Marine Area Relationships from Twenty Sponge Phylogenies. a Comparison of Methods and Coding Strategies

Cladistics 13,1–20 (1997) WWW http://www.apnet.com

Marine Area Relationships from Twenty Sponge Phylogenies. A Comparison of Methods and Coding Strategies

Rob W. M. van Soest1 and Eduardo Hajdu1,2 1Institute for Systematics and Population Biology (Zoölogisch Museum), University of Amsterdam, P.O. Box 94766, 1090 GT Amsterdam, The Netherlands and 2Departamento de Invertebrados, Museu Nacional, Universidade Federal do Rio de Janeiro, Quinta de Boa Vista, 20940–040, Rio de Janeiro, RJ, Brazil and Centro de Biologia Marinha, Cx. Postal 83, 11600–970, São Sebastião, SP, Brazil

Accepted 6 March 1997

Published phylogenies of 20 marine sponge groups are available sponge phylogenies are representative of used to build general area cladograms of marine areas of marine benthic groups, software and hardware limita- endemism under three different methods for which algo- tions are serious obstacles to a successful development rithms adapted for personal computers are available, viz. of marine general area cladograms under any method or COMPONENT, BPA and TAS, and two different coding coding strategy. © 1997 The Willi Hennig Society strategies, Assumption 0 (A0) and “no assumption” (NA). The latter is a recently proposed procedure for handling the distributions of widespread taxa by treating these as separate areas of endemism, rather than as suites of smaller constituent areas. The 20 phylogenies INTRODUCTION contained large numbers of problem data which pre- vented an exhaustive search for all possible equally “best” general area cladograms. The Nelson consensus Marine sponges are organisms fixed on a firm subtrees and their equivalents in parsimony analysis for all strate and possessing a free swimming larva capable of six attempts (viz. three different methodologies under limited propagation (Bergquist, 1978). Sea water tem- two different coding strategies) were compared using perature limits reproductive activity and growth (e.g. their fit with the 20 sponge phylogenies as a measure of Vethaak et al., 1982; Wapstra and van Soest, 1987). quality. Fit was determined using the number of “cospe- Sponge distributions reflect these life history traits. ciations” between a general area cladogram and a taxon area cladogram computed with TreeMap 1.0. No single Barriers between areas of endemism are land masses, method or coding strategy yielded a clearly better fit, oceanic depths and isotherms (van Soest, 1994). Still, each cladogram fitting variously better or worse with var- over geological time, adaptation to almost any marine ious phylogenies. In general, fit with NA coding was habitat has occurred resulting in cosmopolitan distri- higher than with A0 coding, but random tree tests failed butions of higher taxa. At lower taxonomic levels to generate statistically significant support for the con- (genera and species groups) four clearly demarcated clusion that NA coding improves fit. Assuming that patterns are found: cosmopolitan warm water

(tropical/subtropical); restricted Indo-Australian; Arc- Indo-West Pacific), numbers of phylogenies (van Soest, tic-Boreal; and Antarctic-Antiboreal (van Soest, 1994). 1993: four; Hooper and Lévi, 1994: three, Hajdu, 1995: Information on distributions of individual species is seven), and analytical methods (De Weerdt, 1989) and still incomplete and it is estimated that only about half van Soest (1993): Component Compatibility (Zandee the number of species are known to science (Hooper and Roos, 1987); Hooper and Lévi (1994): BPA (Wiley, and Wiedenmayer, 1994). Nevertheless, patterns of 1988). Hajdu (1995) was the first among sponge bioge- species distributions are emerging (van Soest, 1994) ographers to use several alternative methods: and they seem to conform largely to schemes formu- Component Analysis (Nelson and Platnick, 1981; Page, lated by Ekman (1953) and Briggs (1974). At the 1990), BPA, and Panbiogeography (Craw, 1988). In present time, about 35 areas of endemism may be rec- other marine benthic groups attempts at cladistic bio- ognized (van Soest, 1994), defined by the presence of at geography have also been made (e.g. molluscs: Reid, least several, and in many cases large numbers of, 1990). The results of these attempts give conflicting endemic sponge species. There are also many species answers, probably because numbers of phylogenies which have wider distributions covering several of the were inadequate, and no results comparable to those of 35 areas of endemism, for example the Mediterranean– land organisms have so far been obtained (e.g. Crisci et Atlantic distribution which involves the Eastern and al., 1991; Oosterbroek and Arntzen, 1992). However, Western Mediterranean, North-East Atlantic and West whereas marine cladistic biogeographical data in most Africa. Seven of these “wide” distributions have been groups are still scarce and apparently do not show recognized as recurring frequently in distantly related much congruence, there are currently 20 published sponge groups (van Soest, 1994). It is also possible that area cladograms of sponge groups available, together finer patterns of endemism may need to be recognized covering most of the 35 areas of endemism. Although in the future (suggested by Hooper and Lévi, 1994), but this data set is far from perfect—the clades are of so far evidence is lacking. widely different sizes and may cover considerably dif- The large areas of endemism and the nature of the ferent sets of areas of endemism—it prompted us to abiotic barriers indicate that most marine sponges have make the first serious attempt to find general patterns diverged in a process of slow allopatric speciation in of marine area relationships. Because many of the areas the sense of Palumbi (1992). Under this paradigm of of endemism considered are also recognized in other geographical speciation, several attempts have been marine benthic groups, the results presented here may made to analyse marine sponge distributions to find be of wider interest. out whether they can be correlated with present-day distributions of abiotic parameters and with the proposed history of the ocean basins. Methodologies for CLADISTIC BIOGEOGRAPHY these studies varied from comparing lists of species occurring in various areas (Boury-Esnault and Lopes, 1985; van Soest, 1993; Desqueyroux-Faúndez, 1994), The aim of cladistic biogeography is here understood numbers of species of genera occurring in various as the reconstruction of the history of areas of ende- areas (van Soest, 1989, 1993, 1994), and phylogenies of mism (not geographical areas, although these may sponge groups (De Weerdt, 1989; van Soest, 1993; coincide), using phylogenies of organisms as input Hooper and Lévi, 1994; Hajdu, 1995). data (see also Morrone and Crisci, 1995: figure 5). This The latter method, cladistic biogeography (Hum- history of areas of endemism takes the form of a gen- phries and Parenti, 1986), is superior over other eral area cladogram and from this scenarios of analytical methods such as determining degrees of diversification of individual organism groups may be area similarity because it uses phylogenies as building derived in an objective manner. blocks for general area cladogram construction, ensur- Cladistic biogeography methods are still under ing that historical events are not obliterated by development, and several competing approaches have present-day ecogeography. The attempts cited above appeared in the literature. Relatively established, rival were limited in areas (De Weerdt, 1989: only the North methodologies are Component Analysis (Nelson Atlantic considered; Hooper and Lévi, 1994: only the and Platnick, 1981; Humphries and Parenti, 1986;

Page, 1990), and Brooks’ Parsimony Analysis (Brooks, cladograms by employing different coding for the 1981; Wiley, 1988), with its sibling “Component Com- same area in different positions in the cladogram. patibility Analysis” (Zandee and Roos, 1987). A Thus, they potentially solved both the problem of method recently formulated is “Three-Area State- redundant distributions and the single presence of ments” Analysis (TAS) (Nelson and Ladiges, 1991, areas in the general area cladogram. A similar solution 1996). Component Analysis attempts to find general is available for the presence of “wide” distributions in patterns by seeking the congruent parts of individual taxon area cladograms: “no assumption” recoding. taxon area cladograms; BPA and TAS use Wagner Tree parsimony (Farris, 1983) to solve conflicting area relationships. BPA converts all taxon area cladograms into “WIDE” DISTRIBUTIONS AND “NO a combined matrix of areas and taxa, whereas TAS con- ASSUMPTION” RECODING verts the area relationships of all nodes occurring in the taxon area cladogram into suites of three-area relationships, which are subsequently put into a matrix of In line with Brooks’ (1990) proposal, van Soest (1996) areas and “statements” (columns, each of which con- suggested a similar solution for the coding of distribu- tains a three-area relationship). tions of widespread taxa. Different groups have Despite claims of proficiency by its proponents, no different methods of dispersal (range extension with- method has been demonstrated to have a clear advan- out crossing barriers) and are affected differently by tage over the others (Morrone and Carpenter, 1994). the abiotic environment and historical changes in All three methods have pros and cons, and differ in geography. Thus, if general area cladograms are con- their outcome especially when handling conflicting structed from individual cladograms of different data, viz. “wide” distributions, “missing” areas and organism groups, it is the rule rather than the excep- “redundant” distributions. Previous authors devised tion to find smaller and larger, partially or wholly different solutions to deal with these problem data, all overlapping areas of endemism. If the smallest areas involving manipulations of the original data: deleting are used as the areas of endemism, under the prevailing redundancies (Rosen, 1978), “Assumptions 1 and 2” methods, the (partially) overlapping larger areas auto- (Nelson and Platnick, 1981; Platnick, 1981) and “0” matically become “wide” distributions. This appears (Zandee and Roos, 1987), putting question marks in the unfair for taxa with larger distributions, because they matrix (Wiley, 1988), and deleting widespread distri- would not possess proper areas of endemism them- butions (Kluge, 1988). There is cause for questioning selves, only “wide” distributions. If we employ the the legitimacy of these manipulations in the process of operational definition of “area of endemism” (viz. of general area cladogram construction, because in prin- Platnick, 1991; Harold and Mooi, 1994; and Morrone, ciple all distributions contribute to the construction. 1994), as the congruent distributional limits of two or Most cladistic biogeography methods (including more species, many of the “wide” distributions, such Component Analysis, BPA and TAS) optimize their as the marine Indo-West Pacific or the Boreal-Arctic raw data in such a way that the resultant general area distributions, are areas of endemism, with many spe- cladograms present only single relationships for the cies of different organism groups inhabiting the same areas of endemism. This restriction is unwarranted, large area. Manipulating individual parts of these because earth history points rather strongly towards areas under Assumptions 0, 1 and 2 cannot be justified, different area relationships in various geological peri- despite Platnick’s (1991) consideration that relation- ods (cf. Cracraft, 1988; Hallam, 1994). Many (if not ships of the smallest geographical units are the most most) areas of endemism are “composite” (e.g. Duffels informative about earth history. Although applying and De Boer, 1990). If cladistic biogeography aims at Assumption 0 would seem to be reasonable, because it constructing general area cladograms reflecting the represents the empirical data, it is flawed because it historical relationships of areas of endemism, it should inherently assumes that the constituent areas share a allow areas to have multiple relationships. Mickevich taxon that failed to respond to a vicariant event (Nel- (1981) and Brooks (1990) suggested a solution for the son and Ladiges, 1991a, b; Humphries, 1992). This is an presence of redundant distributions in taxon area a priori assumption because the “wide” distributions

