Malte C. Ebach's research investigates the theoretical problems inherent in standard cladistic and cladistic biogeographical approaches and systematic revisions of mid-Palaeozoic trilobite groups.
Chris J. Humphries' research has concentrated on several problems in historical biogeography, both theoretical and empirical, angiosperm systematics, biodiversity measurement and methods of priority selection in conservation.
Christopher J. Humphries
Department of Botany, The Natural History Museum, London, UK
Malte C. Ebach's research investigates the theoretical problems inherent in standard cladistic and cladistic biogeographical approaches and systematic revisions of mid-Palaeozoic trilobite groups.
Chris J. Humphries' research has concentrated on several problems in historical biogeography, both theoretical and empirical, angiosperm systematics, biodiversity measurement and methods of priority selection in conservation.
Cladistic biogeography is about discovering geographical congruence. The agreement of several taxon-area cladograms (TACs) rarely yields a perfect result. Areas may overlap, taxa may not be evenly distributed, and thus, ambiguity may be prevalent in the data. Ambiguity is incongruence and may be resolved by reducing paralogy and resolving potential information. Recently, several new approaches in cladistic biogeography [i.e. Brooks parsimony analysis (BPA), Assumption 0] interpret ambiguity as congruence. These methods are problematic, as they are generational. Methods constructed under the generation paradigm are flawed concepts that are immunized from falsifying evidence. A critique of modified BPA reveals that taking an evolutionary stance in biogeography leads to flaws in implementation.
Area cladistics is a new development in cladistic biogeography. Area cladistics adopts paralogy-free subtree analysis using Assumption 2, to discover the relative positions of continents through time.
Geographical congruence is the result of allopatric (geographical) speciation. Vicariance, dispersal and combinations of both, are recognized causes for allopatric speciation. Area cladistics highlights the concept that all these events occur in response to geological changes (e.g. continental drift) either directly, by geographical boundaries, or indirectly, at the level of ocean currents. Samples of chosen examples all respond to the geological process. The examples include Ordovician–Silurian and Lower Devonian trilobites to yield a general areagram which is a representational branching diagram that depicts the relationships of areas.
Finding one common biogeographical pattern from several unrelated groups is a qualitative approach to interpret the positions of continental margins through time. Area cladistics is not a substitute for palaeomaps that are derived from palaeomagnetic data, but general areagrams adding to the body of knowledge that yields more precise interpretations of the earth's past.
The temptation to form premature theories upon insufficient data is the bane of our profession The Valley of Fear (Doyle, 1993, p. 779).
So said Sherlock Holmes, the alluring character created by Arthur Conan Doyle who has enthralled many generations of `who-done-it' fiction readers. Sherlock Holmes' powers of deduction are world-renowned and he often flabbergasted the clients that entered the sitting room of his private residence at 211B Baker Street. Holmes's statement reveals the success achieved by his amazing powers of deduction. The patch of mud on the sleeve, the grime of chalk on the forefinger, and the alternating shaking or steady hand in a written document are all facts that informed his conclusions. Our hero's disciplined approach prevents him from making any premature deductions. Lestrade, Doyle's bumbling policeman, was all too hasty to contrive conclusions based on tenuous connections, that were later diminished by Holmes' dedicated search for the `facts' before formulating his theories.
Various methods and implementations of cladistic biogeography have been dismissed because of tenuous ideas and dubious inferences. For example, historical biogeography has often claimed to find ways of locating ancestral areas or the means of migration or dispersal of one species or group based on a fragmentary fossil record and with only a vague understanding of the organisms themselves.
Forming a more disciplined approach to the subject, Nelson & Platnick (1981) presented a method based on numerous existing ideas (Brundin, 1966; Hennig, 1966; Rosen, 1978, etc.) that organized relationships of areas based on their shared homologues as represented by components. The concept was that the relationship of areas inhabited by endemic taxa might be resolved if combined with the relationships of taxa yielded by cladistic analysis. Ultimately, the method functions by translating the relationships of taxa into the relationships of areas.
Since Nelson and Platnick's publication, more methods have been formulated (Wiley, 1987; Zandee & Roos, 1987; Brooks, 1990; Page, 1993; Nelson & Ladiges, 1996) to deal with translating the relationships of taxa into the relationships of areas. For most of the methods to work ideally, each taxon should normally inhabit one endemic area. However, this is not always the case. Taxon distribution may overlap with other endemic areas and with the distributions of other taxa. The ideal distribution, that is one taxon per area, is the result of allopatry (separation followed by geographical isolation). Geographical congruence is the result of allopatry in two or more groups of taxa. Overlapping distributions are the result of sympatry or parapatry (formed by non-geographical isolation). Ambiguity is the result of sympatry and other non-geographical isolation mechanisms.
