***For a PDF file of this essay, click here***

With the focus of this study on species, a clear account is needed of what species concept is being employed, what criteria are being used to delimit species, their justification with regard to that concept, and how those criteria can be applied to molecular and morphological data. Here it will be argued that species are real, and the question of how many species there are within a given clade is therefore in principle answerable with a definite number. A clear definition will be provided of what is meant by the term "species concept", and it will be suggested that some so-called species concepts (such as that of van Steenis 1957) are in fact implementations - that is, attempts to make operational - inexplicit species concepts. It will be argued that species concepts should be pattern-based and character-based, rather than mechanistic or history-based, so far as possible, to avoid circularity of evolutionary hypotheses. Following Cracraft (1997), it will be accepted that there is a need for some minimal statement of process (self-perpetuation) in any species concept, though it may be difficult to render this element operational with only morphological data. The species concept adopted here is a phylogenetic species concept which is a modification of that of Nelson & Platnick (1981).

Morphological data have traditionally been the basis of species delimitation, both in Agathis and more generally in systematics, but in Agathis morphology presents special problems. The number of fertile specimens in major European herbaria of each species recognized in the most recent set of revisions (de Laubenfels 1972, 1988) is fewer than five in total for many recently described species (de Laubenfels 1969, 1979) and even for some species described last century, only three or four fertile specimens are known. Leaf morphology appears to be continuously variable within and between most species, but is presently being investigated in detail to check whether multivariate analyses are capable of recovering discrete clusters of points in the data: so far the evidence is unconvincing. The number of fertile specimens so far gathered is insufficient to address species delimitation in a statistically defensible way across the genus as a whole, although some of the better-collected species are clearly highly distinct on these characters . Furthermore, fertile material of Agathis is for logistical reasons very difficult to collect [the phenology is not well understood and in addition many species are extremely tall trees (Table 3.1 in Whitmore 1998)]. Whilst the further investigation of morphological characters in Agathis remains important as a possible source of characters appropriate for species delimitation, molecular data are potentially a major help in elucidating species boundaries in the genus.

Molecular data presents especial difficulties for methods of species detection (henceforth, implementations), as most have been (a) intuitive, and (b) based on the detection of discontinuities in continuously varying morphological data. Three published implementations of the phylogenetic species concept (field for recombination (FFR), population aggregation analysis (PAA), and cladistic haplotype analysis (CHA)) are examined and analysed for their advantages and shortcomings as means of detecting phylogenetic species with molecular data. FFR is rejected as inappropriate for consideration, and CHA is suggested as more appropriate than PAA, where the data permit, as an implementation of the phylogenetic species concept.

Species are 'real' in the sense that they are discovered rather than invented, because only by accepting that species are real, and that there can be a single correct answer to the question "how many species are there?" is debate over their boundaries meaningful rather than purely scholastic (Cracraft 1997). Such an assertion of species reality is a metaphysical assertion of scientific realism - that there is a world external to methods of enquiry - and as such cannot be directly tested, although it is fundamental to scientific enquiry (Psillos 1999). However, partial evidence in favour of the view that species are discovered rather than invented may be had from independent enumerations of species by different cultures. The classical example of this is that Ernst Mayr and the people of the Arfak Mountains of western New Guinea respectively recognized 138 and 137 species of birds from the mountain range (Mayr 1942): a correspondence which would be miraculous if species are arbitrary as some authors have argued (e.g. Levin 1979). As Platnick (2001) has commented, systematists are in various stages of retreat from the notion, prevalent during the 1950s, that species were the only biologically 'real' categories in the taxonomic hierarchy, with some biologists now taking the view that species are wholly dispensable in a totally monophyletic system (Pleijel 1999), or simply artificial amalgamations of the real units of evolution, populations (Levin 1979). By requiring monophyly of all taxa (Pleijel 1999), taxonomists neither recognize the reality of tokogenetic processes nor can provide means of identifying members of 'paraphyletic' species (Olmstead 1995; Crisp & Chandler 1996). Levin's argument that we should be naming units of evolution and that those are populations, not species (Levin 1979), fails further by ignoring the possibility that for some evolutionary processes, species rather than populations are indeed the units of evolution for the simple reason that some species do not usually occur in populations sensu Levin, and that that property has shaped differing evolutionary fates (Vrba 1984).

