Phylogenetic biogeography deconstructed

Authors


Malte Ebach E-mail: male@nhm.ac.uk

Introduction

The confusion of these two classes of ideas is one of the causes that have most retarded the science, and that have prevented it from acquiring exactitude.
A. P. de Candolle (1820), Dictionnaire des Sciences naturelles, p. 383
(Translated by Gary Nelson 1978: 281)

Biogeography is a research programme with one aim, namely to ‘discover why organisms are distributed the way they are today’ (Platnick & Nelson, 1978: 1) by uncovering the ‘patterns of spatial distribution attained by life on earth and the means by which these distributions were achieved’ (Rosen, 1978: 159). The concept of biogeography became manifested in the idea that ‘stations [habitats] are determined uniquely by physical causes actually in operation, and habitations [realms] are probably determined in part by geological causes that no longer exist today’ (de Candolle, 1820: 383–384, 1820). After 150 years of exploration in the latter half of the twentieth century, biogeography has become refined to include a rigorous cladistic approach that uses the relationships of taxa to find common geographical patterns of endemic areas (Rosen, 1978; Nelson & Platnick, 1980, 1981, 1984). Biogeography is about uncovering the biotic and geological history of geographical areas. Recently, however, comparative historical biogeography (the general theory of relationships) has been confused with phylogenetic and ecological biogeography (mechanisms in evolutionary biology), which use distributional phenomena of biota to uncover evolutionary lineages. ‘In short biogeography appears to us to have developed to the extent that it can be reformulated in the terms of the first principles of population ecology and genetics’ (MacArthur & Wilson, 1967: 183).

‘High jacking’ biogeography

Buffon's laws of animal distribution were regarded by Cuvier as veritable discoveries. They set forth some of the fundamental principles of geographic distribution […] Palaeogeography, or the study of the past relations of the land and sea surfaces of the globe, also had its beginnings in Buffon's time.
(Osborn, 1910: 19)

Biogeography was founded as a comparative biological science in the works of Georges-Louis Leclerc Comte de Buffon and Louis Jean Marie Daubenton in the foundation of ‘Buffon's Law’ (migration as causal explanation for biotic distribution) (Buffon, 1761; Daubenton in Buffon, 1749–1767). The works of Buffon, Daubenton and Carl von Linné (Linnaeus, 1781) had influenced the concept of type (individual and biotic diversity) and distribution of Johann Wolfgang von Goethe and the ideas of Alexander von Humbolt, Alphonse P. de Candolle and Phillip Lutley Sclater (de Candolle, 1820; Buffon, 1761; Goethe, 1994; Humboldt, 1816; Humboldt and Bonpland, 1805; Sclater; 1858).

Biogeography is a field that existed long before evolutionary biology and indeed helped in founding the evolutionary ideas of Charles Darwin, Alfred Wallace and Ernst Haeckel. For these evolutionary (Darwinian) biologists, biotic distributions are a result of dispersal away from a centre of origin. To evolutionary biologists, biotic distribution was caused by particular biological processes. They believed the solution to finding why these biotic distributions have occurred was to search for the centres of creation (origin) by retracing the steps of genealogical lineages through space and time. They saw each species lineage as having its own unique history, and extrapolated common biotic distributions to be the result of the same dispersal processes. On an Earth that was believed to be static, retracing species histories (phylogenies) back to the centre of origin was the basis of the new phylogenetic biogeography. Biogeography was thereby ‘high jacked’ by the modern synthesis, namely evolutionary biology and the search for centres of origin (Matthew, 1915; Willis, 1922; Mayr, 1942; Simpson, 1953; Darlington, 1957), most notably in the search for human origins (Haeckel, 1876, 1907, 1909; Leakey et al., 2001). Phylogenetic biogeography had changed biogeography into an evolutionary biological research programme (Darlington, 1957; Huxley, 1900; Mayr, 1942; Willis, 1922; Simpson, 1953).

Darlington's outlook in his study of bird distributions was highly influenced by Darwin:

Nevertheless, I find the distribution of birds very hard to understand. The present pattern is clear enough, though complex. But the processes that have produced the pattern – the evolution and dispersal of birds – are very difficult to trace and understand.
Darlington (1957: 236)

From this statement we see that Darlington had encountered the difficulty of understanding the patterns of biotic distribution since he was relying on finding the evolutionary mechanisms (diachrony), rather than biogeographical relationships (synchrony).