The results of different methods in efforts to resolve area ambiguities resolve with different outcomes. Various classifications of methods have been proposed (e.g. Crisci, 2001; Morrone & Crisci, 1995). van Veller et al. (1999) used the terms priori and posteriori in the vain hope of classifying the main cladistic biogeographical methods into one scheme. These terms may also be defined under the discovery and generation paradigms, respectively. Contrary to discovery, methods constructed under the generation paradigm are based on preconceived beliefs, in this case, of evolution and biogeography, rather than on facts. These methods are like our working Lestrade, eager to find one clue to form a conviction no matter how tenuous.
Cladistic biogeographical methodology may avoid forming premature theories by separating hierarchical relationship from evolutionary history. Cladistic biogeography contains a range of methods, many intent on reconstructing evolutionary history and many that diverge from that aim.
Cladistic biogeography aims to extract the signals of geographical congruence. Geographical congruence is best explained by evolution. Therefore, when we find geographical congruence, it supports the theory of evolution.
The procedure to find congruence starts with a taxon-area cladogram (herein TAC), which is a cladogram that has the taxa replaced with the areas in which they occur. The TAC A(BC) can yield no more information other than that B and C are more closely related to each other than either is to A (Fig. 1). Cladograms can also find congruence. A common pattern of relationships among taxa is evidence of homology. Homology is the best explained by evolution. Although the TAC shares the same structure as a cladogram, it tells only of the geographical distribution of taxa. Taxic homologues are replaced by area homologues. A suite of geographical patterns may reveal one common pattern, either as a resolved or unresolved areagram. A resolved general areagram is a result of congruence among the individual TACs and provides insights into the geographical history of several unrelated groups.
The concept of geographical congruence is simple and many hours are spent producing cladograms to justify, in the minds of many, an easy biogeographical interpretation. Not all taxa inhabit one area, thus a single terminal node or branch of a TAC may contain more than one area, or several branches may contain the same node. The pursuit of `quick and easy' solutions is the bane of cladistic biogeography since its inception.
Most methods are primarily concerned with attaining resolution in the general areagram, whether there is congruence among the individual TACs or not. Some areas yield more information than do others, and areas can be neither excluded nor redefined. Cladistic biogeography is concerned with interpreting the patterns of history. An area is not unlike the crime scene of Holmes' nineteenth century London. The victims' fate is a given fact, however, the motive(s), weapon(s) and circumstances may never be known but can only be speculated upon. Cladistic biogeography is a method consisting of a series of tools but which are only effective so far as they deal with and acknowledge incongruence. Incongruence amongst TACs, herein defined as ambiguity, has led to many methods that implement theories prematurely and cannot accept that a general areagram may not be necessarily resolved. As Holmes' might say the case is unsolvable.
By pursuing pre-determined theories of evolution it is easy to stray from the original aim, to discover congruence. The aim is objective, recognizing that geographical congruence may or may not be there. This however, is directly contrary to Lieberman (2000):
Accepting the fact that there is an analytical approach that can be applied to historical studies is part of the willingness to accept that there are actual historical patterns (Lieberman, 2000, p. 134).
It is subjective to search for congruence because one might believe it to be present. Congruence is evidence for evolution in systematics, but we can only make that inference if it is there. If all that can be found is incongruence, there is no evidence for evolution (homology). Implementation may diverge completely from its aim if it asserts that congruence exists regardless of the result.
In this way, methods are formulated subjectively to seek and treat ambiguity as informative. Cladistic biogeography under these implementations no longer seeks to find evidence for evolution, as both congruence and incongruence are taken as proof. The part evolution plays in causing the distributions of organisms is no longer being questioned, but is regarded as part of the answer in the biogeographical analysis.
The developmental history of cladistic biogeography methodology has been short. Preceding its inception Nelson (1981) warned that:
Biogeography has too long been riddled with definitive authority. What the field needs now is a simplification of theory that will open the field to future investigation. What it needs are questions, not answers. Without them, it will remain what perhaps, it still is: a nineteenth-century science that, for a while, has become interesting once again (Nelson, 1981, p. 528).
In light of recent developments, one may wonder if cladistic biogeography has been opened to `further investigation' over the last 20 years.
Implementing the aim
There are many reasons for why taxa are distributed in particular patterns. Two explanations include dispersal and vicariance. However, amongst the different methods that masquerade under the rubric of historical biogeography, island biogeographers, pan-biogeographers and cladistic biogeographers all have developed separate explanations and preferences for which particular mechanisms (either or only dispersal, vicariance or allopatry) are primarily responsible. An ideal distribution might consist of one species in every area. Unusual distributions might include two species sharing one endemic area, or one species spread over two or more areas. Vicariance and dispersal are both subjective explanations for interpreting distribution patterns.
Clearly, then, neither dispersal nor vicariance explanations can be discounted a priori as irrelevant for any particular group of organisms, and it might seem that the ideal method of biogeographical analysis would be one that allows us to choose objectively between these two types of explanations for particular groups (Nelson & Platnick, 1981, p. 47).
To date no method is able to `choose objectively' between either mechanism. The interpretation of dispersal and vicariance events is pure hypothesis rather than observation. Cladistic biogeography cannot be implemented so as to choose between vicariance and dispersal. Vicariance and dispersal both cause the same patterns. TAC are essentially silent or passive statements of relationship. They do not function as active accounts of speciation (hybridization, reticulation) or geography (migration, centres of origin). Geographical and geological evidence are more appropriate tools for uncovering which mechanism is responsible for speciation.