may also be the result of dispersal over existing A B A X (A + E) barriers. Van Soest (1996) and Hajdu (1995) proposed a simple E A solution to the problem of distributions of widespread taxa and multiple area relationships which avoids the B B

choice of one of the assumptions (and thus was named D C “no assumption” recoding) by the introduction of a E limited set of extra areas. If several species have a sim- D ilar “wide” distribution, the combination of areas C covered by these is given a name separate from its constituent smaller areas, and is treated in the analysis as FIG. 2. A: General area cladograms resulting from Brooks’ Parsi- a separate area. In constructing the general area clado- mony Analysis under Assumption 0 using imaginary organism groups A–D of Fig. 1; note that area E has been “forced” into sis- gram, a “wide” area and its constituent areas are thus ter-group relationship with area A. B: Two general area cladograms considered independently and may find their place in resulting from Brooks’ Parsimony Analysis under “no assumption” different parts of the cladogram. An example was con- coding of the same imaginary organism groups, with widespread structed by van Soest (1996) in which the recoding distributions over areas A+E recoded as “wide” area X; the position of X differs in two equally parsimonious alternatives; note that area procedure and its potential effect was demonstrated. E now has sister-group relationships with area D and (if it is This example is here reproduced in a simplified form acknowledged that X=A+E) with area A or areas A, B (simplified in Fig. 1: consider taxon area cladograms A–D, each after van Soest, 1996). containing five taxa together occurring in five areas. Two taxa, one in clade C and one in clade D, are widespread over areas A and E. General area cladogram construction under BPA Assumption 0 yields the tree of Fig. 2A; under BPA “no assumption” coding with introduced area X (=“wide” area A+E) it yields the A C A A trees of Fig. 2B. The trees of 2A and B differ in the position of area E, and—if it is acknowledged that area X B A + E also partly contains area E—this area occurs in two

C B positions. Area E has historical relationships with both areas D and A/B. D C This recoding method may be applied in any algorithm used for general area-cladogram construction E D (van Soest, 1996). There are no unwarranted synapo- B D A A morphies when Assumption 0 is applied. Similarly, Assumptions 1 and 2 do not need to be applied because B B there are no ambiguous data caused by overlapping distributions (of course, application of Assumptions 1 C A + E and 2 might still be warranted for missing areas or D D redundant distributions). If the “wide” areas and constituent areas all become E C sister areas in the general area cladogram this is taken as evidence of shared history of all constituent areas; if they end up as para- or polyphyletic, then individual FIG. 1. A–D: Four imaginary clades of five taxa each occurring in constituent areas have had independent histories. In five areas A–E, used for general area cladogram construction under the marine environment, with its high frequency of Brooks’ Parsimony Analysis using two different coding strategies, widespread taxa, the recoding method may be viz. Assumption 0 and “no assumption” (after van Soest, 1996). Note that clades C and D each have a taxon widespread over areas expected to salvage area relationships that might be A+E. obscured by Assumption 0 effects. One might argue

Copyright © 1997 by The Willi Hennig Society All rights of reproduction in any form reserved Marine Area Relationships 5 that the same may be achieved by simply omitting dis- Areas of Endemism and Coding Strategies tributions of widespread taxa, but that is not true: the recoded “wide” areas, by their presence and support, Areas of endemism were derived from van Soest do have an impact on the support for various other (1994), who, after tabulating distributions of between parts of the general area cladogram and thus do influ- 4000 and 5000 sponge species over the world’s oceans, ence the final outcome. arrived at formalized set of 35 (non-overlapping) areas Both the currently accepted coding method and the of endemism. They are given in the map of Fig. 4, in “no assumptions” recoding procedure were applied in which only those areas (31) are numbered that occur in the present cladistic biogeography study and the effect the cladograms used below. Van Soest (1994) also of the recoding procedure on the results is evaluated. reported that in a further approximately 2000 species, seven sets of overlapping areas were found to recur repeatedly. Five of these occur in the cladograms used below and are presented in Fig. 5. Assignment of spe- MATERIALS AND METHODS cies to their area of endemism lacks a formal procedure and thus some ad hoc decisions had to be made. Small Sponge Phylogenies extensions over borders of areas of endemism were accepted for practical reasons. As stated above we implemented two different Twenty cladograms (Table 1, Fig. 3) of sponge genera codings for distributions of widespread taxa occurring or species groups were used. Sources of these cladog- in our data: rams are given in Table 1. Clades are widely different sizes and may cover considerably different sets of areas 1. Assumption 0, in the way it is implemented by of endemism. No individual clade covers all 31 areas: Wiley (1988): species distributions occupying two or the largest set is 26 (Rhabderemia). more areas of endemism (areas 1–31 in Fig. 4) were

TABLE 1 Phylogenies of Sponge Groups Used for General Area Cladogram Construction, Number (N) of Species and Number (N) of Areas under Two Coding Procedures

Taxon N species N areas Source A0 NA

Haliclona oculata group 9 6 4 De Weerdt, 1989 Acervochalina 6 5 4 De Weerdt, 1989 Haliclona aquaeductus group 11 5 5 De Weerdt, 1989 Haliclona fistulosa group 6 4 3 De Weerdt, 1989 Haliclona angulosa group 7 4 4 De Weerdt, 1989 Haliclona arenata group 10 4 4 De Weerdt, 1989 Haliclona rosea group 8 6 6 De Weerdt, 1989 Clathria procera group 9 15 4 Hooper and Lévi, 1994 Tethya spp. 14 14 9 Bergquist and Kelly-Borges, 1991 Didiscus 6 7 5 Hiemstra and van Soest, 1991 Myrmekioderma 5 9 4 van Soest, 1993 Mycale immitis group 17 15 16 Hajdu, 1995 Mycale australis group 10 8 8 Hajdu, 1995 Ceratopsion-Thrinacophora 16 12 11 Hooper and Lévi, 1994 Ptilocaulis-Reniochalina 14 6 6 Hooper and Lévi, 1994 Rhabderemia 26 17 15 van Soest and Hooper, 1993 Pachastrella 9 13 11 Malonado, 1993 Ianthella-Anomoianthella 6 6 5 Bergquist and Kelly-Borges, 1995 Zyzzya c.s. 20 12 11 van Soest et al., 1994 Acarnus 21 17 14 van Soest et al., 1991 A0=Assumption 0 coding procedure, NA=“no assumption” coding.

