25).

Also, in the ammoxenids, the posterior median spinnerets have moved anteriorly, so that they separate the anterior lateral spinnerets (character 26), a feature paralleled only in distantly related groups such as the Caponiidae.

Cephalothoracic Characters

Gnaphosoids are often most easily recognized by their modified posterior median eyes. In most spiders those eyes have normal, rounded lenses; seen in profile, for example, the lenses of the posterior median eyes have the same shape as do those of the posterior lateral eyes. In gnaphosoids, however, the posterior median eyes are flattened, and are often irregularly shaped; these modified lenses (character 9) presumably do not function in normal vision.

Another feature often used to diagnoses gnaphosoids is the presence of an oblique depression on the palpal endites (character 10), although the character does occur sporadically in some "clubionoids." Some of the taxa included in the matrix have additional endite modifications as well. Members of the genus Lampona and its close relatives have distinctively shaped endites, with a rather evenly rectangular outline (character 1; fig. 15). Those taxa and others also have a deep and sharply demarcated longitudinal groove along the median edge of the endites, opposite the labium (character 21; fig. 15).

In some of the genera, the carapace is heavily coated with large tubercles (character 20); similar tubercles are often found on the sternum, but their presence may be correlated with the carapace tubercles and they are therefore not coded as an independent character.

At the front of the carapace, there is usually a small, triangular sclerite (the chilum) situated between the anterior edge of the carapace and the proximal margins of the chelicerae. In some cases, the chilum may be fused to the carapace, producing a medially prolonged carapace outline (state 1 of character 12), or the chilum may instead be lost entirely (state 2 of character 12).

Some of the taxa analyzed here have a distinctive cheliceral seta that seems not to have been noted before. The promargin of the cheliceral fang furrow is generally bordered by an entire series of setae that normally parallel each other in size, shape, and orientation. However, the seta which originates closest to the base can be quite modified; in the taxa showing character 13, that seta is greatly elongated, bent at a right angle just beyond its origin, extends across the other promarginal cheliceral setae toward the midline, and is distally plumose (figs. 9--11). The modified seta is found in most of the taxa here considered lamponids, but also in some outgroup taxa as well.

Centrothele and its relatives have an additional cheliceral modification, consisting of an oval area, situated on the anterior surface of the chelicerae, not far dorsal of the promarginal setae. This oval area is relatively weakly sclerotized and much paler in coloration than the remainder of the paturon. This feature (character 17) also occurs in some of outgroup taxa (such as Zelotes), where it also seems not to have been noted before.

Several features of the sternum and the adjacent epimeric sclerites, which are situated between the lateral edge of the carapace and the coxae, may also be informative. In some lamponids, the sternal surface is notably rugose (character 3), rather than smooth. In some, the lateral edges of the sternum are fused to the epimeric sclerites (character 4). The epimeric sclerites can also fuse, producing a single long sclerite that extends above all of the leg coxae (character 8). The latter fusion can happen regardless of whether the epimeric sclerites also fuse to the sternum, but when both characters co-occur, the coxae then each open from a circular foramen (i.e., the coxae are each completely encircled by sclerotized cuticle).

Some lamponids have a distinctive divided scopula, composed of sharply bent hairs (figs. 31, 32), on the ventral surface of metatarsi and tarsi I and II (character 2). All these genera have fewer leg spines than do typical ground spiders, but in some genera, the leg spination is dramatically reduced (character 29); in these cases, there is at most a single proventral spine at the distal tip of tibiae III and IV, with an occasional dorsal spine on femur IV, or spines may be lacking entirely.

Other Abdominal Characters

The pedicel is often highly modified in these spiders. In some genera, the pedicel sclerites form a tube (character 6), which either bears an anterior extension on its ventral surface, or has the entire tube extended anteriorly. In other genera, the anterior edge of the pedicel becomes truncated ventrally (character 19), fitting across the breadth of the posterior margin of the sternum with only a narrow sliver of unsclerotized cuticle separating the pedicel and sternum, or (in extreme cases) even fusing with the posterior margin of the sternum.

Males of these groups commonly have a dorsal abdominal scutum, but in some genera the females may also have a scutum, which may be a dorsal scutum much like that of the male (only smaller; state 1 of character 5), or instead may be an anterior scutum, situated on the anterior face of the abdomen, just above the pedicel (state 2 of character 5).

