Norntates PUBLISHED by the AMERICAN MUSEUM of NATURAL HISTORY CENTRAL PARK WEST at 79TH STREET, NEW YORK, N.Y
Total Page:16
File Type:pdf, Size:1020Kb
AMERICAN MUSEUM Norntates PUBLISHED BY THE AMERICAN MUSEUM OF NATURAL HISTORY CENTRAL PARK WEST AT 79TH STREET, NEW YORK, N.Y. 10024 Number 3056, 32 pp., 27 figures, 4 tables March 23, 1993 A Reanalysis of Mygalomorpl Spider Families (Araneae) PABLO A. GOLOBOFF1 ABSTRACT The higher classification of the mygalomorph ina should be relimited to include the spiders is reviewed. A data set of 42 taxa and 71 Microstigmatidae and Rastelloidina. The group characters is analyzed, applying a new method for Crassitarsae should include also the Microstig- character weighting. Most families were repre- matidae. The group formed by the Diplurinae and sented by two or more genera, so that their mono- Rastelloidina plus Crassitarsae is called Bipectina. phyly could be tested. The trees that best conform Three of the currently recognized families appear to the data have Atypidae plus Antrodiaetidae, as paraphyletic: Cyrtaucheniidae (in terms of Mecicobothriidae, Hexathelidae, non-diplurine Domiothelina), Nemesiidae (in terms of Micro- "diplurids," and diplurines as successive sister stigmatidae and Theraphosidae plus Paratropidi- groups ofRastelloidina plus Crassitarsae. Neither dae plus Barychelidae), and Dipluridae (in terms Fomicephalae nor Tuberculotae appear as mono- ofCrassitarsae plus Rastelloidina). Support for the phyletic groups. The groups Atypoidea (restricted monophyly of Hexathelidae is weak; the data are to Atypidae plus Antrodietidae) and Avicularioi- explained almost as well by trees in which the dea (the rest of the Mygalomorphae) are resur- family is paraphyletic. rected. The groups Orthopalpae and Quadrithel- INTRODUCTION The higher classification of the spider in- ra incognita. The mygalomorphs had been fraorder Mygalomorphae was recently ana- traditionally divided into atypoids (atypids, lyzed by Raven (1985). Prior to Raven's work, antrodiaetids, and mecicobothriids, families the relationships among the families were ter- with a complex male palpal bulb, traces of I Cornell University/American Museum ofNatural History Graduate Training Program in Arthropod Systematics. Copyright © American Museum of Natural History 1993 ISSN 0003-0082 / Price $3.40 2 AMERICAN MUSEUM NOVITATES NO. 3056 abdominal segmentation, and six spinnerets) simony analysis. Using these new tools, Ra- vs. non-atypoids (the remaining families, with ven (personal commun.), Coddington (per- male bulb sclerites fused, completely unseg- sonal commun.), and I have now reanalyzed mented abdomen, and only four spinnerets). the relationships among the families of My- Platnick (1977) showed that the atypoids galomorphae using the characters mentioned shared only plesiomorphies, and were thus in Raven (1985). Given Raven's data, if ev- likely to be a paraphyletic group. Raven (1978, ery one of the 15 families he considered is 1979, 1980, 1981, 1982), Raven and Platnick treated as a terminal and scored as having (1981), Gertsch and Platnick (1979, 1980), the character states he considered plesiom- and Platnick and Shadab (1976) tried to re- orphic for each family, Raven's familial tree solve some groups, notably the "Dipluroi- (1985: fig. 1) is one among several most par- dea" (mecicobothriids, microstigmatids, di- simonious trees. This, however, does not plurids, and possibly all or some guarantee optimal results: even ifeach ofthe "pycnothelids"). Finally, Raven (1985) di- 8 cladograms presented by Raven is in itself vided the mygalomorphs into two groups, maximally parsimonious, putting all those Fornicephalae and Tuberculotae, both in- cladograms together may result in a less than cluding some atypoids and some non-aty- optimal solution. In fact, some lines of evi- poids. The interested reader is referred to Ra- dence suggest that some of the families con- ven (1985), and references therein, for a more sidered by Raven (1985) might not be mono- historical perspective. phyletic. The best example is the Nemesiidae; Raven's (1985) paper was the first to at- Raven (1985: 28) discussed the possibility tempt an understanding of the relationships that some of the proposed synapomorphies within the Mygalomorphae as a whole. Con- of Nemesiidae might actually be synapo- sequently, it became a starting point for the morphies of the more inclusive group Cras- higher systematics in Mygalomorphae. Cod- sitarsae (secondarily modified in most non- dington and Levi (1991) rightly referred to nemesiid Crassitarsae). This would leave the Raven's results as "impressive." monophyly ofNemesiidae unsupported, and At the time of Raven's study, however, it would be unwarranted to use "Nemesi- computerized parsimony programs were not idae" as a terminal in the analysis. Other cases readily available, and analysis could be done are discussed below (see under Taxa). only by hand. Raven explicitly attempted to The aim of this paper is to present a rean- minimize homoplasy for his data, which had alysis of the relationships of the Mygalo- nothing less than 240 terminals (genera). For morphae. I intend to show that