A species concept is therefore a statement of what a species is, rather than how we may discover it (Cracraft 1997), though many concepts may be regarded as doing both at once. The corollary of this argument is that if species are real, a method or methods of detection are particular scientific procedures with merits or demerits that are separable from the merits and demerits of a conception of how species are defined. A method of species detection - that is, a procedure for establishing whether two or more sets of individuals belong to the same or different species - is here referred to as an implementation. An implementation is therefore an attempt to make operational a particular species concept, which may or may not be made explicit. Particular implementations are likely only to be worthwhile as operationalizations of a single species concept - for example, Doyle's FFR technique (Doyle 1995) for species delimitation would be considered irrelevant by adherents of a topology-based species concept such as Templeton (1989).

Species concepts have been enumerated and reviewed several times recently (Luckow 1995; McDade 1995; Hull 1997; Mayden 1997; Dupré 1999; Lherminer & Solignac 2000), and these have included attempts to classify the many competing concepts (Davis 1995; Luckow 1995; Mayden 1997). Debates have included the possibility or otherwise of a monistic species concept, (Ereshefsky 1992; Hull 1997; Dupré 1999), the importance or relevance of historical criteria (Baum & Donoghue 1995; Davis 1997), the question of species monophyly (Crisp & Chandler 1996) and the issue of whether species should be treated as individuals or as classes (Hull 1964, 1965; Rieppel 1986; Baum 1998). As Farris (1985) has pointed out, some of these debates are not of obvious relevance to biological problems despite their apparent philosophical interest. However, differing species concepts are not always best understood as directly comparable, because 'species concepts' as traditionally understood may include either a primary statement of what a species is (a species concept in the sense used here), or a discovery procedure for finding them (an implementation in the sense used here), or both. The implications of this, in terms of seeing some so-called 'species concepts' as merely implementations of species concepts in the sense used here, will be discussed later.

Following Luckow (1995) a distinction may be made between mechanistic concepts, where species are defined as participants in some evolutionary process; and historical concepts, the end results of evolutionary processes. Within the field of historical concepts, Luckow (1995) recognizes a further distinction between ancestor-based and character-based concepts, a distinction she regards as that analysed by Rieppel (1994) as between pattern-based and process-based concepts. These groups of concepts will be examined in turn.

Mechanistic concepts such as the biological species concept of Mayr (1942) and the recognition species concept of Paterson (1985) are generally flawed for at least three reasons. Firstly, by identifying a particular speciation process as the defining element in our understanding of species, they remove the possibility that the investigation of speciation processes will be anything other than trivial; secondly they are inherently impossible to implement satisfactorily because they depend on observations of processes rather than of patterns; and thirdly they are highly dependent on the biology of the species under study - Mayr's concept is inapplicable to asexual organisms, and the applicability of Paterson's concept to flowering plants is very questionable. Even in sexual organisms, it is possible for morphologically and ecologically highly distinct species to persist over time whilst still apparently capable of interbreeding (Van Valen 1976). An additional practical problem with mechanistic species concepts is that the data that would be necessary for any suggested implementation is extremely difficult to collect: furthermore, supposed implementations of mechanistic concepts (e.g. Mayr 1992) have turned out on subsequent analysis to be nothing of the sort (Whittemore 1993). Some authors have suggested that allele topologies should be used in species delimitation (Templeton 1989; Avise & Wollenberg 1997; Avise 2000), but the realization that for some genes [e.g. the MHC complex of immunologically important genes in primates (Gaur et al. 1992) and self-incompatibility genes in angiosperms (Clark & Kao 1991)], allele topologies do not coalesce anywhere near as low down the taxonomic hierarchy as commonly accepted species boundaries, and that in any case there is no necessary reason for expecting gene trees and species trees to be isomorphic, has led other workers to regard gene-trees as irrelevant for species delimitation (Davis & Nixon 1992; Doyle 1992, 1995; Luckow 1995).