One of the oddest features of Lampona and its relatives seems not to have been noted before. The ventral surface of the abdomen bears, just behind the epigastric furrow, a pair of oval, invaginated sclerites, one behind each booklung cover (figs. 20, 37, 38; character 11). These invaginated sclerites have a distinctive anterior rim, which is often retracted under the epigastric sclerite itself (at least in preserved specimens). The function of these peculiar sclerites is unknown; scanning electron microscopy has revealed no pores or sensillae on their surface, and there seem to be no anatomical connections to either the respiratory or reproductive organs. In other spiders, similar invaginated sclerites occur in females and are used by males as holdfasts during mating, but in these animals the sclerites are identical in both sexes. They are usually very conspicuous, and although some corinnids and liocranids have flat sclerites, of various shapes, in analogous positions on the abdomen, these invaginated sclerites have not yet been noticed in other taxa. In two new genera that contain only very small species (Pseudolampona and Paralampona), the sclerites are much less conspicuous, and were initially thought to be absent. Compound microscopic examination of the relevant parts of the cuticle indicates that they are present, although they are not heavily sclerotized and are relatively much smaller than in the remaining genera that bear them. Those two genera are therefore coded as having the character, but since its small size might represent either an early stage in their evolution, or a subsequent reduction correlated with a reduced body size, the relative size is not coded as an additional feature.

Genitalic Characters

In many genera considered here, both the male and female genitalia are relatively simple, offering few clues about higher-level relationships. Nevertheless, a few characters can be coded as representing likely synapomorphies. In some cases the epigynal area forms part of an elongate sclerite, with edges parallel to the lateral edges of the booklung covers and separated from the rest of the epigastric sclerite by only a narrow strip of unsclerotized cuticle on each side; in these cases (character 7), the epigynal sclerite reaches all the way to the pedicel.

In these genera, a median apophysis may be present on the male palp, as in most hunting spiders, or reduced (character 14), being represented at most by a slight sclerotization situated within a membranous conductor, or lost entirely. In some cases, the male embolus bears a distinctive apophysis at its base (character 18); in others the embolus may be greatly elongated and twice recurved (character 22). The male tegulum may have a distinctive bulge near its anteromedial corner (character 28), and the female epigynum may have a pair of blind ducts (character 27).

Analysis

Although phylogenetic analyses are still most commonly carried out with equally weighted characters, it seems clear that this is usually inappropriate. If analysis of a data set indicates that some of the characters are homoplasious, then the analysis itself provides evidence that not all the characters deserve equal weight. At least two reasonable methods of analysis which allow the total suite of characters to determine the weights of any individual character are available, and both are used here.

Successive approximations weighting (Farris, 1969) was applied to the character matrix using Hennig86, version 1.5 (Farris, 1988). The states of the few multistate characters were presumed to be independent modifications, and were therefore considered unordered. Four runs using the h, h*, m, and m* tree-generating algorithms, with the results then subjected to the bb* branch swapper, each produced a total of 780 cladograms (length 51, consistency index 0.62, retention index 0.87). Three rounds of successive approximations weighting produced a stable solution (24 cladograms of length 278, consistency index 0.87, retention index 0.96).

One of those 24 resulting cladograms was chosen as the preferred solution (fig. 2), for the following reasons. In 16 out of the 24 cladograms, the new genera Pseudolampona and Paralampona do not appear as sister taxa, falling either as unresolved at the basal lamponid node (six cases) or as paraphyletic (with either one of those two genera treated as closer to other lamponids than to each other, ten cases). While the evidence is not overwhelming, the similarities between those two genera in size and carapace shape (not coded in the matrix), as well as their joint loss of all leg spines, suggest that they do indeed form a monophyletic group.

In two of the remaining eight cladograms, the basal lamponine dichotomy between the new genus Lamponella and the other eight lamponine genera was not recovered, producing a trichotomy at that level. Although most of the many odd features of Lamponella may be autapomorphic to it, the loss of the median apophysis in the remaining genera suggest that they do indeed form a monophyletic group.

The remaining six cladograms differ only with regard to relationships within the centrothelines. All six include two components; they each group Centrothele with the new genera Centrina and Centrocalia, and they each place the two new genera Notsodipus and Longepi as sister taxa. One (fig. 2) shows no further resolution. Three resolve the first group, with Centrothele and Centrina shown as sister taxa, relative to Centrocalia. This result was contrary to my expectations, and seems to be produced by associating the new genus Bigenditia with this component. Bigenditia and Centrocalia are both extremely heavily sclerotized, which is reflected in their both having their sternum fused to the epimeric sclerites (the only coded character in which Centrocalia differs from Centrothele and Centrina). This association allows the program to consider the fusion to evolve in the group of all four genera and then become lost in Centrothele and Centrina. Fusion of the sternum to the epimeric sclerites seems to have occurred independently twice within the lamponines as well (once in the clade including the new genera Lamponicta, Lamponega, and Lamponusa, and a second time in Lamponoides) and a fourth origin, correlated with increased sclerotization, therefore seems more likely than the reversal required by these three cladograms.