The role of the method in cladistic biogeography is to find evidence that either supports or falsifies a hypothesis. Geographical congruence is one such hypothesis. The hypothesis is investigated using a method such as cladistic biogeography. Evidence of the factors that influence distributions of organisms, whether evolution, geology or climate, may only be found as remnants or relicts in the geological record. Cladistic biogeographical methodology may provide evidence for or against geographical congruence, rather than recreate a scenario of earth's biotic history. One may, however, be able to resolve the factors that cause ambiguity. Not all ambiguity is automatically the result of incongruence and there are several ways in which these may also be resolved.
Taxa that are widespread and areas that appear as duplicated in area cladograms are the known forms of ambiguity. Ambiguity is incongruence and must either be resolved or ignored, because the aim of cladistic biogeography is to discover geographical congruence. Inability to resolve ambiguity disallows congruence. The duplication of areas, herein called geographical paralogy (Nelson & Ladiges, 1996), contains no information on area relationships, thus areas A and B are paralogous or uninformative in the areagrams A(AA) and A(BB). The TAC A(A(B(B(CC) contains only three areas and thus one informative relationship, A(BC) (Fig. 2). The areas were originally based on six taxa but it is the relationships of the three areas that form the hypothesis of area relationships. Zandee (1995) believed that by resolving paralogy and widespread taxa (i.e. using Assumption 2), information with respect to interrelationships of areas is removed, thus violating Hennig's (1966) auxiliary principle. The auxiliary principle states that homoplasy should not be assumed beyond necessity. However, ambiguity in cladistic biogeography cannot be compared with homoplasy in cladistics. Homoplasy may be defined as `a character that specifies a different and overlapping group of taxa from another character' or `any character that is not a synapomorphy (homology)' (Kitching et al., 1998, p. 208, their italics). Zandee (1995, Chapter 6, p. 4) believes homoplasy to be assumed when widespread taxa are rejected when using Assumption 1 (Nelson & Platnick, 1981) and Assumption 2. However, taxa and their characters are never used in any component-based analysis. Once a cladogram has been converted to a TAC, only the areas are analysed. Thus, Hennig's (1966) auxiliary principle can never be violated in cladistic biogeography.
Widespread taxa also cause ambiguity and result in two or more areas occupying a terminal node or branch. Unlike geographical paralogy, widespread taxa cause a surfeit of potential information and may be resolved into a simple answer. The TAC AB(BC) contains the area relationships A(BC) or B(BC). The latter, is paralogous and is interpreted as BC. Hence A(BC) + BC yields one result only, A(BC). Widespread taxa are potentially informative as they contain more than one possible set of relationships. Resolving potential information also conforms to Hennig's (1966) auxiliary principle.
Existing methods that deal with resolving ambiguity include; Assumptions 0, 1, 2, subtree analysis, and Brooks' parsimony analysis (BPA). Several reviews (Crisci et al., 1991; Morrone & Carpenter, 1994; Crisci & Morrone, 1995; Biondi, 1998; Humphries & Parenti, 1999; Ebach & Edgecombe, 2001) discuss the implementation of the above methods. The implementations of each method differ greatly so that one TAC may be resolved to yield different results. Cladistic biogeography aims to discover geographical congruence, rather than generating its presence. Thus when designing methods it is necessary to be very clear about the distinction between generation and discovery.
THE ART OF DISCOVERY
The aim of cladistic biogeography is also its concept. A concept is defined as an idea or assumption. In the case of cladistic biogeography, the concept is that geographical congruence is evidence for allopatry. In order to validate the concept, a method is required. The role of the method is to find evidence to validate the concept. If the concept cannot be validated one of two actions may be implemented. Either accept the concept and construct an adjusted method (herein patch method) that qualifies the concept, or, reject the concept. By rejecting the concept, the method may construct a new concept that in turn constructs new methods. These two approaches are known as generation and discovery, respectively.
The generation paradigm never constructs new concepts. Examples of the generation paradigm are a posteriori methods [Assumption 0, BPA and modified BPA (mBPA)]. These methods immunize results from scrutiny by introducing qualifiers. Constructed methods that reject, that is fail to qualify, the concept, are replaced by patch methods (Fig. 3a). Patch methods justify the use of errata. Patch methods thus act as immunizers to falsification. Methods under the generation paradigm are unable to explore new ideas or concepts.
The discovery paradigm allows for concepts to be rejected. Methods that fail to qualify concepts are used to establish new concepts (Fig. 3b). Patch methods or any other immunizers do not exist under the discovery paradigm. Discovery explores new concepts and ideas and includes a priori methods.