For the late eighteenth century biogeographers, areas, geological in origin, were seen as a cycle of the gradual deposition and erosion of sedimentation on a Lyellian static Earth, which contrasted sharply with the views of the catastrophists who believed that the Earth evolved in an abrupt manner. In this sense biogeography must be answerable to constant dispersal and evolution. Therefore, the Darwinian (evolutionary) paradigm is interested in how species appear in one place and then disperse away from the centre of origin, and how ecological, behavioural, physiological (genetic) attributes affect speciation. The evolutionary approach uses biotic distributions to explain and understand the history of evolutionary lineages. After all, the distributions of biota are a biogeographical phenomenon that have helped explain the evolutionary histories of species. ‘Darwin thought that biogeography was an interesting and, indeed, a critical subject, otherwise he would not have introduced his evolutionary theory to the world through the medium of this distinct discipline’ (Craw et al., 1999: 5). Others have thought the same: the ‘study of insular biogeography has contributed a major part of evolutionary theory and much of its clearest documentation’ (MacArthur & Wilson, 1967: 3).

Sclater (1858) had a prescient notion that by uncovering the relationships of areas we would understand biotic distributions. ‘The world is mapped out into so many portions, according to latitude and longitude, and an attempt is made to give the principal distinguishing characteristics of the Fauna and Flora of each of these divisions; but little or no attention is given to the fact that two or more of these geographical divisions may have much closer relations to each other than to a third’ (Sclater, 1858, 130). Sclater's concern for a classification of biogeographical areas is justified. Comparative biology, that is systematics (cladistics), is a classification scheme for taxa. Biogeography has developed from the ideas of Buffon and Linnaeus as a classification scheme for endemic areas. The search for ancestor–descendant and event-based lineages is not an aim of systematics and biogeography. Lineages do not tell us about classification of taxa and areas, rather they are scenarios, stories of unobserved species history and evolutionary events that are based on a fragmentary fossil and geological record and rely on statistical models.

Discovering proximal (synchronic) area relationships and the classification of endemic areas represent the two main tasks of biogeography. Uncovering area relationships (the geography in biogeography) may tell us something about overall biotic evolution (the bio in biogeography). Biogeography is not about finding individual species histories (ancestor–descendant relationships and event based scenarios). Biogeography is about the proximal relationships of biotic areas of endemism, yet the majority of biogeographers fail to investigate biogeographical problems and misinterpret the discipline.

From a sociological point of view, the aims of any research programme lies in the intentions of those individuals that are considered to be part of that programme, in this case biogeographers. A biologist working on molecular data may ask the same biogeographical question as an ecologist, but the molecular phylogeneticists and the ecologists will use their own ’home-grown’ methods to answer the same problem. Hence, this explains why we have phylogeography (Avise, 2000) and experimental biogeography (Haydon et al., 1993) as two insular and blinkered approaches in an integrated and interdisciplinary field. Will there ever be a ‘whole’ biogeography? As has been so clearly stated before there ‘are no professional biogeographers – no professors of it, no curators of it. It seems to have few traditions. It seems to have few authoritative spokesmen’ (Nelson, 1978: 269). So who are the individuals considered to be the ‘few’ spokesmen and spokeswomen of biogeography?

‘Many of the greatest scientists of their eras were biogeographers, including Charles Darwin, Alfred Russell Wallace […] Joseph Dalton Hooker, George Gaylord Simpson, Ernst Mayr, Robert MacArthur and Edward O. Wilson’ (Brown & Lomolino, 1998: 10–11). The evolutionary biologists above have left a legacy of hypothesizing and modelling for historical evolutionary and ecological mechanisms of species lineages rather than answering the biogeographical problem of overall biotic distribution. They are the protagonists of the modern phylogenetic biogeography– a discipline disinterested in biotic distribution and proximal area relationships, but one that recreates species histories in order to understand evolutionary lineages, originating in one place and then moving and dispersing away from that centre of origin. As the origins of species are unknowable then pursuit of those origins falls into the same trap as looking for ancestors in palaeontology. In fact, Nelson (1983: 469) was moved to say that ‘[F]rom creation myth came the notions of centre of origin and dispersal, which are, and continue to be, employed in biogeography as ‘‘explanatory’’ devices.’