Of the remaining three cladograms, one places the Notsodipus plus Longepi clade as the sister group of Prionosternum Dunn, whereas one instead places the Notsodipus plus Longepi clade as the sister group of all other centrothelines. Neither of these contradictory hypotheses seems well-supported, and the preferred solution (fig. 2) therefore shows just the two components, within the centrothelines, that are indisputably supported by the data.

Implied weighting (Goloboff, 1993) was also applied to the matrix, using Pew-Wee, version 2.6 (Goloboff, 1997). This program downweights homoplasious characters using a concave function of the homoplasy; the range of available settings for that concavity function (1--6) was used, in conjunction with the mult*20 command, which randomizes the order of the taxa, creates a weighted cladogram, submits to it tree-bisection and reconnection, and repeats the process 20 times using a different random number seed. At each concavity, the program found the same three cladograms (with, of course, different total weights, of 233.1, 247.0, 255.4, 260.9, 264.6, and 267.6 for concavities 1--6, respectively).

One of those three cladograms is identical to the preferred solution (fig. 2) except that the clade including Pseudolampona plus Paralampona is once again not recovered, producing a trichotomy among those two genera and a group including all other lamponids. The second cladogram also fails to resolve Lamponella as the sister group of the other eight lamponine genera, and the third also resolves a clade including all nine centrotheline genera other than Longepi and Notsodipus. For the reasons cited above, none of these three solutions is seen as an improvement, although the basal dichotomy within the centrothelines resolved in the third cladogram remains a possibility for future testing.

Fig. 2. Preferred cladogram for 30 genera; see text for discussion.

Conclusions

The superfamily Gnaphosoidea is resolved at node 56, supported by flattened posterior median eyes and obliquely depressed endites; the genera Centrothele and Asadipus Simon cluster as gnaphosoids rather than in their current position as tracheline corinnids. The Ammoxenidae cluster at node 30, united by the anteriorly advanced posterior median spinnerets and the loss of piriform gland spigots. The true Gnaphosidae cluster at node 31, based on their widened piriform gland spigot shafts; the Prodidomidae cluster at node 32, based on the elongated bases of their piriform gland spigots; and the gnaphosids and prodidomids cluster at node 33, supported by having the piriform gland spigots larger than the major ampullate gland spigots.

The Lamponidae, as here relimited, cluster at node 54, supported by the pair of oval, invaginated abdominal sclerites and the modified first seta on the cheliceral promargin (despite homoplasy, in the latter character, both inside and outside the group). The lamponids are placed as the sister group of the Gnaphosidae plus Prodidomidae, based on their joint loss of a distal article on the anterior lateral spinnerets (i.e., the interruption of the sclerotized ring around the spigots on those spinnerets). All these higher level relationships will be reexamined, of course, when the substantial Australian faunas of the more plesiomorphic gnaphosoid families (such as the Gallieniellidae and Trochanteriidae) can be added to the matrix. It will be interesting to see whether character 8 (fusion of the epimeric sclerites above the leg coxae) continues to support a separation between the more plesiomorphic gnaphosoid families (such as the Ammoxenidae) and the Lamponidae, Prodidomidae, and Gnaphosidae.

Within the Lamponidae, the new (albeit weakly supported) subfamily Pseudolamponinae clusters at node 39 (supported by the complete loss of leg spines and an uncoded carapace shape character), and appears to represent the sister group of the other two subfamilies recognized, the Lamponinae (node 42) and Centrothelinae (node 51). The latter two subfamilies are united at node 52 by the deep and sharply demarcated longitudinal groove occupying the median side of the endites.

The Lamponinae cluster at node 42, supported by their peculiar endite shape and tubular pedicel sclerites. A rugose sternal surface unites Lampona with the genera Lamponina and Lamponoides at node 36, with the latter two genera clustering at node 35 (supported by their long epigynal sclerite and by the presence of a dorsal abdominal scutum in females, paralleled elsewhere). The new genera Lamponata and Lamponova join that clade at node 40, supported by the presence of divided scopulae on the anterior tarsi and metatarsi.

The three new genera Lamponusa, Lamponega, and Lamponicta cluster at node 38 (sternum fused with the epimeric sclerites, paralleled in Lamponina), with the latter two genera being resolved as sister taxa at node 37, based on their fused chilum. Lamponella appears to represent the sister group of all other lamponines (node 41), retaining a median apophysis on the male palp that is lost in the other genera.

The members of the new subfamily Centrothelinae cluster at node 51, well supported by their highly tuberculate carapace (and sternum), an anterior (rather than dorsal) abdominal scutum in females, an anteriorly truncated ventral pedicel sclerite, a weakly sclerotized spot on the anterior surface of the chelicerae, and, of course, the row of three highly modified cylindrical gland spigots found on the posterior median spinnerets of females. Genitalic characters support the two centrotheline clades recognized (an embolar apophysis at node 45, and blind epigynal ducts and a male tegular bulge at node 34).