H. oculata group H. fistulosa group H. arenata group 1+2+3 (32) 10 1+2 (32) 5+6 (33) 10 5 5 10 2+5+6 (33) 5+6 (33) 10 5 2+5+6 (33) 2 5 5 5 10 5+6 (33) 5 5 5 5 5+6 (33) Acervochalina H. angulosa group 5+6 (33) 10 10 H. rosea group 2+5+6 (33) 2+5+6 (33) 10 3 2+5+6 (33) 5+6 (33) 5+6 (33) 5+6 (33) 1+2 (32) 6 5 3 6 5+6 (33) 2 2+5 (33) 2+5+6 (33) 5+6 (33) 5 5+6 (33) 10 30 2+5+6 (33) 19 5 16+18+19 (35) 5+6 (33) 16+17+19 (35) 5 13+14+15+17+18+19+20 (35) 13+14+15+18+19+25+26+27 (35) 1 5+6 (33) 13+14+17+18+19 (35) 6 16+17 (35) 6 22+24+29 (36) H. aquaeductus group C. procera group

Tethya Mycale I Ptilocaulis/ 5+6 (33)+8+21+24 16 Reniochalina 30 24 25 30 24 9 30 24 4+13+16+20+25 (35) 10 13+14+15+16+17+18 (35) 13 27 30 4 10 16 31 27 13+14+15+16+18 (35)+10 24 27 26 31 12 24 10 14 24 5 27 24 2 19 26 11 30 26 28 30 23 14 4+18+30 (35) 12 19+30 (35) 10+11(34) 20 12 19 Didiscus 7 6 4 14+15+16 (35) Mycale II 26 Ceratopsion/ 14+15+16 (35) Thrinacophora 4 25 20 Myrmekioderma 16 10 14+15+16+17+19 (35) 14 4 10+11(34) 25 19 13 14 24 14+15+16 (35) 25+26+27 (36) 26 7 10 11 20 10 30 16 16+20 (35)

Rhabderemia Zyzzya c.s. Acarnus 10 10 26 15 14+15+16+17+18+19+20 (35) 10 12 17 16+18+19 (35) 14 13 14+15+16+18+19 (35) 5+6 (33) 10 14+15+16+18+19 (35) 14 14 13 6 15 20 6 10 8 5+6 (33) 14+15 (35) 15 15+16 (35) 14 5+6 (33) 22 14 14+15+16+18+19 (35) 20 14 10 15 14 11 6+12 (33) 14 11 18+19+20 (35) 20 12 29 16 4+17 (35) 24 6 10+11 (34) 5 10 4+14+15+16+18+19 (35) 25+26+27 (36) 5 6 16 26 9 5 16 5 16+30 (35) 5 14 24 5 9 5+6+7(33)+10+11(34)+ 15+16+17+18+19 (35) 10 14+16+20+21(35)+8+12 16 Pachastrella 6 Ianthella 19 15 18 20 26 16 19 15

FIG. 3. Legend on facing page.

Copyright © 1997 by The Willi Hennig Society All rights of reproduction in any form reserved Marine Area Relationships 7 broken down into their constituent areas and presence PC/ATs (286 and 486). COMPONENT 2.0 (Page, 1993) in each of these areas was indicated by scoring a “1” in was run on the PC/ATs (the program is not available the matrix. Of the taxa in the cladograms, 28% (64 out on Macintosh). Because of structural limits in the soft- of 231) are widespread in one or more areas of ende- ware, the 20 cladograms had to be converted into nine mism. Missing areas were coded as “?”. (despite Page’s manual’s stated limit of 10) by adding 2. ”No assumption” recoding as proposed by van cladograms into larger cladograms (all seven Chalini- Soest (1996): species distributions occupying more dae cladograms; both Mycale immitis group and M. than one area of endemism were assigned to separately australis group; Zyzzya c.s. and Acarnus; Pachastrella recognized “wide” areas of endemism (areas 32–36 in and Tethya/Didiscus and Myrmekioderma; Fig. 5). The constituent areas were scored as “0”, the Ptilocaulis-Reniochalina and Clathria procera group). This “wide” areas as “1” (missing areas as “?”). Table 3 lists was thought to have little impact on the end result as all distributions of widespread species encountered in only a few very general area statements were artifi- the 20 cladograms and their assignment to each of the cially added by this manipulation. COMPONENT 2.0 five “wide” areas. This recoding has a profound effect has a default procedure for “resolving” problem data (Tables 1, 2) on the number of areas (which rises to 36) in the taxon-area cladograms. The program was set to occupied by the sum of all clades, on the number of seek maximally resolved trees using the NNI (Nearest areas occupied by individual clades (Table 1), and on Neighbour Interchange) procedure. The program has a the number of taxa found in areas (Table 2). maximum tree number buffer of 1000 trees after which trees are only added if more resolved than the least resolved tree in memory. The limitations of this proto- Construction of General Area Cladograms col are severe and it is feared that “islands” containing more resolved trees may easily be overlooked. The Cladistic biogeography was performed in the follow- Nelson consensus of all minimal-value trees was used ing way: as the general area cladogram. In the “no assumptions” procedure (NA) areas were coded according to Table 3. 1. Replacement of species names by names of areas BPA under A0 and NA coding procedures was per- of endemism; taxon area cladograms are left as they formed following Wiley (1988). The matrices consisted have been published without further resolution of 428 columns (they are available on request from the (except when using COMPONENT 2.0 which has a senior author). The computer program PAUP 3.1.1 default resolution of problem data). (Swofford, 1993) for Macintosh (set at 10 Mb RAM) was 2. Direct derivation of general area cladograms, used to generate maximally parsimonious trees from using subsequently Component Analysis, BPA and the matrices in the “unordered characters” option, TAS. using the “Random Stepwise Addition” procedure set 3. Recoding of distributions of widespread species at 250 repetitions. The trees found in this way were into one of five recognized “wide” areas (cf. Table 3 for subsequently swapped using TBR until all maximally all recoded taxa distributions); subsequently, deriva- parsimonious trees that could be generated from them tion of general area cladograms using Component were examined, or until the memory buffer of the com- Analysis, BPA and TAS. puter was reached. In the NA analysis, the memory buffer limit of the Macintosh was reached before all Five computers were used, two Macintoshes with trees were fully analysed, but sufficient trees (memory 68040 processors, a Power Macintosh 7100, and two limit was reached while swapping on tree No. 114) were completely swapped through, yielding 18,000 trees. Attempts to bring down the number of trees by subsequent weighting with the rescaled CI failed, due FIG. 3 Area cladograms of 20 sponge clades used as building to extreme slowness of the analysis. Of the saved trees blocks for general area cladograms. Sources for the cladograms are given in Table 1. Mycale I=M. immitis group, Mycale II=M. australis the “50% majority rule consensus plus other combina- group. Numbers on the branches are areas listed in Tables 2 and 3 ble/compatible components” was used as the general and shown in Figs 4 and 5. area cladogram. In the A0 general area cladogram,

three trichotomies occurred, which had to be solved 10,000 trees. Attempts to bring down the number of randomly for the quality evaluation (cf. below). trees by subsequent weighting with the rescaled CI TAS was performed across computer platforms failed, due to extreme slowness of the analysis. Of the using the program “TAS” (Nelson and Ladiges, 1992), saved trees the “50% majority rule consensus plus running on a PC and, following its manual, subse- other combinable/compatible components” was used quently running the output file on a Macintosh using as the general area cladogram. PAUP. The output files are matrices of areas and large numbers of columns with three-area statements (matrices are available on request from the senior author), Evaluation of General Area Cladograms which were treated in the same way as described above for BPA, only the number of repetitions in the The two coding procedures and three methods “Random Stepwise Addition” procedure was set at 100 yielded six general area cladograms, which were all repetitions. Due to the fact that matrices of TAS were considerably different (cf Figs 6, 7). Since the three extremely large, computation times and memory buff- methods are fundamentally different, and no clearly ers were stretched to the limits. In the NA analysis, the “best” method has emerged from previous studies memory buffer limit of the Macintosh was reached (Morrone and Carpenter, 1994), the output trees of the before all trees were fully analysed, but sufficient trees three methods were evaluated quantitatively. This was (memory limit was reached while swapping on tree achieved by using the observed fit of each general No. 55) were completely swapped through, yielding area cladogram with all 20 individual taxon area

2 31 31 5 8 3 6 7 13 17 10 16 17 9 12 11 14 15 23 18 19 21 30 26 25 24 22 20 27 28

FIG. 4. Areas of endemism, numbered 1–31, based on shallow-water sponge distributions (modified after van Soest, 1994: figure 1).

Copyright © 1997 by The Willi Hennig Society All rights of reproduction in any form reserved Marine Area Relationships 9 cladograms as a measure of their quality. Since the in the fit with the sponge cladograms were tested with methodology of constructing general area cladograms the Student’s t-test (Spiegel, 1990). finds its basis in the paradigm of vicariance speciation TreeMap also has an option to generate random trees (coevolution of areas and biota), the degree to which which may be used to compute whether an observed area relationships of a given general area cladogram fit number of cospeciations between a general area clado- the taxon cladograms may be considered a measure of gram and a taxon area cladogram is significantly its efficiency in predicting vicariance events in those higher than a random cladogram with the same number of branches. TreeMap was used to compute taxon cladograms. numbers of cospeciations for each of the six general The computer program TreeMap 1.0 (Page, 1994b, area cladograms with each of all sponge cladograms. It 1995) has an option for the measure of fit between two was also used to compute numbers of cospeciations cladograms, viz. the number of “cospeciations”. Such a shared between each of 1000 random cladograms with cospeciation is explained as evidence of coevolution of the same number of branches as the general area clado- areas of endemism and their endemic taxa. The higher grams and each of the 20 sponge cladograms, to check the number of cospeciations of a given general area whether the general area cladograms performed sig- cladogram and the 20 sponge cladograms (expressed nificantly better than random trees. Finally, TreeMap as a percentage of the number of nodes of each sponge was used to compute numbers of cospeciations which cladogram), the better that general area cladogram pre- 1000 random cladograms with the same number of dicts scenarios of vicariance for these sponge groups. branches as each of 20 sponge cladograms shared with The differences between the general area cladograms each of the six general area cladograms, to check

34 35 35 35

FIG. 5. ”Wide” areas of endemism, numbered 32–36, based on shallow-water sponge distributions (derived from van Soest, 1994: figures 10–16).