A new theoretical framework, or more of the same?

Recently, in an article in the Journal of Biogeographyvan Veller et al. (2003) attempt to justify a ‘new’ theoretical framework within historical biogeography. They champion the claim that historical biogeography has not one, but two research programmes, namely cladistic and phylogenetic biogeography (see Brooks & McLennan, 2002). The so-called ‘phylogenetic’ biogeography is no more than a repeat of the pre-tectonic ideology that looks for centres of origin and direct lineages. Phylogenetic biogeography can be deconstructed at historical, theoretical and methodological (implementational) levels.

history

Willi Hennig and Lars Brundin were systematists and biogeographers practicing before the tectonic revolution. To them, evolution explains diversity effectively in the sense that biotic distributions can be revealed by way of individual species histories. The original biogeographical question of realms and their interrelationships, as stated by Sclater (1858), were to them irrelevant. Biotic realms are geographical areas defined by the biotas that inhabit them. This pre-tectonic concept in evolutionary biology gave rise to what we now know as phylogenetic biogeography– the use of species histories to solve biotic distributions. The tectonic revolution in geology had effectively removed any doubt as to why there are fossil distributions amongst different biotas. However, the pre-tectonic model of a gradually changing relatively static Earth is still embedded in many evolutionary biological concepts and methods. Now, phylogenetic biogeography (in the sense of Hennig, 1966 and Brundin, 1966) has become irrelevant and should be assigned to the history books. It has neither place in the explanation of biotic distributions nor justification as a viable biogeographical method.

Many of the evolutionary biologists, from Darwin to Hennig, had got it wrong, and we now see that phylogeny does not explain biotic distribution, rather biogeographic patterns may explain phylogeny. In other words cladograms should not be confused with phylogenetic trees. Seen from this perspective, biogeography has the ability to answer questions of biotic distributions without the need for species histories and centres of origin to be sought.

The significance of biogeography as a field in its own right was recognized by Léon Croizat (Croizat, 1958, 1964; Croizat et al., 1974), Gareth Nelson (Nelson, 1970, 1978; Nelson & Platnick, 1980, 1981, 1984; Nelson & Ladiges, 2001), Norman Platnick (Platnick & Nelson, 1978,1984; Platnick, 1981, 1988, 1991), Danielle Rosa (Rosa, 1918) and Donn Rosen (1978). On a dynamic earth, geography is equivalent to geology, which in turn affects the biota (Ebach & Humphries, 2002). Consequently, evolutionary relationships are the result of geography and tell us about the relationships of areas (Ebach & Humphries, 2002; Ebach, 2003a,b). Two significant fields, panbiogeography (Croizat, 1952, 1964; Craw et al., 1999) and cladistic biogeography (Rosen, 1978; see Humphries & Parenti, 1999) embrace this understanding. Cladistics reformed the flawed gradistic (palaeontological) theory of ancestor–descendant lineages in systematics, and cladistic biogeography reformed the equally faulty phylogenetic biogeographical theory in biogeography (see Platnick, 1979).

theory

Hennig's (1950, 1965, 1966) ideas were to revise the phylogenetic concepts of the European idealistic morphologists (Joseph Kälin, Adolf Näf and Danielle Rosa), and to flesh out a method that realized Darwin's (Adolf Remane's and Walter Zimmerman's) desire to classify organisms as genealogies in the English translated and revised version of his 1950 textbook, Phylogenetic Systematics (Hennig, 1966).

Hennig developed a biogeographic method (the chorological method and progression rule), which later influenced the biogeography of Brundin (1966). In the same volume, Hennig also devised the parasitological method, later to be used by Brooks (1981) to develop Brooks Parsimony Analysis (BPA). The realization that biogeography was about areas and not species histories (event-based scenarios) influenced the development of vicariance biogeography by Gareth Nelson, Norman Platnick and Donn Rosen (Nelson, 1969; Platnick & Nelson, 1978; Rosen, 1978; Nelson & Platnick, 1980, 1981, 1984; Humphries & Parenti, 1999), when cladistics was combined with the vicariance concepts of Croizat (Croizat, 1952, 1964; Craw et al., 1999). This later became known as cladistic biogeography (Humphries & Parenti, 1986).