Assumption 0 (Zandee & Roos, 1987) is a method constructed under the concept that geographical congruence is best explained by both allopatry and sympatry. Assumption 0 was constructed on the basis that TACs retain taxic homologues. In the TAC, A(B(CD), the widespread areas C and D are resolved on the basis that the taxa in the area C share a taxic synapomorphy with the taxa in D. Therefore, C is more closely related to D and an extra node and synapomorphy are added to create A(B(CD). Hence, the widespread areas C and D share a unique synapomorphy that is the result of sympatry. The verdict of Assumption 0 is that sympatry is congruent. What Assumption 0 fails to do is justify the unique synapomorphy once two or more other TACs are combined topographically. For example, the resolved TAC, A(B(CD), of another unrelated group, does not share the same unique synapomorphy of CD. Thus both TACs cannot justifiably be combined based on topography alone. They remain as two separate and unique histories. As more than one TAC is needed to discover geographical congruence, Assumption 0 does not qualify the concept.
Several patch methods such as BPA (discussed below) have been devised to deal with the failure of Assumption 0. The misunderstanding and failure to accept any flaw in the ailing concept stems from the indoctrination of standard approaches in cladistics.
Interpretation of components and taxa
Widespread taxa and paralogous nodes come about as the result of unsuccessful allopatric speciation, sympatric speciation, poor sampling, loss of data, missing areas and human error. Whatever causes ambiguity necessarily obscures the discovery of congruence. Congruence exists because components (nodes on areagrams) share similar histories. Ambiguous nodes do not share similar histories therefore it is plausible to assume that every ambiguous node has a unique history. These unique histories are difficult to determine by simply interpreting them from an areagram. The aim of cladistic biogeography is not to explain incongruence. To discover the individual histories of components we need to look beyond cladistic biogeography at the individual organisms and the areas that they occupy. Components do not reflect the same information for every node in each cladogram. Cladograms yield topologies that are used to compare and combine TAC's, but such a simple practice can be easily misinterpreted.
The concept of synapomorphy lies in two taxa sharing a set of characteristics that are unique. Synapomorphy is evidence for evolution, however, in standard parsimony analysis (sensuPlatnick, 1993) it is generated as proof for evolutionary transformation series. Components are derived from nodes as structures in a branching diagram. The features that lead to synapomorphies between two taxa are not replicated, but rather overlaid onto another branching diagram. That same synapomorphy that is congruent in a cladogram may be paralogous in a TAC if it were to contain a duplication of the same area. Components and nodes do, however, share the same function, to form branches and terminal nodes.
The TAC A(A(B(B(CC) is replicated by a cladogram that consists of six different taxa (Fig. 4). The five nodes that occur on the cladogram are unique synapomorphies consisting of homologues yielded from and data set, such as morphology or DNA sequences. The TAC however, does not consist of unique components. Thus the branching diagram is reverted to a statement about areas, rather than taxon relationships. The unique synapomorphies in cladograms yield non-informative B(CC), B(CC), A(BC) and informative statements A(BC), the sum total of which is the simple area relationship A(BC) (Fig. 5). Generative methods however, treat areagrams as cladograms, a common misconception. Here, components are treated exactly as nodes, and they are said to contain information about areas and the relationships of taxa. This is the latest generative patch method termed modified Brooks Parsimony Analysis (mBPA) (sensuLieberman, 2000).
The art of generating: a critique of BPA
I ought to know by this time that when a fact appears to be opposed to a long train of deductions, it invariably proves to be capable of bearing some other interpretation. A Study in Scarlet (Doyle, 1993, p. 49).
Cladistic biogeographers reading the papers of van Veller and colleagues would get the impression that the only worthwhile method in historical biogeography is BPA or mBPA (sensuLieberman, 2000). To us it fails to address the objections made by critics such as Page (1990), Nelson & Ladiges (1996), Humphries & Parenti (1999) and see Ebach (2001). van Veller (2000) in his PhD thesis and various other contributions (Green et al., 2002; Brooks et al., 2001; van Veller & Brooks, 2001; van Veller et al., 1999, 2000, 2001, 2002c; van Veller, 2002b) all suggest that it is possible to classify biogeographical methods into a priori and a posteriori methods, on the grounds that the former somehow remove or manipulate raw data so as to obtain results, and the latter, naturally, is more complete by using all available data. In so doing the authors malign the techniques and aims of so-called a priori methods, such as component analysis (and Reconciled Tree Analysis) and we believe that this dichotomy is in false apposition. The difference as we see it is more acute. BPA is a method using Hennigian cladistics for describing evolutionary scenarios rather than a method for determining the relationships of areas using cladistics. Furthermore, the division into primary and secondary BPA (mBPA) suggests that there is a cladistic means of discriminating between dispersal, extinction and speciation processes, in our view false beliefs. Cladograms are effectively silent about any process as they determine patterns. Cladistic analysis is an inductive procedure that is absolutely silent about these things.
Thus, as a method mBPA has misconceptions of component analysis in three ways; the alleged removal of data of a priori methods, a misunderstanding of Assumptions 1, 2 and 0, and consequently misplaced views as to requirements of a valid method.