TABLE 2 TABLE 3 Areas of Endemism and their Frequency (N) of Occurrence in 20 Recoded Distributions under NA Coding Procedure Sponge Cladograms “Wide” areas Area combinations N taxa N taxa Area recognized in NA encountered in A0 under A0 under NA 32 1+2+3 (1) Arctic 4 1 1+2 (2) Boreal East Atlantic 14 3 (3) Boreal West Atlantic 3 2 33 2+5+6 (4) Japan-China Sea 8 4 5+6 (5) North-East Atlantic 47 22 2+5 (6) Western Mediterranean 35 11 6+12 (7) East Mediterranean 3 2 5+6+12 (8) California 3 3 5+6+7+12 (9) Eastern Pacific 4 3 (10) Caribbean 28 23 34 10+11 (11) Brazil 8 4 (12) West Africa 7 5 35 4+13+16+20+25 (13) Red Sea 10 4 4+14+15+16+18+19 (14) Western Indian Ocean 30 13 4+17 (15) Central Indian Ocean 21 6 4+18+30 (16) Indo-Malaya 30 9 13+14+15+16+17+18 (17) Central Pacific 10 1 13+14+15+16+18 (18) NW Australia (Dampierian) 16 1 13+14+15+17+18+19+20 (19) NE Australia (Solanderian) 22 6 13+14+15+18+19+25+26+27 (20) South Africa 12 7 13+14+17+18+19 (21) South-West Africa 2 2 14+15 (22) SW Atlantic 2 1 14+15+16 (23) Easter Island 1 1 14+15+16+17+19 (24) New Zealand 12 11 14+15+16+17+18+19+20 (25) SW Australia (Flindersian) 7 3 14+15+16+18+19 (26) SE Australia (Peronian) 11 8 14+16+20+21 (27) Tasmania (Maugean) 7 4 15+16 (28) Magellan 1 1 15+16+17+18+19 (29) Antarctic 2 1 16+17 (30) New Caledonia 11 8 16+17+19 (31) Boreal Pacific 2 2 16+18+19 (32) Boreal-Arctic —3 16+19 (33) Mediterranean-Atlantic — 26 16+30 (34) Cental West Atlantic — 5 18+19+20 (35) Indo-West Pacific — 28 19+30 (36) Antiboreal-Antarctic — 3 36 22+24+29 Areas 32–36 were only used in the NA procedure. A0=Assumption 0 coding procedure, NA=“no assumption” coding. 25+26+27 All widespread distributions encountered in the sponge cladograms are listed in the right-hand column and the “wide” areas of endemism to which they are assigned are listed in the left-hand column. Area numbers refer to the areas listed in Table 2. A0=assumption 0 coding procedure, NA=“no assumption” coding.

whether the fit between taxon area cladograms and the general area cladograms was significantly better than pure random match. The use of TreeMap was justified despite the fact that its algorithm is similar to the over those derived by BPA or TAS. In order to compare COMPONENT 2.0 algorithm, because only the num- the output trees a single evaluation method is neces- bers of cospeciations were used, not the other output sary and TreeMap is the only available method to date. items computed by TreeMap. The number of cospecia- Seven of the taxon cladograms (Acervochalina, tions is an objective measure of fit, not favouring Haliclona aquaeductus group, Haliclona angulosa group, COMPONENT 2.0 derived general area cladograms Haliclona arenata group, Rhabderemia, Ianthella, and

Copyright © 1997 by The Willi Hennig Society All rights of reproduction in any form reserved Marine Area Relationships 11 1 Arctic 2 Boreal E-Atlantic 4 Japan-China 12 W-Africa 17 C-Pacific 7 E-Mediterranean 21 SW-Africa Brazil 11 6 W-Mediterranean 8 California 13 Red Sea 20 S-Africa 18 NW-Australia 19 NE-Australia Ocean 14 W-Indian 15 C-Indian Ocean 16 Indo-Malayan 10 Caribbean 31 Boreal Pacific 22 SW-Atlantic 5 NE-Atlantic 24 New Zealand 25 SW-Australia 26 SE-Australia 27 Maugean 29 Antarctic 9 E-Pacific 30 New Caledonia 23 Easter Island 28 Magellan 3 Boreal W-Atlantic 53 28 t Analysis 42 100 100 100 100 20 100 100 C 28 100 100 64 100 100 “50% majority rule 68 100 100 100 100 100 84 100 91 91 91 TAS A0 TAS 41 67 1 Arctic 30 New Caledonia 2 Boreal E-Atlantic 5 NE-Atlantic 6 W-Mediterranean 8 California 21 SW-Africa 12 W-Africa 22 SW-Atlantic 7 E-Mediterranean 24 New Zealand 29 Antarctic 31 Boreal Pacific 4 Japan-China 17 C-Pacific Ocean 14 W-Indian 15 C-Indian Ocean 16 Indo-Malayan 19 NE-Australia 18 NW-Australia 13 Red Sea 20 S-Africa 10 Caribbean Brazil 11 28 Magellan 23 Easter Island 25 SW-Australia 26 SE-Australia 27 Maugean 9 E-Pacific 2 Boreal W-Atlantic 50 60 85 100 100 100 50 100 100 B 100 100 54 100 54 100 60 69 100 100 100 50 85 100 100 100 100 100 BPA A0 BPA 100 100 1 Arctic 5 NE-Atlantic 6 W-Mediterranean 10 Caribbean 4 Japan-China 9 E-Pacific 3-Boreal W-Atlantic 13 Red Sea 12 W-Africa 7 E-Mediterranean Brazil 11 28 Magellan 23 Easter Island Ocean 14 W-Indian 15 C-Indian Ocean 16 Indo-Malayan 18 NW-Australia 19 NE-Australia 17 C-Pacific 25 SW-Australia 26 SE-Australia 27 Maugean 20 S-Africa 30 New Caledonia 8 California 21 SW-Africa 22 SW-Atlantic 24 New Zealand 29 Antarctic 31 Boreal Pacific 2 Boreal E-Atlantic A General area cladograms generated underAssumption (A0). 0A=Nelson consensusoutput tree ofCOMPONENT 2.0, B= COMPONENT A0 FIG. 6. FIG. consensus plus other combinable/compatible components” output of Brooks’ Parsimony Analysis (BPA), C=do. of Three-Area Statemen (TAS).

Zyzzya c.s.) and one of the general area cladograms 1. three disparate areas 1, 3 and 30; (BPA A0) contained one or more polytomies. Since 2. 11 subtropical-boreal-antiboreal areas, in which TreeMap accepts only fully resolved cladograms, the the positions of areas 8, 21, 12, 22, 7 are added stepwise; trees were randomly resolved for the evaluation proce- 3. nine Australian/Indo-West Pacific areas; dure. Differently resolved trees were compared 4. four South American areas; regarding their number of cospeciations, but in general 5. three South Australian areas. these were identical, so only a single resolved cladog- The general area cladogram has a fit with the sponge ram was used for the evaluation. cladograms of 12.5–55%, average 31.25%. It performs significantly (P<0.005) better than 1000 random trees with the same number of branches in 14 of the 20 RESULTS cospeciation comparisons (Table 4). The reverse test showed that of 20 sponge cladograms, only two (Acervochalina and Mycale australis group) had a signif- General Area Cladograms Generated Under A0 icantly higher number of cospeciations than 1000 Coding random cladograms with equal numbers of branches (Table 4). The COMPONENT 2.0 general area cladogram The TAS general area cladogram (Fig. 6c) (length (Fig. 6A) (minimal value 1279, Nelson consensus of 43107 steps, CI 0.665, RI 0.496, “50% majority rule con- 1000 trees containing 90 clusters) has nine single areas sensus plus other combinable/compatible components” of 4592 trees) shows three single areas 9 which are added stepwise to the main cladogram and (Eastern Pacific), 30 (New Caledonia), and 22 (SW three area clusters: Atlantic) and five clusters:

1. nine areas in Australian/Indo-West Pacific 1. three arctic areas; waters; 2. seven disparate areas; 2. three areas in the southern half of South America; 3. 10 areas likewise disparate; 3. 11 areas spread over the northern hemisphere; the 4. five Southern Ocean areas, but also added to this positions of the areas 4, 9, 3, 13, 12 and 7 are added a single area, 5 (NE Atlantic); stepwise to a cluster of five North Atlantic areas. 5. two southern South America areas.