As a point of contention we refute the idea that the method developed to investigate parasite–host relationships had anything to do with the phylogenetic biogeography in the sense of Brundin (1988). The method devised as phylogenetic biogeography by Wiley (1979, 1981, 1987) and Brooks (1981) and revised by van Veller (2000) relates more to Hennig's parasitological method, and as a consequence has tended to muddy the waters when trying to understand how the methods work.

Within cladistic biogeography, taxic relationships on cladograms form proximal relationships. That is, relationships based on shared characters (synapomorphies), rather than on ancestor–descendant relationships. In fact, the cladogram is no more than a statement of proximal relationship that can be easily transformed into a statement of area relationships once the taxa have been replaced by the areas in which they occur. This means that the nodes, which were once diagnosed by synapomorphies, are now comprised of a simple junction called a component. This simple replacement turns the cladogram into an areagram, which may be combined with other areagrams that share the same areas (Fig. 1).

Figure 1.

Cladograms, areagrams and assumptions. (a). Cladogram for three taxa, T1–T3, and two nodes (T4, T5), the areas in which they occur, A–C, and the areagram derived by replacing taxa for areas. (b) The only possible area relationship that can be derived from the cladogram in Fig. 1a. (c) Original areagram in Fig. 1a with five nodes coded into a matrix, coding terminal and internal nodes with 1s, and showing that node 4 (n4) provides the only statement of relationship.

The purpose of using areagrams rather than cladograms for uncovering area relationships is that cladograms cannot be combined. Each cladogram is a unique statement of a taxic relationship. Although area relationships are unique, areagrams are not unique amongst themselves. Therefore, more than two areagrams may share the same area relationship and may be combined to uncover a general area relationship. Areagrams may be congruent, that is, they share the same general area relationship, or they are complimentary, which means that they share relationships that do not conflict. Areagrams may be incongruent, which means they do not share the general relationship but two or more sets of relationships; i.e. they are in conflict. Usually, areagrams have topographical anomalies such that certain taxa occur in two or more areas or several different taxa occur in the same areas. Such areas are termed as either widespread or paralogous (see Nelson & Ladiges, 1996). Therefore, A(B, CD) means that two areas, C and D are represented by one node, as shown by the subtended underlining of CD. Consequently, in terms of maintaining the veracity of the original taxic relations the statement A(B, CD) contains two relationships, A(B, C) and A(B, D). An areagram paralogous, say for areas A and B, can be reduced to statements of relationship that represent the areas just once. Therefore, A(A(A(B(B(B, C), simply shows that B and C are more closely related to each other than either are to A; i.e. A(B, C). In either case, no data has been removed, rather, areas have been represented once, as multiples of areas are nonsensical and do not add or take anything away from statements of relationship.

The so-called phylogenetic biogeography of van Veller et al. (2003) has its origins in the parasitological method of Hennig, which has been shown by Page (1990) to be about general comparisons of hosts and associate relationships in cladograms. Page (1990, 1994) considered that gene trees and species trees, hosts and parasites, and taxa and areas could all be resolved using the same protocols. However, we consider that co-evolutionary methods (gene trees and species trees and hosts and parasites) are fundamentally different from biogeographical ones. In parasites and hosts two cladograms are compared to one another in order to find a co-evolutionary relationship between two symbiotically associated monophyletic groups. Wiley (1981, 1987), Brooks (1981) and colleagues (van Veller et al., 1999, 2000, 2001, 2002, 2003; Lieberman, 2000; Brooks & McLennan, 2002) have used this method to compare taxa and areas in place of host and parasites. In their approach, the taxa and their synapomorphies are not replaced as areas and components, but instead remain as unique phylogenetic comparisons. Crucially, these associations do not discover any biogeographical patterns; as we know from history, phylogeny does not explain biogeography, but rather biogeography may explain phylogeny. Thus, as a biogeographical method, the parasitological method has failed valiantly (see Page, 1990).