The confusion about removal of data is a misunderstanding of the difference of taxa and areas, or biological homologues and geographical homologues. Consider a three taxon statement, for example for three species, a, b, c. The relationships amongst these three taxa can be resolved by an informative character (the presence of similar homologues in two of the taxa) to give an expression of relationship, a(bc). Defining a group (abc), as monophyletic is not a resolution as we understand it, but merely an assertion of the original problem to be solved. Consider that the three species, a, b and c occur in the following areas: a in A, b in B and C and c in C and D. Using Rosen's (1978) method for replacing the taxa with the areas in which they occur would give the following area cladogram A(BC,BD). Remembering that the homology supporting the statement that b and c are more closely related to each other than either are to a, means that the biological homology cannot be violated in area cladograms. The area cladogram for a, b and c is complicated by the fact that the species are widespread (two species occurring in two areas) and that there is geographical paralogy (i.e. the areas of each species overlaps with at least one other). There are only four areas, A, B, C and D. What are their relationships? By removing the paralogy from the area cladogram (using Assumption 2) so as to represent each area just once, the relationships have the following possibilities: A(BD), A(CB), A(BD) or A(CD). Summarizing the information content of the four three-item statements would give A(BCD). All four areas are represented and the original homology for the species relationships, a(bc), has not been violated. So the question we could ask is, does the use of component analysis and Assumption 2 remove data?
In his PhD thesis van Veller (2000; see also van Veller et al., 1999), developed two criteria for consistency of all pattern-based methods with respect to their capacity for finding general area cladograms for different monophyletic groups. van Veller et al. (1999) claimed that if one assumes a priori that inclusive sets of independent, and thus additive processes under Assumptions 0, 1 and 2, this should result in inclusive sets of cladograms (requirement 1). Secondly, van Veller et al. (1999) argued that these sets of area cladograms obtained for different monophyletic groups should be compared under the same assumption to obtain valid general area cladograms (requirement 2). The second requirement is trivial, as only bad practice would compare results under different assumptions. However, the first requirement is more serious. It fails through the use of Assumption 0 by confusing the relationship of taxa and areas to one another. By treating areas taxically, i.e. areas at the terminals of branches are treated in exactly the same way as species in a fundamental cladogram; the areas are considered the same kind of terminal entity as species and other biological taxa. Thus species a might occur only in areas A and B but a false node, A + B would be added into the area cladograms. By way of an example consider the following cladogram: a(bc). Taxon a occurs in areas AC, taxon b occurs in area C and c occurs in A, B and C. The initial area cladogram would be represented as AC(C,ABC). Under BPA, coding for each of the four nodes, rather than simply representing the one informative node, would yield one solution, B(AC) because species a occurs in both areas A and C. Under Assumption 2 and using component analysis, a method that requires only the informative node as critical, delivers the solution, A(BC). The difference occurs purely because BPA introduces additional internal nodes for taxa occurring in two or more areas, a in AC in this case providing the only signal. In other words, the result B(AC) would never appear under Assumption 2 and therefore could not be nested within it. Consequently requirement 1 is violated by the use of Assumption 0 in the first place.
The preference for using the a posteriori method, BPA (or mBPA) and Assumption 0 is misguided, as it only really considers species histories (Ebach, 2001). Species (or other taxa) are treated as if they are characters and areas are treated as if they are taxa. Such an approach causes spurious results by introducing area relationships on the basis of widespread distributions rather than sister group relations between areas of endemism.
AREA CLADISTICS: DISCOVERING THE RELATIVE POSITIONS OF CONTINENTS THROUGH TIME
There is again a general question that remains uninvestigated: whether there is a relation between biotic similarity and the geological relationships of the areas occupied by organisms (Nelson & Platnick, 1981, p. 508).
Cladistic biogeography discovers the relationships between endemic areas, it does not speak of lineages, migrations or centres of origin (Nelson, 1978). This discovery may be significant, for relationships of areas may tell us something about continental drift and other geological and geophysical events. A new technique called area cladistics amalgamates cladistic biogeography (sensuNelson & Platnick, 1981; Nelson & Ladiges, 1996) and the application of cladistics to terrane analysis pioneered by Young (1984, 1995). Area cladistics (sensuEbach, 1999; see Ebach & Edgecombe, 2001) finds the relative positions of areas through time, given the following starting assumptions:
1. Allopatric speciation results in geographically congruent data, and
2. Continental drift is directly or indirectly responsible for allopatric speciation.
Allopatric speciation and geographical congruence
Geographical isolation causes populations of various organisms to become disjunct and isolated. Typical barriers include glaciers, mountain chains and rift valleys. The physical barriers may affect whole biotas at one time. A series of geographical barriers, for instance, may divide up one area into several smaller areas over time. This will affect the original fauna and flora and separate it into distinct smaller biotas. Isolation and environmental change will, over a period of time, cause the populations of species in each biota to become genetically isolated and eventually genetically independent. Thus geographical isolation is responsible for allopatric speciation. Apart from mutation, sympatry and parapatry are amongst the few other non-geographical evolutionary mechanisms. Physiological and behavioural differences that are unique in one population will cause sexual and genetic isolation within a species. Such isolation events do not affect different faunas at the same time and many different sympatric speciation events may take place within each species. Mutation is similar in that it occurs in individuals or small populations that are affected by a high degree of chemical or physical changes. Sympatry and mutation do not form geographically specific patterns and thus cannot be detected by cladistic biogeography. The topology of the areagram is a reflection of geographical isolation. Geographical isolation events are directly (mountain chains) or indirectly (ocean currents) caused by continental drift. The movement of continents builds mountains and form basins. High mountains in high latitudes form glaciers; wide rifts form seas or oceans. Continental drift may be directly or indirectly responsible for allopatric speciation and geographical congruence in areagrams. Drift does not only move continents, it drives evolution (sensu Croizat, 1964).
Aims of area cladistics
Area cladistics aims to discover the relative positions of continents and provides independent evidence for continental drift. Three-item analysis is the most robust method for yielding cladograms, as it makes no assumptions regarding character evolution. Paralogy-free subtree analysis using Assumption 2 is the best method to yield areagrams (Nelson & Ladiges, 1996; Ebach, 1999; Humphries & Parenti, 1999; Ebach & Edgecombe, 2001) because it aims only to find congruence and utilize potential information.
In an area cladistic analysis areas are treated as `endemic'. Endemicity is herein defined as the known geographical distribution of a taxon. An endemic area is not the area of origin. Centres of origin are impossible to define using either cladistic biogeography or area cladistics (see Ebach, 1999). Geographical distribution is defined as the ecological or physical range boundaries of a taxon. A species may be confined by two mountain ranges. However, if the organism is a fresh water crustacean that lives only in streams, then distribution between mountain ranges is restricted even further (Fig. 6). Streams wander over time. Thus the sedimentological extent is defined by channel gravel and sand deposits that indicate the maximal range of that taxon. Areas that are now unoccupied cannot be defined accurately.
As a thought experiment the same area described above, when confined by a mountain range, may lie on Devonian shallow marine sediments. These sediments indicate that the rock was once a reef inhabited by numerous organisms preserved in situ, including a single specimen of a marine arthropod. The distribution of that single specimen can only be based on the extent of the reef. A reef is a biozone, a unique ecological and physical boundary. Biozones are distinct and characterized by their palaeontological and lithological composition. They are distinguished from other neighbouring environments interbedded within the same formation. Biotas that are not in situ are not deposited in their original biozone. The extent of such faunas may be determined in terms of biotic composition and lithology.
Geographically extant areas containing fossil taxa are not treated as the same area. This can be applied to the above stream example. Although they inhabit the same space they are disjunct over time and constitute two quite different areas. The same principle applies to fossil faunas inhabiting two different time zones and biozones. Long ranging taxa that inhabit two or more biozones in the same space can be treated as widespread taxa.
The importance of sampling different data sets
In a cladistic biogeographical analysis, more than two TACs are essential to find a geographical pattern. The greater the number of TACs the more information can be extracted concerning geographical patterns. The same is true for sampling different data sets. If we were to examine twenty TACs that represent twenty genera of one monophyletic taxonomic family we may be biasing our result towards one form of sexual reproduction, habitat, ecological zone or climatic region. The majority of corals for instance live in warm, shallow water facies that are confined to the tropical and subtropical climatic regions of the world. They spawn and have a larval and nektonic stage. The distribution of corals may be indirectly affected by sea currents, changes in temperature and geology as reefs may form on extinct marine volcanoes called guyots. An increase in the variety of unrelated taxa, for example fishes, for instance will increase the amount of information concerning those areas. Fishes also spawn, however, they are nektonic throughout their lives and may move away from currents. They do not rely on guyots nor are they greatly affected by air temperature or slight sea-level fluctuations. An increase in the variety of data prevents any physical biases from influencing the distribution of one taxonomic group to distort the data.
Once endemic areas are established and data assembled into TACs, a cladistic biogeographical analysis can be undertaken successfully using the parameters (sampling of different data sets, etc.) that are specified above.
Minimal trees and general areagrams
Currently only one program exists for implementing a paralogy-free subtree analysis using Assumption 2. TASS (Nelson & Ladiges, 1995) runs via an MS-DOS interface. The program requires trees to be entered manually. Subtrees that are found among the various combinations of area relationships are constructed into a three-item matrix at the end of the analysis. The matrix is then imputed into any parsimony program such as PAUP* v4.0 (Swofford, 1999) or NONA (Goloboff, 1998) to yield a most parsimonious tree or consensus tree. The characters from the three-item matrix are optimized onto the most parsimonious tree, yielding a summary of homologues. These programs are problematic, as the matrix does not reflect homology. The use of a three-item matrix alleviates this problem. However, programs such as PAUP* use optimizations that yield different results. NONA has a non-ambiguous character transformation option, which is perhaps the best way to manually implement a minimal tree using parsimony software. A minimal tree is the tree that best fits the data (Williams, 1996; Kitching et al., 1998). Minimal trees are similar to cliques or strict consensus.