The general area cladogram has a fit with the sponge The general area cladogram has a fit with the sponge cladograms of 12.5–66%, average 29.9%. It performs cladograms of 0–55%, average 23.8%. It performs sig- significantly (P<0.005) better than 1000 random trees nificantly (P<0.005) better than 1000 random trees with with the same number of branches in 15 of the 20 the same number of branches in eight of the 20 cospe- cospeciation comparisons (Table 4). This is the highest ciation comparisons (Table 4). The reverse test showed fit of all general area cladograms. However, the reverse that of 20 sponge cladograms, three (Didiscus, Myrme- kioderma and Mycale australis group) had a significantly test is less successful: of the 20 sponge cladograms, higher number of cospeciations than 1000 random only two (Acervochalina and Mycale australis group) had cladograms with equal numbers of branches (Table 4). a significantly higher number of cospeciations than The differences between the three general area clado- 1000 random cladograms with equal numbers of grams in the fit with the sponge cladograms were branches (Table 4). tested with Student’s t-test and found to be significant The BPA general area cladogram (Fig. 6B) (length 629 between the COMPONENT and TAS results (t=3.46 at steps, CI 0.672, RI 0.609, “50% majority rule consensus 19 df), and between BPA and TAS (t=3.24), but not plus other combinable/compatible components” of between COMPONENT and BPA (t=1.39). Most of the 1170 trees) shows a single area 9 (East Pacific) and five cospeciation percentages, however, were not signifi- clusters: cant in the random taxon area cladogram test (Table 4).

Copyright © 1997 by The Willi Hennig Society All rights of reproduction in any form reserved Marine Area Relationships 13 33 Arctic-Boreal 1 Arctic 4 Japan-China 12 W-Africa 6 W-Mediterranean 34 CW-Atlantic 27 Maugean 7 E-Mediterranean 17 C-Pacific 35 Indo-W-Pacific 22 SW-Atlantic 25 SW-Australia 6 California 13 Red Sea 15 C-Indian Ocean Ocean 14 W-Indian Atlantic 33 Med.- 10 Caribbean 31 Boreal Pacific 20 S-Africa 21 SW-Africa 30 New Caledonia 19 NE-Australia 5 NE-Atlantic 24 New Zealand 26 SE-Australia 16 Indo-Malayan 36 Antarctic-Antiboreal 29 Antarctic 9 E-Pacific 2 Boreal E-Atlantic 3 Boreal W-Atlantic 18 NW-Australia Brazil 11 28 Magellan 23 Easter Island 27 97 50 26 66 57 64 f Three-Area f 22 29 92 43 28 53 75 97 C 53 62 73 65 61 39 60 91 53 69 42 64 65 67 56 53 45 31 TAS NA TAS 28 2 Boreal E-Atlantic 1 Arctic 5 NE-Atlantic 33 Med.-Atlantic 6 W-Mediterranean 4 Japan-China 8 California 13 Red Sea 20 S-Africa Ocean 14 W-Indian 15 C-Indian Ocean 35 Indo-W-Pacific 19 NE-Australia 16 Indo-Malayan 17 C-Pacific 18 NW-Australia 21 SW-Africa 22 SW-Atlantic 10 Caribbean Brazil 11 28 Magellan 23 Easter Island 27 Maugean 12 W-Africa 7 E-Mediterranean 3 Boreal W-Atlantic 32 Arctic-Boreal 34 CW-Atlantic 9 E-Pacific 25 SW-Australia 26 SE-Australia 24 New Zealand 36 Antarctic-Antiboreal 29 Antarctic 31 Boreal Pacific 30 New Caledonia 97 44 93 77 94 27 70 80 95 67 84 88 73 68 29 50 36 78 71 72 B 31 72 35 32 85 85 36 54 64 38 74 72 BPA NA BPA 71 81 “no assumption” coding procedure (NA). A=Nelson consensus tree output of COMPONENT 2.0, 5 NE-Atlantic 1 Arctic 10 Caribbean 13 Red Sea 35 Indo-W-Pacific 36 Antarctic-Antiboreal Ocean 14 W-Indian 16 Indo-Malayan 25 SW-Australia 26 SE-Australia 33 Med.-Atlantic 6 W-Mediterranean 15 C-Indian Ocean 17 C-Pacific 9 E-Pacific 20 S-Africa 4 Japan-China 24 New Zealand 31 Boreal Pacific 2 Boreal E-Atlantic 3 Boreal W-Atlantic 7 E-Mediterranean Brazil 11 28 Magellan 23 Easter Island 12 W-Africa 34 CW-Atlantic 8 California 19 NE-Australia 27 Maugean 18 NW-Australia 21 SW-Africa 22 SW-Atlantic 29 Antarctic 30 New Caledonia 32 Arctic-Boreal A General area cladograms generated under FIG. 7. FIG. B=“50% majority rule consensus plus other combinable/compatible components” output of Brooks’ Parsimony Analysis (BPA), C=do. o Statement Analysis (TAS). Analysis Statement COMPONENT NA

TABLE 4 Quality Evaluation of General Area Cladograms Computed with COMPONENT 2.0 (COMP.), Brooks’ Parsimony Analysis (BPA) and Three-Area Statements Analysis (TAS) under Two Different Coding Procedures, Assumption 0 coding (A0) and “No Assumption” Coding (NA) using Page’s (1995) TreeMap 1.0 Software

Taxa A0 NA COMP. BPA TAS COMP. BPA TAS 121212 121212 Haliclona oculata group 1* 13* 1* 13* 1* 13* 3* 38* 2* 25* 2* 25* Acervochalina1 2* 40* 2* 40* 0* 0* 3* 60* 3* 60* 3* 60* Haliclona aquaeductus group1 2* 20* 2* 20* 1* 10* 3* 30* 3* 30* 3* 30* Haliclona fistulosa group 1* 20* 1* 20* 0* 0* 1* 20* 2* 40* 1* 20* Haliclona angulosa group1 1* 14* 1* 14* 0* 0* 3* 42* 2* 28* 3* 42* Haliclona arenata group1 2* 22* 2* 22* 1* 11* 2* 22* 3* 33* 2* 22* Haliclona rosea group 2* 28* 2* 28* 1* 14* 5* 70* 4* 56* 5* 70* Clathria procera group 1* 11* 1* 11* 1* 11* 2* 22* 2* 22* 2* 22* Tethya spp. 2* 16* 3* 24* 2* 16* 2* 16* 2* 16* 3* 24* Didiscus 2* 40* 2* 40* 2* 40* 3* 60* 2* 40* 2* 40* Myrmekioderma 2* 50* 2* 50* 1* 25* 2* 50* 2* 50* 3* 75* Mycale immitis group 8* 50* 8* 50* 7* 44* 10* 63* 8* 50* 8* 50* Mycale australis group 6* 66* 5* 55* 5* 55* 7* 77* 5* 55* 3* 33* Ceratopsion-Thrinacophora 4* 26* 4* 26* 4* 26* 6* 40* 6* 40* 5* 33* Ptilocaulis-Reniochalina 6* 48* 6* 48* 6* 48* 6* 48* 6* 48* 5* 40* Rhabderemia1 10* 40* 9* 36* 8* 32* 10* 40* 9* 36* 11* 44* Pachastrella 2* 25* 2* 25* 2* 25* 4* 50* 4* 50* 4* 50* Ianthella-Anomoianthella1 1* 20* 1* 20* 1* 20* 3* 60* 3* 60* 2* 40* Zyzzya c.s.1 5* 27* 5* 27* 5* 27* 9* 48* 6* 32* 8* 43* Acarnus 6* 30* 7* 35* 6* 30* 9* 45* 9* 45* 7* 35*

1One or more trichotomies randomly resolved. First it was determined whether the general area cladograms had phylogenetic significance by comparing the number of cospeciations (column 1) shared with the 20 sponge cladograms and those of 1000 random trees with the same number of branches as the general area cladograms. Data are considered significant (*) if the number of cospeciations is less than that of 50 out of 1000 random trees with the same number of branches as the various sponge cladograms. Secondly, it was attempted to determine which of the general area, cladograms may be considered the best by measuring fit with the 20 sponge cladograms deduced from the number of cospeciation events expressed as a percentage of the total number of nodes of the sponge cladograms (column 2). If all nodes coincide with nodes in the general area, cladogram cospeciation is 100%, if no nodes coincide, cospeciation is 0%. Significance (*) determined as above.

If fit is decisive, then the general area cladograms (Mediterranean-Atlantic); also included however, are generated under A0 may be valued as COMPONENT areas 1 (Arctic) and 36 (Antarctic-Antiboreal); =BPA>TAS. 2. three areas forming a trans-Pacific tract; 4 (Japan), 31 (Boreal Pacific) and 24 (New Zealand); 3. two Boreal-Atlantic areas; General Area Cladograms Generated Under NA 4. five disparate areas; Coding 5. two Eastern Australian areas.