method

In their efforts to extract biogeographic signals from area data, Nelson & Platnick (1981) proposed the use of two biogeographical assumptions (which they cryptically called Assumptions 1 and 2) in historical biogeography. These assumptions allow the combination of areagrams (as defined above) and solve the problem of paralogous, widespread and missing data. Zandee & Roos (1987) and Wiley (1987) used another ‘assumption’, one that mistakenly equated areagrams with evolutionary trees. The so-called Assumption 0 (Zandee & Roos, 1987) and BPA, secondary BPA (Brooks et al. (2001) and modified BPA (Lieberman, 2000), treats taxa and areas as host–parasite associations by using the distributions of taxa, nodes and areas (see Fig. 1a,b). Methodologically, Assumption 0 and BPA ignore taxic relationships within the original areagram/evolutionary tree and actually use paralogous, missing and widespread data to obtain solutions. Assumption 0 and BPA code taxa, nodes and areas into a data matrix (Fig. 1c) and then the matrix using a regular parsimony algorithm finds a new areagram onto which the areas are optimized. The result is a cladogram of new relationships that does not reflect upon the relationships within the original areagram. For instance, in Fig. 2a the relationship AC(C(ABC), which contains only one possible relationship, A(C, B) (Fig. 2b), is resolved by Assumption 0 and BPA as B(A, C) (Fig. 2c). van Veller et al. (2003; Fig. 2) incorrectly claim that Assumption 1 also has the same conflicting relationship. However, Assumptions 1 and 2 rely on relationships that are in the areagram and so could never uncover the spurious relationship B(A,C). Brooks et al. (2001) have attempted to show that BPA can recover the original tree from a data matrix. The example they give is an exercise in circular reasoning1. A data matrix will recover any branching diagram (without the original synapomorphies) if that branching diagram is broken down into its fundamental components (taxa, nodes and areas). This is void once any noise occurs in the signal, such as paralogy or widespread taxa, and introduced into the matrix. Instead of uncovering relationships in the original areagram, they produce a completely new cladogram that shares no relationship with the original, as shown in van Veller et al. (2003); fig. 2b).

Figure 2.

Cladograms, areagrams, assumptions and widespread taxa. (a) Cladogram for three taxa, T1–T3, with T1 widespread in areas A and C and T3 in areas A–C, two nodes (T4, T5), the areas in which they occur, A–C, and the areagram derived by replacing taxa for areas. Note that n = node and w = widespread taxa. (b) The only possible area relationship that can be obtained from the areagram in Fig. 2a. (c) Original areagram in Fig. 2a with five nodes coded into a matrix, coding terminal and internal nodes with 1s; showing that node 1 (n1) provides the only statement of relationship, and a new areagram showing that a spurious relationship has been introduced, B(A, C).

Assumptions and misconceptions

The example discussed in van Veller et al. (2003, Fig. 2) is herein reproduced as Fig. 2. The areagram AC((C(ABC)) (Fig. 2a) contains only one possible relationship, namely A(C, B) (Fig. 2b). Assumptions 1, 2 and sub-tree analysis uncover the correct answer; however, Assumption 0 and BPA find B(A, C) (Fig. 2c; van Veller et al. (2003, Fig. 2b), a relationship that cannot exist, as taxon 1 (T1 in Fig. 2a) does not occur in area B. The generation of a new cladogram means that it is not based on any relationships found in the fundamental cladogram as shown in Fig. 2a. Adding in a matrix that represents any branching tree consisting of the original number of terminal nodes is artificially generating a result.

Consider the example in Fig. 3. A taxon-areagram AB(C, D) contains three taxa, T1–T3 found in four areas, one species of which, T1, is widespread in two areas, A and B (Fig. 3a). Assumption 0 has treated the widespread taxa as monophyletic and therefore introduces an extra node. Therefore AB(C, D) is resolved as (A, B)(C, D) (Fig. 3b). Assumption 1 treats the widespread node as artefactual and recovers three possibilities that, combined, form AB(C, D) (Fig. 3c). Assumption 2 also treats the widespread taxa as artefactual and uncovers the one combined areagram from seven potential relationships (Fig. 3d). The relationships uncovered by Assumptions 1 and 2 relate directly to the relationships of areas within the original areagram. Therefore, AB(C,D) contains seven possible relationships for further analysis with different taxa, if we assume them to be either monophyletic or non-monophyletic. Assumption 0 differs, however, as it does not use the relationships in the original areagram. The taxa and areas are recombined in a data matrix (as shown in Fig. 1c) to form a new areagram. Therefore, the new areagram, paradoxically, cannot make any assumptions as it is newly generated and consists only of one relationship, (A, B)(C, D). If Assumption 0 is not an assumption in the sense of Nelson & Platnick (1981) then what is it?