Amongst historical concepts, "ancestor-based" concepts, sensu Luckow (1995), are those in which species are conceived of as being historically-defined units where species are hypothesized as "systems of common descent even if they have no modifications (apomorphies) by which we can recognize them" (de Queiroz & Donoghue 1990). "Character-based" concepts are those such as Nelson and Platnick (Nelson & Platnick 1981): "species are the smallest detected samples of self-perpetuating organisms that have unique sets of characters". Since we clearly cannot ever know for certain what the reality of species boundaries is, the division between such concepts comes down to what classes of implementation are desirable means to the recognition of species. Whilst the belief that species are real discrete entities makes the minimal assumption that there exists a real external world about which empirical data may allow us to make inferences, the appeal in ancestor-based concepts to shared common descent rests on both an understanding of evolution and on another step of inference. Whereas character-based concepts provide for the inference of existence of real entities (species) directly from empirical data (characters), ancestor-based concepts infer the existence of real entities from inferences (about historical process) which are themselves inferences from empirical data (characters). For this reason, because of the unnecessary multiplication of levels of inference required by ancestor-based concepts, and because the retreat from the primary observational data necessarily increases the metaphysical assumption content of any implementation (Luckow 1995), character-based concepts - that is, character-based methods of implementation - are superior.

The species concept used here is guided by the considerations advanced by Cracraft (1997), that a species concept should include:

(i) Mention or implication of reproductive cohesion, so as to include both males and females in the same species;

(ii) A notion of diagnosability, such that species may be distinguished from one another;

(iii) Criteria for ranking populations or aggregations of populations at the species level rather than at some other level of the Linnean hierarchy;

These considerations may be rephrased as (i) a minimal mechanistic criterion of inclusion, i.e. that tokogenetic interaction must be possible within a species, (ii) a character-based criterion of exclusion, i.e. diagnosably different groups of individuals must be in different species, and (iii) a basis for recognizing these groups of individuals at the species level rather at a higher or lower rank. Nelson and Platnick's (1981) definition of a phylogenetic species concept, cited above, attempts to satisfy criterion (i) by providing that samples of organisms be populations which are self-perpetuating, (ii) by requiring that species have unique sets of characters, which makes them diagnosable, and (iii) by making 'species' the lowest rank displaying fixed unique character states which is recognized. It is adopted here with adjustments from Cracraft (1997): species are here understood as:

"the smallest population or group of populations within which there is a parental pattern of ancestry and descent and which is diagnosable by unique combinations of character-states"

Because tokogenetic interaction - reticulating relationships - within species (part of the 'self-perpetuating') is expected to ensure that actually interbreeding populations are not distinguishable by the possession of unique sets of characters, this species concept fulfils all three of Cracraft's requirements. Infraspecific ranks are accordingly not employed to refer to diagnosably different populations which happen not to be reproductively isolated from one another. Infraspecific taxonomy has been considered to have many advantages (Stace 1986), but there can be no clear justification for placing a taxon at any given infraspecific rank rather than as a species if it is diagnosably different. Infraspecific taxonomy may however be occasionally useful to name and convey information about notable variants which occur only within larger populations [such as radiate Senecio vulgaris L.: (Marshall & Abbott 1982)].