The COMPONENT 2.0 general area cladogram The general area cladogram has a fit with the sponge (Fig. 7A) (minimal value 867, Nelson consensus of 1000 cladograms of 20–77%, average 45.2%, which is the trees containing 63 clusters) shows 12 single areas highest score of all general area cladograms. It per- added stepwise to various parts of the tree, and five forms significantly (P<0.005) better than 1000 random clusters: trees with the same number of branches in 11 of the 20 cospeciation comparisons (Table 4). The reverse 1. 12 presumably Tethyan areas, including the test showed that of 20 sponge cladograms, six “wide” areas 35 (Indo-West Pacific) and 33 (Acervochalina, Haliclona angulosa group, Haliclona rosea

Copyright © 1997 by The Willi Hennig Society All rights of reproduction in any form reserved Marine Area Relationships 15 group, Mycale immitis group, Mycale australis group, 1. six disparate areas, including “wide” area 34 and Pachastrella) had a significantly higher number of (Central West Atlantic); cospeciations than 1000 random cladograms with 2. 14 predominantly Tethyan areas, but including equal numbers of branches (Table 4). areas 22 (SW Atlantic), 25 (SW Australia), 8 (Califor- The BPA general area cladogram (Fig. 7B) (length nia), and 31 (Boreal Pacific); 572 steps, CI 0.748, RI 0.648, “50% majority rule consen- 3. five areas in the SE Pacific and the Southern sus plus other combinable/compatible components” Ocean, and in addition area 5 (NE Atlantic); of 18,000 trees which do not represent the complete set 4. two boreal areas; due to memory limits) shows three single areas; 12 5. three southern South American areas. (West Africa), 7 (Eastern Mediterranean) and 30 (New The general area cladogram has a fit with the sponge Caledonia), and six clusters: cladograms of 20–75%, average 40.45%. It performs significantly (P<0.05) better than 1000 random trees 1. five Arctic-North Atlantic-Mediterranean areas, with the same number of branches in seven of the 20 including “wide” area 33 (Mediterranean-Atlantic); cospeciation comparisons (Table 4). The reverse test 2. 12 Indo-West Pacific areas, and in addition a showed that of 20 sponge cladograms, four (Acervoch- single area 22 (SW Atlantic), including “wide” area 35 alina, Haliclona angulosa group, Haliclona rosea group (Indo-West Pacific); and Myrmekioderma) had a significantly higher number 3. four South American areas, and in addition area of cospeciations than 1000 random cladograms with 27 (Maugean); equal numbers of branches (Table 4). 4. two boreal areas, and in addition the “wide” area The differences between the three general area clado- 34 (Central West Atlantic); grams regarding the fit with the sponge cladograms 5. two South Australian areas, and in addition area 9 were tested with Student’s t-test and found to be not (East Pacific); significant (t of COMPONENT/TAS=1.19, of BPA/ 6. three Southern Ocean areas (including “wide” TAS=0.91, of COMPONENT/BPA=0.73). Just as with area 36, Antiboreal-Antarctic), and in addition area 31 A0, most of the cospeciations were not significant in (Boreal Pacific). the random taxon area cladogram test, although they were higher than in A0. If fit is decisive then The general area cladogram has a fit with the sponge the general area cladograms may be valued as cladograms of 16–60%, average 43.12%. It performs COMPONENT=BPA=TAS. significantly (P<0.05) better than 1000 random trees with the same number of branches in five of the 20 Differences Between A0 and NA Results cospeciation comparisons (Table 4). The reverse test showed that of 20 sponge cladograms, four (Haliclona Cospeciations with the 20 sponge cladograms angulosa group, Mycale australis group, Pachastrella and expressed as percentages of the number of nodes of Ianthella) had a significantly higher number of cospeci- these cladograms (Table 4) were consistently higher in ations than 1000 random cladograms with equal NA than in A0, with two exceptions: Tethya under numbers of branches (Table 4). BPA, and Pachastrella under TAS. Differences were sig- The TAS general area cladogram (Fig. 7C) (length nificant under all three methods: Student’s t for 8217 steps, CI 0.734, RI 0.637, “50% majority rule COMPONENT’s A0/NA=5.64 (P<0.005), BPA’s A0/ consensus plus other combinable/compatible compo- NA=3.63 (P<0.01), and TAS’s A0/NA=4.6 (P<0.005). nents” of 10,000 trees which do not represent the In contrast, however, in NA comparisons fewer of the complete set due to memory limits) shows five single observed cospeciations between general area cladog- areas added stepwise to various parts of the tree and rams and the various taxon cladograms were five clusters: significantly higher than 1000 random cladograms

TABLE 5 Resemblance of General Area Cladograms Constructed with Three Different Methods, COMPONENT 2.0 (COMP.), Brooks’ Parsimony Analysis (BPA) and Three-Area Statements Analysis (TAS), under Two Different Coding Procedures, Assumption 0 coding (A0) and “No Assumption” Coding (NA), as Determined by Comparison of the Number of Cospeciations Expressed as a Percentage of the Total Number of Nodes of the Cladograms, using TreeMap 1.0 Software (Page, 1995)

A0 NA COMP. BPA TAS COMP. BPA TAS

COMP. — 57 43 — 57 49 BPA 33 — 40 37 — 49 TAS 27 60 — 40 51 — Read horizontally, the figures refer to the general area cladograms as “host” cladograms on which the other general area cladograms are mapped as “parasites”. Read vertically, the figures refer to the general area cladograms as “parasite” cladograms which are mapped on the other “host” cladograms.

with equal numbers of branches than those observed in TAS: Area 33 (Mediterranean-Atlantic) has sis- A0 comparisons (Table 4). ter-group relationships with Indo-West Pacific areas, In order to detect whether differences in the general while its constituent areas continue to maintain their area cladograms would decrease by adopting the NA different positions in groups of subtropical–temperate coding, all three general area cladograms of both A0 areas also occupied in the A0 general area cladogram. and NA were mapped on each other and their resem- Area 34 (Central West Atlantic) has sister-group rela- blance determined, expressed as the number of tionships with northern hemisphere areas, while its “cospeciations” (Table 5). The results show that the constituent areas have disparate relationships with resemblance of trees in NA is on average higher than in Indo-West Pacific areas (area 10, Caribbean) and A0. However, the differences were not significant southern South American areas (area 11, Brazil). Cold- (t=1.28 with 5 df), so it cannot be concluded that NA water area 32 (Arctic-Boreal) has a single position, whereas 36 (Antarctic-Antiboreal) occupies a position coding converges the differences between the three alongside its constituent areas. methods by removing conflict. Regarding the effects on the positions of “wide” areas and their constituent areas in the two coding pro- Conclusions cedures, the following discrepancies may be noted: COMPONENT: Area 35 (Indo-West Pacific) has sis- Coded in one currently prevailing way (A0), the con- ter-group relationships with the Caribbean, while its struction of general area cladograms under constituent areas are sister to South Australia. Area 36 COMPONENT and BPA yielded considerably differ- (Antarctic-Antiboreal) is related to presumed Tethyan ent results of statistically equal quality; under TAS, areas, while its constituent areas show disparate results are again different and slightly inferior to those groupings in both general area cladograms. Area 33 of the first two methods. Fit between general area (Mediterranean-Atlantic) is related to its constituent cladograms and sponge cladograms ranges between 25 areas. Areas 32 (Arctic-Boreal) and 34 (Central West and 30% of the number of nodes of the sponge cladog- Atlantic) are in single positions far removed from their rams. When distributions of widespread species are constituent areas. recoded according to the “no assumption” procedure BPA: Areas 32 (Arctic-Boreal) and 34 (Central West (NA), the three general area cladograms are again sub- Atlantic) have sister-group relationships away from stantially different, and the quality of all three is their constituent areas which remain in the same posi- statistically equal. Recoding yields a significantly tions as occupied in the A0 general area cladogram. higher fit between general area cladograms and taxon The other “wide” areas occupy positions alongside area cladograms, reaching averages of 40–45%. Recod- their constituent areas. ing also tends to diminish differences between the

Copyright © 1997 by The Willi Hennig Society All rights of reproduction in any form reserved Marine Area Relationships 17 three analytical methods, but this is not statistically sig- whereas Morrone and Carpenter (1994) used nificant. Recoding frequently results in a different COMPONENT 2.0’s option for tree mapping to evalu- position of the “wide” areas away from their constitu- ate their general area cladograms, using the output ent areas, but in several cases they ended up in the data dubbed “items of error” for the comparison. same clade. TreeMap has a similar function and comparable out- In view of the substantial differences between the put, but in addition to delivering “items of error” it trees and the equally low fit with the cladograms from also allows computation of possible “host switching” which they were built, no single general area cladog- (analogous to dispersal in biogeography). This option ram may be objectively indicated as a representation of could be used to deduce a likely scenario of coevolu- area relationships. Intersection (using COMPONENT tion, dispersal and extinction of a given organism 2.0) of the three general area cladograms of both A0 group from a comparison of a general area cladogram and NA yielded hundreds of possible trees and and its taxon area cladogram. (Adam’s) consensus showed very few area relationships common to all three. Programs and Hardware Limitations