Figure 3.

Assumptions 0, 1 and 2. (a) Cladogram with three taxa (T1–T3) with taxon 1 widespread to areas A and B. (b–d) No of possible cladograms for further analysis using assumptions 0–2. (b) Under assumption 0 only one cladogram is possible as areas A and B in taxon T1 are taken as given. (c, d) Under assumptions 1 and 2 areas A and B are treated as unresolved and hence have a greater number of possible solutions under further analysis; three possible cladograms under Assumptions 1 and 7 (2 × 5; with 3 in common) under Assumption 2.

‘Each column in the final data matrix represents the distribution over a terminal taxon or monophyletic group’ (Zandee & Roos, 1987: 309), that is, the distribution of taxa in a monophyletic group. In this sense Assumption 0 and BPA use nodes in areagrams to represent ancestral taxa in the same way as character states are used as synapomorphies in cladograms. Another similar method also uses taxa as synapomorphies. ‘In PAE (Parsimony Analysis of Endemicity), parsimony analysis is applied to the ‘synapomorphic’ (i.e. shared) taxa of different sample localities in order to obtain relationships between the biotas as sampled at these localities. PAE, therefore, produces area cladograms of sample localities directly from geographical distributions (Rosen, 1984). Cladistic biogeography uses taxonomic characters to fingerprint areas, and PAE uses whole taxa’ (Rosen, 1978: 457). PAE differs from Assumption 0 and BPA in that it uses the distributions of sampled taxa that may or may not be monophyletic. The theory and implementation, however, is exactly the same, with the exception that BPA and Assumption 0 attempt to insert a matrix that represents a fully resolved tree.

Deriving cladograms from the PAE (Assumption 0 and BPA) matrix disassociates the cladogram from the original tree. Therefore, the statement of relationships in the original tree is independent of the generated cladogram. Take the areagram ABCD(ABC(AB, A)) (Fig. 4a). There is only one relationship contained within the areagram, namely D(C(B,A)) (Fig. 4b). All methods find this relationship. However, only Assumption 0 and BPA find exactly the same relationship in unresolved nodes that, indeed, contain any other relationship or form a completely unresolved bush (Fig. 4c–e). The matrix, independent of the original relationships in the diagram will still find a branching diagram consisting of four branches and five nodes despite there being only one informative node, namely D(A, B, C).

Figure 4.

Widespread areas, matrices and assumptions. (a) Cladogram and areagram for four taxa (T1–T4), with T1 present in areas A–D, T2 in areas A–C, T3 in areas AB and T4 in A. (b) Resolved areagram using Assumptions 1, 2 & 0 and BPA. (c) resolved areagram using Assumption 0 and BPA based on three widespread nodes (w), as coded into a matrix with one informative node and one uninformative (basal) node. (d) resolved areagram using Assumption 0 and BPA based on two widespread nodes (w), but with a totally different topology. (e) resolved areagram using Assumption 0 and BPA based on two widespread nodes (w) from a totally unresolved cladogram.

The criticism of Assumptions 1 and 2 given by Zandee & Roos (1987) is that ‘both assumptions [1 and 2] permitted a researcher to conclude that inconsistent monophyletic groups were actually paraphyletic or polyphyletic if that helped strengthen the general pattern of vicariance. They concluded that component analysis gave less than parsimonious results and that simplifying assumptions represented unwarranted ad hoc mechanisms that protected the hypothesis of a single general pattern of vicariance from falsification. Once again the spectre of fiddling with the data to achieve a predetermined result was raised, this time from within the systematics community’ (Brooks & McLennan, 2002: 176–177: original italics).