some of the that many taxa, finding the optimal tree(s) is, hypotheses of monophyly proposed by Ra- because of the extremely large number of ven (1985) are indeed weakly supported or cladograms to be considered, a very difficult counterindicated by the data, and that a dif- task. Raven (1985) divided the mygalo- ferent tree is a better hypothesis of mygalo- morphs into 15 families, proposed a familial morph relationships. This paper also consti- cladogram, and provided separate generic tutes the first application of a new method cladograms for most of the families. The for character weighting. The method is being number of terminals in the cladograms var- described in detail elsewhere (Goloboff, in ied from 7 to 26. It appears very likely, then, prep.), but a short discussion can be found that some of the trees that Raven presented in the section on "Methods." were not actually most parsimonious trees; This paper uses data that have been col- in fact, some cladogram nodes are said to be lected over several years, in the course of supported by characters which are ambigu- studies focused on aspects other than the re- ously, or even incorrectly, optimized. lationships of mygalomorph families. I wish Not that Raven was unaware of this prob- to express my gratitude to a number of ar- lem, of course. In recent years, powerful mi- achnologists who have helped me in the crocomputer parsimony programs have been course of such studies: Maria E. Galiano, developed (see Platnick, 1989; Fitzhugh, Norman I. Platnick, Robert J. Raven, Fred- 1989), the use of which has also increased erick A. Coyle, Martin J. Ramirez, Emilio A. general awareness ofthe complexities ofpar- Maury, Charles Griswold, and Jonathan A. 1 993 GOLOBOFF: REANALYSIS OF MYGALOMORPHS 3 Coddington. Specimens kindly lent by Rob- (Goloboff, in prep.). Given two trees A and ert J. Raven (Queensland Museum, Bris- B, choosing B instead of A may save a step bane), Norman I. Platnick (American Mu- for one character and require an additional seum ofNatural History, New York), Emilio extra step in another. Even when the two trees A. Maury (Museo Argentino de Ciencias Na- require the same number ofsteps, ifthe char- turales, Buenos Aires), Herbert W. Levi (Mu- acter in which we could save some steps shows seum ofComparative Zoology, Harvard), Ana more homoplasy on the trees being com- T. Da Costa (Museu Nacional de Rio de Ja- pared, tree A should be kept. If by choosing neiro), and Jonathan A. Coddington (Smith- B instead we would save a step in a character sonian Institution, Washington) were rele- which shows less homoplasy, B would be a vant for the present study. Maria E. Galiano better choice than A. It is possible to take kindly shared her unpublished observations into account automatically the relative of postembryonic development of several amounts ofhomoplasy ofthe characters when mygalomorphs. I also wish to express my counting steps to compare trees. This can be gratitude to a number of nonarachnologists done by searching for trees in which the "total (nobody is perfect!) who have also helped me: fit" (i.e., the sum of the fits for all the char- James M. Carpenter, James S. Farris, James acters) is maximum, where fit for each char- K. Liebherr, Kevin C. Nixon, and Quentin acter is measured with a nonlinear, concave D. Wheeler. My wife Claudia A. Szumik used function of the homoplasy (see Goloboff, in the computer program developed for this pa- prep., for a full discussion). A well-known per to analyze her own data on embiids, and concave function of the homoplasy is the helped eliminate many errors. Norman I. (character) consistency index, c, ofKluge and Platnick, Robert J. Raven, Frederick A. Coyle, Farris (1969). The effect of using that func- and Maria E. Galiano provided helpful crit- tion to measure fit is that the difference in fit icisms ofdifferent versions ofthe manuscript. between two trees will be smaller for char- Financial support from the American Mu- acters with more extra steps. seum ofNatural History and the Department An MS-DOS computer program, Pee-Wee of Entomology of Cornell University are 1.45 (available from the author), was used to greatly appreciated. calculate the fit for each character using the consistency index (multiplied by 10, with one METHODS significant digit), and to search for trees with Abbreviations. Abbreviations used in this higher total fits. Pee-Wee's branch-swapping study are standard for the group: PLS, pos- algorithms are less effective than those ofoth- terior lateral spinnerets; PMS, posterior me- er parsimony programs (such as Hennig86); dian spinnerets; ALS, anterior lateral spin- to increase the chance of finding all or most nerets; STC, superior tarsal claws; ITC, ofthe optimal trees, several runs were made, inferior tarsal claw. using different initial trees. Parsimony analysis. A data set in which For some of the characters, the terminals 42 taxa were scored for 71 characters was may have more than one state. Those steps analyzed. As pointed out by Coddington and "within" terminals were counted to calculate Levi (1991), cladistic analysis is essentially a the fit for the corresponding characters.