Working with a modest number of 20 cladograms, DISCUSSION computing time and internal memory necessary for executing BPA and especially TAS exceeds that of the Quality of the Input Cladograms current generation of PCs. It is clear that real-world general area cladogram building is hampered by these limitations. For a proper and exhaustive data treat- Cladograms used for input were incorporated indis- ment the algorithms need to be implemented on new criminately because of the limited number available in generation computers with faster and more powerful the current literature. However, not all of the 20 clado- microprocessors. COMPONENT 2.0 has an inbuilt grams represented certified monophyletic groups. memory buffer of 1000 trees which prevents an Tethya species were only taken from South-East Pacific exhaustive search, and it is suspected that the tree mor- areas, whereas the genus has a worldwide distribution. phology with a high number of areas added stepwise The species groups of North Atlantic Chalinidae to the tree reflects this non-exhaustive search. BPA and almost certainly have sister species in the North TAS have probably better prospects as methods for Pacific, which were not included; clades of Chalinidae general area cladogram building because they can in other parts of the world oceans were likewise not make use of the powerful options of PAUP and included. Northern and especially southern cold water Hennig86 to reduce the number of maximally parsimo- areas were underrepresented in the 20 sponge cladog- nious trees by adopting additional criteria, e.g. rams. These may have contributed to cause the large (subsequent) weighting of characters, Dollo parsi- numbers of different area relationships in the 20 clades, mony, and filtering for maximally resolved trees. and consequently in the general area cladograms. Measures of fit used to evaluate quality of general area cladograms are also influenced by the quality (extent Coding Procedures and Problem Data of monophyly and completeness) of the various cladograms. Distributions of widespread taxa were treated here in two different ways: firstly by implementing Assumption 0 (Zandee and Roos, 1987), which implies Developing Scenarios From Mapping Cladog- accepting “wide” distributions as empirical evidence rams on General Area Cladograms for sister-group relationships of their constituent areas. It would have been useful also to implement Assump- Page’s (1995) TreeMap 1.0 software is here used to tion 2 (Nelson and Platnick, 1981), but as implemented evaluate the fit of alternative general area cladograms, by COMPONENT 2.0 this involves subsequent

deletion of all but one of the constituent areas of a contain constituent areas that would show fit with a “wide” distribution of a given taxon in a given cladog- compared cladogram if their presence were made ram. With the large numbers of “wide” distributions visible. A formal procedure for salvaging such infor- this would have led to the necessity of introducing an mation would be a useful addition to the NA coding unmanageable number of alternative cladograms, each practice. to be drawn in the analysis with every other alternative cladogram. Likewise, TAS cannot at present deal with Assumption 2 other than by manual analysis, so it is Other Marine Groups not possible to predict here whether Assumption 2 would have performed better (Humphries, 1992). In this study only sponge cladograms were used, but The second coding procedure, here dubbed “no theoretically at least, coevolution of areas and organ- assumption” (Hajdu, 1995; van Soest, 1996) accepts isms extends to many other benthic marine taxon distributions as empirical and handles them invertebrates and even fish. Future studies could accordingly by assigning them to separate “wide” include cladograms of many different groups, such as areas of endemism, thus foregoing the need to imple- corals (Wallace et al., 1991; Pandolfi, 1992), molluscs ment Assumptions 0, 1 or 2. The NA recoding (e.g. Reid, 1990), worms (e.g. Fitzhugh et al., 1994), procedure gives full credit to the original “raw” data fishes (e.g. Bellwood, 1994) and algae (e.g. Lüning and and avoids the need to divide “wide” distributions into Tom Dieck, 1990; Cheshire, 1995). However, care smaller constituent areas (Assumptions 0, 1 and 2). should be exercised when including groups with very Since about 28% of the taxa used in this study are wide- different dispersal capabilities. spread, it is no surprise that A0 coding and NA coding yielded considerably different results. NA coding shows a better fit with the original data, probably ACKNOWLEDGEMENTS because of the multiple positions of geographical loca- tions in the same general area cladogram. Brooks’ (1990) proposal for handling redundant dis- Dr G. Nelson made TAS 1.0 available to one of us (EH), while Dr R. D. tributions by coding them differently (“Caribbean 1”, M. Page made TreeMap 1.0 available though anonymous FTP. Drs F. Schram and J. Parnell allowed us to use their (campus-licensed) copies of “Caribbean 2”, etc. for each different sister-group rela- COMPONENT 2.0 and PAUP 3.1.1 respectively. R. Vonk and M. C. van tionship between the Caribbean and other areas found Soest generously conceded us the use of their computers for periods of in the taxon area cladograms) is theoretically applica- weeks. Dr Lynne Parenti gave material support to RVS through the Soci- ble, but faces practical difficulties, as the number of ety for Integrative and Comparative Biology (formerly the American areas may rise dramatically. For example, in the 20 Society of Zoologists). EH is grateful to the Conselho Nacional de Desen- cladograms used in this study, area 6 (Western Medi- volvimento Científico e Tecnológico (CNPq/Brazil, process no. 201577/ 90.0) and Fundação de Amparo à Pesuisa do Etado de São Paulo terranean) occurred redundantly in seven of the 12 (FAPESP/Brazil, process no. 95/6717-4) for financial support. Three cladograms in which it was represented, and demon- anonymous reviewers as well as the editors contributed significantly to strated 11 different sister-group relationships. Brooks’ the final manuscript. method should likewise formulate suggestions limit- ing the number of possible alternative areas. Both for Brooks’ redundant “alternative” areas, as well as for REFERENCES the constituent and “wide” areas of endemism, the vexing practical problem remains of determining the precise extent of areas of endemism. A formal procedure for such recognition is still wanting, despite Bellwood, D. R. (1994). A phylogenetic study of the Parrotfishes family Scaridae (Pisces: Labroidei), with a revision of genera. proposals and discussions by Morrone (1994) and Records Aust. Mus. Suppl. 20, 1–86. Harold and Mooi (1994). Bergquist, P. R. (1978). “Sponges”. Hutchinson & Co., London. Measurements of fit in the quality comparisons may Bergquist, P. R., and Kelly-Borges M. (1991). An evaluation of the genus Tethya (Porifera: Demospongiae: Hadromerida) with have underestimated those of the NA coded cladog- description of new species from the Southwest Pacific. Beagle 8, rams because implicitly, many of the “wide” areas 37–72.