The consensus or minimal tree functions as the tree that represents the combined common patterns of the imputed TACs. This tree shows the degree of geographical congruence and is described as the general areagram. The general areagram is the final stage of a cladistic biogeographical analysis (see Ebach & Edgecombe, 2001, Fig. 9). It tells us about the relationships of areas. Once this information is applied to reconstructing the past, we start an area cladistic analysis.
Palaeomagnetics and palaeolongitude
The magnetism of minerals can be used to measure the direction and distance of the magnetic poles. The palaeomagnetism that occurs in rocks of varying ages indicates that minerals in oceanic and continental crusts are aligned at a particular direction to indicate palaeolatitude (Fig. 7a). Continental drift helps to explain the irregular directions of these minerals in igneous rocks. These magnetic readings form similar patterns in rocks of similar age, indicating that the continents have moved over time, thus displacing the original reading. Palaeomagnetic maps are constructed from palaeomagnetic data. The alignment of minerals in rocks of a particular age can indicate the palaeolatitude of rocks and in turn the distance of continents from the poles (Fig. 7b). Although palaeomagnetics can place continents at particular latitudes, it can however,
provide no information on palaeolongitude, and so the E–W separation of the continents is indeterminate (Kearey & Vine, 1990, p. 58).
Other geological data, such as sedimentology and fossil data usually provide information regarding palaeolatitude rather than palaeolongitude. For instance, the extent of reefs, evaporite basins and fossil organisms are controlled by climate that is graded palaeolatitudinally.
Area cladistics: a technique to align continents
The aim of area cladistics is to use the area relationships that are found in a general areagram in order to find the relative positions of continents through time.
The concept of area cladistics is derived from the notion that allopatry is responsible for geographical congruence. Therefore, if B and C are more closely related to each other than either are to A and that allopatry was responsible for the pattern, then B and C are not only more closely related to A, but they are also geographically closer to one another than they are to A (Fig. 8). Therefore degrees of relationship equal relative distance.
The method of area cladistics begins by overlaying the relationships of areas onto palaeomaps for any given time. For example, early Cretaceous areas would be plotted onto an early Cretaceous palaeomap. Palaeomaps may be constructed from existing palaeontological information and from palaeomagnetic data. Given these relationships, then B and C are geographically closer to each other than either are to A. A hypothetical example using ten areas is given in Fig. 9(a). The existing map has five continents contrived from palaeomagnetic data. Because of the nature of palaeomagnetic data the continents are constrained to palaeolatitude. A general areagram depicting the relationships of areas acts as an alternative source of information (Fig. 9b). The relationships of the areas define the relative geographical positions of the areas. The relationships between the areas F, G and H, do not correspond to the geographical position of the areas on the palaeomap. By aligning the continents palaeolongitudinally (Fig. 9c) so that they correspond to the relationship of the areas given in the general areagram, we see that the areas F, G and H are geographically closer to each other than they are to other areas. The implementation of area cladistic techniques to existing palaeomaps does not exclude palaeomagnetic data; rather it fine-tunes the weaknesses in the geological data, namely the unknown palaeolongitudes, to accommodate the biological data. Area cladistics is a unique way of combining biological and geological techniques to acquire a detailed result.
An empirical area cladistic example is given in Fig. 10. The global area cladistic analysis consists of four Lower Devonian trilobite TACs that include the genera Acanthopyge (Lobopyge), Paciphacops, Leonaspis and Cordania were combined to form a general areagram (see Ebach & Edgecombe, 2001). The relationships of areas in the resulting general areagram reflect the positions of plotted areas on the palaeomagnetic map (Fig. 11). New South Wales, however, has a closer association to the Appalachian province than it does to Victoria. The areas may be adjusted by aligning the palaeolongitude of Laurentia so it is closer to the eastern Australian coastline. This alignment neither affects the palaeomagnetic data nor the other relationships with the general areagram.
The Lower Devonian trilobite example is based only on four TACs of one particular kind of organism. As stated above, many TACs of many various organisms need to be combined in order to yield a more accurate result.
Area cladistics may be also used at a local scale. Neighbouring areas that share a similar tectonic history and similar faunas are ideal to use in a local analysis. A series of tectonic events may have isolated a fauna repeatedly to yield several different but related faunas (Fig. 12). Similar approaches may be used at smaller levels as taxa represent groups of organisms that may include populations. Ecological areas formed and defined by tectonic events such as glaciation, mountain ontogeny or rift basins may be used as endemic areas. An example of a local analysis is seen in the Ordovician–Silurian trilobite example using the TACs of three genera (Figs 13–15) Hibbertia (see Ebach & McNamara, in press), Wallacia and Distyrax (see Ebach & Edgecombe, 2001) that mainly inhabit Laurentia. The areas are plotted onto a palaeomagnetic map and compared with the general areagram (Fig. 16). In this example, the biological data does not conflict with the palaeomagnetic data, both in palaeolongitude and latitude. As in the Lower Devonian example, more varied data are needed for greater accuracy.