Areagrams are based on the relationships of taxa within a cladogram. These relationships are our primary evidence and the biogeographic data secondary. Therefore, ‘a cladogram that includes no taxa inhabiting a given area, can provide no information about the relationships of that area, and hence cannot possibly conflict with information supplied about that area by a second cladogram’ (Platnick, 1988: 311). Monophyletic groups (known as clades) are based on shared homologues (synapomorphies) in cladograms. Monophyletic groups are assumed, based on synapomorphies. Therefore, monophyly and non-monophyly are two hypotheses of a taxic relationship. Therefore, the cladogram, A(B(C,D), may be monophyletic or non-monophyletic depending on the synapomorphies or taxonomy. At no time will the relationship between the taxa change: C and D will always be more closely related to each other than they are either to A or B.

Biogeographical evidence (areas of endemism) acts as secondary evidence whether or not it contradicts the hypothesis of taxic relationships. If the biogeographical data are ambiguous, say widespread, then we can alter the hypothesis of the area relationships without affecting the taxic relationships. In the areagram AB(C, D) (Fig. 2a), for instance, there are two possible relationships, namely A(C, D) and B(C, D). These relationships will always persist whether we treat the relationships as monophyletic or non-monophyletic. Regardless of the hypothesis of relationships of taxa and areas, the hypothesis will not change the proximal topographical (synchronic) relationship.

If one were to incorrectly consider cladograms to be phylogenetic trees, then the hypothesis of taxic relationships becomes integral to the tree. So why do Brooks, McLennan, van Veller and Zandee confuse cladograms and trees? The answer it appears lies in the origins of Assumption 0 and BPA.

Trees and cladograms

If phylogenetic systematics is to provide the comparative framework for studies of evolutionary processes, then we must be able to equate at least some cladograms with trees
(Wiley, 1987: 234).

Phylogenetic trees are hypotheses ‘of genealogical relationships among a group of taxa with specific connotations of ancestry and an implied time axis’ (Kitching et al., 1998: 213), and a cladogram is a branching diagram that ‘includes no connotation of ancestry and has no implied time axis’ (Kitching et al., 1998: 202). However, ‘[u]p to this point we have equated ‘branching diagram’ with ‘phylogenetic tree’ because we have been dealing with the unique historical relationships among species based on their traits’ (Brooks & McLennan, 2002: 179). Although they correctly claim ‘cladogram’ literally means ‘branching diagram’ (Brooks & McLennan, 2002: 179), the ‘branching diagrams’ that Brooks & McLennan (2002) use throughout their book show ‘ancestral species’ or the ‘origin’ of characters (Brooks & McLennan, 2002, Figs 2.5, 2.8, 2.9a) and, thus, are clearly phylogenetic trees. The confounding of cladograms and phylogenetic trees dates back to Wiley (1979, 1981). Although Wiley correctly defines cladogram and phylogenetic trees (Wiley, 1981: 97) he then ruins his argument by stating, ‘I have attempted to show that there is a one-to-one relationship among all dichotomous cladograms and trees with identical topologies. Of course, this requires that the entities comprising both have specific evolutionary connotations’ (Wiley, 1981: 98). Platnick addresses this statement, ‘[w]hether by ‘Genealogy’ and ‘Tree’ Wiley means a cladogram, his conclusion is false, for the fact that a character can parsimoniously be considered a synapomorphy of a group does not mean that in actual evolutionary history it did not instead arise in parallel in two or more subgroups’Platnick (1988: 309). One fact is clear: ‘[t]he distinction between cladograms and trees is important because many people have taken the cladogram to be a statement about evolution’ (Kitching et al., 1998: 17).

Despite Wiley (1981: 106) stating that the answer to whether we can find ancestral species (or evolutionary ‘event-o-grams’ see Wiley, 1979: 217) is ‘no’, he still advocates the search for ancestors if one were to use morphology, stratigraphy and biogeography (in fact, some of the auxiliary criteria of Hennig, 1966). Ironically, Wiley agrees with Hull's (1980) incorrect assertion that ‘the difference between testing ancestor–descendant hypotheses and sister-group hypotheses is a matter of the quality of the data needed and not a matter of any inherent difference between the two types of relationships’ (Wiley, 1981: 107). In this way Wiley conflates the argument by stating that although cladograms are not phylogenetic trees, one still can optimize ancestor–descendant relationships onto the cladogram (see Brooks & Wiley, 1986). However, ‘[t]he important point is that evolutionary trees are very precise statements of singular history but their precision is gained from criteria other than character distributions’ (Kitching et al., 1998: 17). This striking contradiction is no different than creating the ‘evolutionary event-o-grams’ that Wiley so strongly abhors! Cladograms are not phylogenetic trees.