Bergquist, P. R., and Kelly-Borges, M. (1995). Systematics and bioge- Humphries, C. J. (1992). Cladistic biogeography. In “Cladistics. A ography of the genus Ianthella (Demospongiae: Verongida: Practical Course in Systematics” (P. L. Forey, C. J. Humphries, I. L. Ianthellidae) in the South West Pacific. Beagle 12, 151–176. Kitching, R. W. Scotland, D. J. Siebert, and D. M. Williams, Eds), Boury-Esnault, N., and Lopes, T. (1985). Les Démosponges littorales pp. 137–159. The Systematics Association Publication 10. Oxford de l’Archipel des Açores. Ann. Institut Océanog. 61, 149–225. Science Publications. Clarendon Press, Oxford. Briggs, J. C. (1974). “Marine Zoogeography”. McGraw Hill, New Humphries, C. J., and Parenti, L. R. (1986). “Cladistic Biogeography”. York. Oxford Monographs on Biogeography. Clarendon Press, Oxford. Brooks, D. R. (1981). Hennig’s parasitological method: a proposed Kluge, A. G. (1988). Parsimony in vicariance biogeography: A quan- solution. Syst. Zool. 30, 229–249. titative method and a Greater Antillean example. Syst. Zool. 37, Brooks, D. R. (1990). Parsimony analysis in historical biogeography 315–328. and coevolution: Methodological and theoretical update. Syst. Lüning, K., and Tom Dieck, I. (1990). The distribution and evolution Zool. 39 (1), 14–30. of the Laminariales: North Pacific–Atlantic relationships. In “Evo- Cheshire, A. C., Conran, J. C., and Hallam, N. D. (1995). A cladistic lutionary Biogeography of the Marine Algae of the North analysis of the evolution and biogeography of Durvillea (Phaeo- Atlantic” (D. J. Garbary, and G. R. South, Eds), pp. 187–204. phyta). J. Phycol. 31, 644–655. NATO/ASI Ser. G22. Springer Verlag, Berlin. Maldonado, M. (1993). The taxonomic significance of the Cracraft, J. (1988). Deep history biogeography: Retrieving the historical pattern of evolving continental biotas. Syst. Zool. 37, 221–236. short-shafted mesotriaene reviewed by parsimony analysis: Vali- dation of Pachastrella ovisternata Lendenf. (Demospongiae: Craw, R. C. (1988). Continuing the synthesis between panbiogeogra- Astrophoridae). Bijdragen tot de Dierkunde 63, 129–148. phy, phylogenetic systematics and geology, as illustrated by Mickevich, M. F. (1981). Quantitative phylogenetic biogeography. In empirical studies on the biogeography of New Zealand and the “Advances in Cladistics—Proceedings of the First Meeting of the Chatham islands. Syst. Zool. 37, 291–310. Willi Hennig Society” (V. A. Funk, and D. R. Brooks, Eds), pp. Crisci, J. V., Cigliano, M. M., Morrone, J. J., and Roig-Juñent, S. (1991). 202–222. New York Botanical Garden, New York. A comparative review of cladistic approaches to historical bioge- Morrone, J. J. (1994). On the identification of areas of endemism. Syst. ography of southern South America. Aust. Syst. Bot. 4, 117–126. Biol. 43, 438–441. Desqueyroux-Faundez, R. (1994). Biogeography of Chilean marine Morrone, J. J., and Carpenter, J. M. (1994). In search of a method for sponges (Demospongiae). In “Sponges in Time and Space” (R. W. cladistic biogeography: An empirical comparison of Component M. van Soest, Th. M. G. van Kempen, and J. C. Braekman, Eds), pp. Analysis, Brooks Parsimony Analysis, and Three Area Statements. 183–189. Balkema, Rotterdam. Cladistics 10 (2), 99–154. Duffels, J. P., and De Boer, A. J. (1990). Areas of endemism and com- Morrone, J. J., and Crisci, J. V. (1995). Historical biogeography: Intro- posite areas in East Malesia. In “The Plant Diversity of Malesia” duction to methods. Ann. Rev. Ecol. Syst. 26, 373–401. (P. Baas et al., Eds), pp. 249–272. Kluwer Academic Publishers, Nelson, G., and Ladiges, P. (1991a). Standard assumptions for bioge- Netherlands. ographic analysis. Aust. Syst. Bot. 4, 41–58. Ekman, S. (1953). “Zoogeography of the Sea”. Sidgwick & Jackson, Nelson, G., and Ladiges, P. (1991b). Three-area statements: Standard London. assumptions for biogeographic analysis. Syst. Zool. 40, 470–485. Farris, J. S. (1983). The logical basis of phylogenetic analysis. Adv. Nelson, G., and Ladiges, P. (1992). Manual TAS Three-area state- Cladistics 2, 7–36. ments. A method for historical biogeography. Fitzhugh, K., Giangrande, A., and Simboura, N. (1994). New species Nelson, G., and Ladiges, P. (1996). Paralogy in cladistic biogeogra- of Pseudofabriciola Fitzhugh, 1990 (Polychaeta: Sabellidae: Fab- phy and analysis of paralogy-free subtrees. Am. Mus. Novitates riciinae), from the Mediterranean Sea. Zool. J. Linn. Soc. 110, 219– 3167, 1–58. 241. Nelson, G., and Platnick, N. I. (1981). “Systematics and Biogeogra- Hajdu, E. (1995). “Macroevolutionary Patterns Within the Demos- phy: Cladistics and Vicariance”. Columbia Univ. Press, New York. ponge Order Poecilosclerida. An Integrated Approach to Oosterbroek, P., and Arntzen, P. (1992). Area-cladograms of Biogeography of the Marine Realm, with New Data Derived from circum-Mediterranean taxa in relation to Mediterranean palaeo- the Cosmopolitan Demosponge Genus Mycale.” PhD Thesis, geography. J. Biogeog. 19, 3–20. University of Amsterdam. Centrale Drukkerij, Universiteit van Page, R. D. M. (1990). Component analysis: a valiant failure? Amsterdam. Cladistics 6, 119–136. Hallam, A. (1994). “An Outline of Phanerozoic Biogeography.” Page, R. D. M. (1993). COMPONENT 2.0 for MS Windows 3.0 Oxford Biogeography Series 10. Oxford University Press, Oxford. Manual. Harold, A. S., and Mooi, R. D. (1994). Areas of endemism: Definition Page, R. D. M. (1994a). Maps between trees and cladistic analysis of and recognition criteria. Syst. Biol. 43 (2), 261–266. historical associations among genes, organisms and areas. Syst. Hiemstra, F., and van Soest, R. W. M. (1991). Didiscus verdensis spec. Biol. 43, 58–77. nov. (Porifera: Halichondrida) from the Cape Verde Islands, with Page, R. D. M. (1994b). Parallel phylogenies: Reconstructing the his- a revision and phylogenetic classification of the genus Didiscus. tory of host-parasite assemblages. Cladistics 10 (2), 155–174. Zool. Mededelin Leiden 65 (4), 39–52. Page, R. D. M. (1995). TreeMap 1.0. User’s guide. Hooper, J. N. A., and Lévi, C. (1994). Biogeography of Indo-west Palumbi, S. R. (1992). Marine speciation on a small planet. Trends Pacific sponges: Microcionidae, Raspailiidae, Axinellidae. In Ecol. Evol. 7 (4), 114–118. “Sponges in Time and Space” (R. W. M. van Soest, Th. M. G. van Pandolfi, J. M. (1992). Successive isolation rather than evolutionary Kempen, and J. C. Braekman, Eds), pp. 191–212. Balkema, centres for the origination of Indo-Pacific reef corals. J. Biogeog. 19, Rotterdam. 593–609. Hooper, J. N. A., and Wiedenmayer, F. (1994). Porifera. Zoological Platnick, N. I. (1981). Widespread taxa and biogeographic congru- Catalogue of Australia 12, 1–624. ence. In “Advances in Cladistics” (V. A. Funk, and D. R. Brooks,

Eds), pp. 223–228. The New York Botanical Garden, Bronx, New van Soest, R. W. M., Zea, S., and Kielman, M. (1994). New species of York. Zyzzya, Cornulella, Damiria, and Acheliderma (Porifera: Poeciloscle- Platnick, N. I. (1991). On areas of endemism. Aust. Syst. Bot. 4 rida), with a review of fistular genera of Iophonidae. Bijdragen tot commentary (unnumbered page). de Dierkunde 64 (3), 163–192. Reid, D. G. (1990). Trans-Arctic migration and speciation induced by Spiegel, M. R. (1990). “Statistics”. 3rd edition. Schaum’s Outline climatic change: The biogeography of Littorina (Mollusca: Gastro- Series in Mathematics. McGraw-Hill Inc., New York. poda). Bull. Marine Sci. 47 (1), 35–49. Swofford, D. L. (1993). PAUP. Phylogenetic analysis using parsi- Rosen, B. R. (1988). From fossils to earth history: Applied historical mony. Version 3.1.1. Illinois Natural History Survey. biogeography. In “Analytical Biogeography” (A. A. Myers, and P. S. Giller, Eds), pp. 437–478. Chapman and Hall, London. Vethaak, A. D., Cronie, R. J. A., and van Soest, R. W. M. (1982). Ecol- Rosen, D. E. (1978). Vicariant patterns and historical explanations in ogy and distribution of two sympatric closely related sponge biogeography. Syst. Zool. 27, 159–188. species, Halichondria panicea (Pallas, 1766) and H. bowerbanki van Soest, R. W. M. (1989). The Indonesian sponge fauna: A status Burton, 1930 (Porifera, Demospongiae), with remarks on their report. Netherlands J. Sea Res. 23, 225–232. speciation. Bijdragen tot de Dierkunde 52 (2), 82–102. van Soest, R. W. M. (1993). Affinities of the marine Demosponge Wallace, C. C., Pandolfi, J. M., Young, A., and Wolstenholme, J. fauna of the Cape Verde Islands and tropical West Africa. Courier (1991). Indo-Pacific coral biogeography: A case study from the Forschungs-Institut Senckenberg 159, 205–219. Acropora selago group. Aust. Syst. Bot. 4, 199–210. van Soest, R. W. M. (1994). Demosponge distribution patterns. In Wapstra, M., and van Soest, R. W. M. (1987). Sexual reproduction, “Sponges in Time and Space” (R. W. M. van Soest, Th. M. G. van larval morphology and behaviour in Demosponges from the Kempen, and J. C. Braekman, Eds), pp. 213–223. Balkema, southwest of the Netherlands. In “Taxonomy of Porifera from the Rotterdam. N.E. Atlantic and Mediterranean Sea” (J. Vacelet, and van Soest, R. W. M. (1996). Recoding widespread distributions for N. Boury-Esnault, Eds), pp. 281–307. NATO/ASI Ser. G13. general area cladogram construction. Vie et Milieu 46 (2), 155–161. de Weerdt, W. H. (1989). Phylogeny and vicariance biogeography of van Soest, R. W. M., and Hooper, J. N. A. (1993). Taxonomy, phylogeny and biogeography of the marine sponge genus Rhabderemia North Atlantic Chalinidae (Haplosclerida, Demospongiae). Topsent, 1890 (Demospongiae, Poecilosclerida). Scientia Marina Beaufortia 39, 55–88. 57, 319–351. Wiley, E. O. (1988). Parsimony analysis and vicariance biogeogra- van Soest, R. W. M., Hooper, J. N. A., and Hiemstra, F. (1991). Taxon- phy. Syst. Zool. 37, 271–290. omy, phylogeny and biogeography of the marine sponge genus Zandee, M., and Roos, M. C. (1987). Component compatibility in Acarnus (Porifera: Poecilosclerida). Beaufortia 42, 49–88. historical biogeography. Cladistics 3, 305–332.