Local analyses can also be made at a much smaller scale with taxa confined to one continent or several basins. A similar approach termed phylogeography (Avise, 2000) views the world from a taxic/ancestor-descendant perspective. Phylogeography has re-invented dispersal biogeography. It completely ignores Platnick & Nelson's (1978; see also Nelson & Platnick, 1981) original ideas (but see Riddle, 1996). Phylogeography is limited in its perspective, as it has not overcome the logical hurdles already addressed in cladistic biogeographical methodology over the last two decades. Prior knowledge, it seems, is neither assumed nor necessary in phylogeography.
Axioms of area cladistics
The concept of allopatry provides the underlying rationale for the method of cladistic biogeography. In turn this justifies a number of axioms that are central to the method of area cladistics. Area cladistics incorporates several lines of reasoning that can be described with six axioms:
An endemic area is the known distribution of a taxon.
A taxon may be interpreted as a species or any monophyletic group (i.e. population or among species).
Congruence is a result of allopatric speciation, the direct or indirect consequences of continental drift.
Allopatric speciation will affect unrelated taxonomic groups occurring in similar areas. This is not true for parapatry or sympatry. Non-geographical isolation is unlikely to affect more than two species at one time.
The relationship of areas equates to the relative geographical proximity of areas.
The relationships of areas, as indicated by the terminal components on the general areagram, are the outcomes of close geographical association (proximate distance). The general areagram functions as a branching diagram that indicates relative distance and components operate as nothing more than bifurcating junctions.
Area cladistic data may not be necessarily consistent with other forms of data.
Results that conflict with palaeomagnetic data may be resolved by aligning continents in palaeolongitude. Palaeomagnetic data are also qualitative. However, because of the nature of palaeomagnetic data, palaeolongitudes may be imprecise.
Sampling of different data sets will yield consistent results.
Organisms that are distantly related, inhabit different niches and have variable ontogenies, will not be influenced in the same way by environmental factors such as water depth, temperature or ocean currents. These factors may be dismissed as influencing particular patterns. Numerous data sets of two or more taxonomic groups are vital to ensure accurate results.
Palaeomagnetic and area cladistic data provide independent evidence for continental drift.
The combination of palaeomagnetic and area cladistic data will yield high quality palaeomaps with which to interpret patterns of continental drift and palaeoclimates.
AREA CLADISTICS AND PALAEOMAGNETICS: A NEW COLLABORATION
At present, area cladistics has not been used very much and the reason is quite simple: there are few cladists in palaeontology. To date, where palaeontologists have disagreed with palaeomagnetic re-constructions, their own palaeontological re-constructions are based on a single taxonomic group.
Palaeontologists that are restricted to one particular taxonomic group, also find problems with palaeomagnetic data. The distribution and relationships of their organisms do not match the positions of areas in palaeomagnetic maps. This is an easy criticism to make, because using the distribution of only one particular taxonomic group is not exhaustive, and does not `test' the validity of palaeomagnetic data. Area cladistics is independent evidence for continental drift, and palaeontologists in two ways may use it – either solely use area cladistics to test the accuracy of palaeomagnetic data, or combine the methods of both area cladistics and palaeomagnetics. Palaeomagnetic data should never be completely disqualified. It provides valuable information about palaeolatitude in cases where fossil evidence is absent. One should be able to test palaeomagnetic data and use other non-geological evidence to determine its accuracy. `Accuracy' is defined as the corroboration of geological and palaeontological evidence to interpret the earth's history. Such corroboration is an immense benefit to palaeomagnetics, in sorting out problematic data. Palaeomagnetic data that conflict with taxic distributions of particular groups can be easily used given the relative incompleteness of the fossil record. However, the way to test data thoroughly is to use a robust method, namely area cladistics, that samples numerous different data sets of unrelated taxa to find a congruent pattern. The combination of area cladistics and palaeomagnetic data can yield highly accurate palaeomaps that both independently and collectively yield evidence to support continental drift.
Area cladistics provides palaeontologists and palaeomagneticists with independent evidence for continental drift, a new way to align continents and a test of corroboration. It also provides palaeobiogeography with an exciting future.
Palaeobiogeography has the potential to become a powerful methodological tool to rival palaeomagnetics. The future relevance of palaeobiogeography lies in the development of new ideas, and area cladistics is one such promising idea. However, before we reach the goals of collaboration, we need to proceed with a viable method, one that does not accept `insufficient data' as admissible.
We approached the case you remember, with an absolutely blank mind, which is always an advantage. We had formed no theories. We were simply there to observe and to draw inferences from our observations.
The adventure of the cardboard-box (Doyle, 1993, p. 895).
BIOSKETCHESMalte C. Ebach's research investigates the theoretical problems inherent in standard cladistic and cladistic biogeographical approaches and systematic revisions of mid-Palaeozoic trilobite groups.Chris J. Humphries' research has concentrated on several problems in historical biogeography, both theoretical and empirical, angiosperm systematics, biodiversity measurement and methods of priority selection in conservation.