The contradictions, seen clearly throughout Wiley's work (Wiley, 1979, 1981, 1985, 1987), are inherent in Assumption 0: ‘this assumption was ‘‘discovered’’ through conversations between myself, M. Zandee and M. Roos (Wiley, 1987: 297)’, and in BPA, ‘Wiley called this application of standard phylogenetic systematic methodology Brooks Parsimony Analysis’ (Brooks & McLennan, 2002: 177; Brooks & Wiley, 1986: 240–250).

Goodbye a priori, a posteriori

Van Veller et al. (van Veller et al., 1999, 2000, 2001, 2002; van Veller, 2000; Brooks et al., 2001; Green et al., 2002;) have pointed out ad nauseam that Assumption 0 and BPA are a posteriori methods. Perversely, using the classification of van Veller (2000) to distinguish a priori and a posteriori methods, Assumption 0 and BPA should be categorized as a priori, rather than a posteriori methods. Such a view is reinforced by the statement of van Veller et al. (2003: 220): ‘modification of data in the taxon-area cladogram by invoking certain assumptions to justify modifying input data in order to provide maximum fit of widespread and sympatric taxa to a single area cladogram’. Assumption 0 and BPA certainly modify the original data by transforming the distribution of taxa and areas on a tree into a matrix and generating a tree with independent statements of relationship, thereby losing the original relationships on the original areagrams. The cladogram that is produced fits the data perfectly (but incorrectly), as the algorithm will find all parsimonious trees and find a consensus tree. This type of modification is an example of precisely the sort of a priori assumption that van Veller et al. (2003) profess to disdain.

Deconstructed

In castigating component and subtree analysis van Veller et al. (2003) have condemned their own methods to the ‘a priori’ category. They still have failed to address numerous critical reviews of their preferred methods, namely Assumption 0 and BPA (Platnick, 1988; Page, 1990; Ebach, 2001), through a constant reiteration that their version of historical biogeography has two quite separate research programmes. We consider that cladistic biogeography fulfils the aims of historical biogeography and stays as the sole contender for understanding and discovering proximal area relationships. What remains of the phylogenetic biogeographical research programme is an anthology of evolutionary biogeography attempting to explain evolutionary histories (species histories, co-speciation histories) instead of attempting to answer biogeographical questions. An accurate description of van Veller et al.’s ontology of phylogenetic ‘biogeography’ may be summed up like this: ‘A species cannot have two different histories... therefore its occurrence in two different areas cannot have been caused by vicariance but must be the result of dispersal’ (van Veller et al., 2003: 324). This statement, namely that all species subjected to vicariance events are required to respond by speciating reflects a misunderstanding of historical biogeography. Nothing in the real world requires widespread taxa to be a result of dispersal (for example, see Page, 1988), nor newborn waifs to be delivered by storks (Platnick, 2003: pers comm.).

Cladistic biogeography remains as an empirical biogeographical field that has progressed in the twenty-first century and encompasses both the geological and biological sciences.

Footnotes

  • 1

    It must be noted that TAS and TASS (Nelson & Ladiges, 1995) both code subtrees as 1s in data matrices but only as informative statements. False nodes by accepting widespread areas as given are not coded. It is worth noting that although subtree analysis has been worked out theoretically and methodologically, it still needs to be a method for full, rather than partial, implementation. BPA and PAE on the other hand are purely implementations, and we consider that the method behind them is PAE analysis and the co-evolution theory (not biogeography).

Biosketches

Malte Ebach's research investigates the history, theory, philosophy and methodology of comparative biology (systematics and biogeography), and trilobite systematics.

Chris Humphries’ research has concentrated on several problems in historical biogeography, both theoretical and empirical, angiosperm systematics (particularly Asteraceae, Nothofagaceae and Malvaceae), biodiversity measurement and methods of priority selection in conservation.

David Williams’ research includes systematics and biogeography of diatoms, and the history, philosophy and theories of comparative biology, particularly cladistics (three-item analysis).

Ancillary