Implementation of the species concept adopted should therefore aim to ensure that species delimitation be populational in scope, so as to take into account evidence/absence of evidence of genetic interaction and an understanding of character variability. Two of the most 'operational' so-called species concepts have been the 'morphological species concepts' of Du Rietz (1930: "The smallest natural populations permanently separated from each other by a distinct discontinuity in the series of biotypes") and van Steenis (1957), who argued that correlated differences in two morphological characters should be the basis for species delimitation. These concepts rely solely on morphological data, make no explicit reference to reproductive cohesion (Du Rietz's 'permanently separated' is not necessarily the same as 'totally reproductively isolated'). Van Steenis's (1957) concept in particular is best understood as being an implementation for morphological data only of a species concept which is not actually explicitly described but is probably not dissimilar from that of Nelson & Platnick (1981). His criterion of two morphological characters for species delimitation in the Flora Malesiana project (1957) is understandable in terms of the number of well-studied species in temperate floras known to show inter- and intra-populational differences in one morphological character but which still interbreed both in the wild and under laboratory conditions (e.g. Lotus corniculatus L., Jones & Turkington 1986; Senecio vulgaris, Marshall & Abbott 1982), and indeed of sexual dimorphisms which have led over-zealous workers to propose different species for material from different sexes (e.g. in the parrot Eclectus roratus (Statius Muller), and in Pleistocene Australian populations of Homo sapiens L.: Flannery 1994). Correlated differences in two morphological characters as a criterion for delimitation can therefore be understood as a guarded implementation of the concept adopted here in the absence of any mechanistic data whatsoever.

With both the growing availability and utility of molecular methods in systematics, and ongoing debate about methods and principles in systematic biology, different implementations of the phylogenetic species concept (in the sense adopted here) have been developed explicitly for dealing with molecular data. The best-known implementations are the 'field for recombination' (FFR) approach of Doyle (1995) and the 'population aggregation analysis' (PAA) of Davis and co-workers (Davis & Manos 1991; Davis & Nixon 1992). More recently, cladistic haplotype analysis (CHA), a modified version of PAA involving some topological considerations, has been advanced alongside an analysis of the effects of different interpretations of PAA (Brower 1999). These approaches differ in their consideration of the fundamental units to be considered during species delimitation - individuals in FFR, populations in PAA and CHA, and may also differ in their practical results when applied to real data (e.g. Bailey 2001).

The field-for-recombination (FFR) approach is based on the rejection of an assumption that gene trees and species trees are necessarily isomorphic (Doyle 1992, 1995) and the rejection of the suggestion by Vrana and Wheeler (1992) that individuals are appropriate terminal taxa for phylogenetic analysis, because of the impossibility of placing a heterozygous organism in a single position on a gene tree (Doyle 1995). Furthermore, since it is clear that making species delimitation dependent on allele-tree topologies as has been suggested by some (Baum & Shaw 1995) would be unworkable for particular variable genes (Clark & Kao 1991; Gaur et al. 1992), the correct approach is instead to view gene pools as the terminal units for phylogenetic analysis. It also attempts to get round the difficulties inherent in defining a spatially bounded population constituting a subset of a putative species, by basing itself on the comparison of individual rather than populational attributes. Individuals are examined for their alleles, and heterozygous individuals are informative about species delimitation, because the possession of overlapping alleles by different individuals implies the possibility of reproductive exchange between them. Because of the possibility that some loci will be less diverse in terms of the set of possible alleles and therefore less informative about gene-pool size, Doyle (1995) suggested the use of a multilocus approach (ml-FFR) in which only groups of individuals which appear not to overlap on alleles for any locus be separated as species.

Whilst FFR, and especially ml-FFR, ostensibly offers a potentially useful approach to delimiting species, it suffers from two particular problems. Firstly, the recognition of species based on analysis of their polymorphisms cannot be rendered congruent with the aim of the phylogenetic species concept here adopted of recognizing species based on unique combinations of character states, and cannot therefore be used as an implementation of that concept . Nixon and Wheeler (1990) make the useful distinction between attributes (things that may vary between individuals), traits (attributes that vary within a species) and characters (attributes that are fixed within species). FFR, contra Doyle (1995: 584) relies for species delimitation on an analysis of attributes that vary within a species (traits) rather than the differences in fixed attributes (characters).

Population aggregation analysis (PAA) was introduced in a study of the uses of allozyme data for species delimitation in Puccinellia (Davis & Manos 1991), and subsequently amplified by Davis & Nixon (1992). It has since been employed in practical studies of species delimitation (e.g. Chamberlain 1998; Bailey 2001) The essential principle behind PAA is that only attributes that are fixed within populations (characters sensu Nixon & Wheeler (1990)) are appropriate for the delimitation of species, a view which clearly separates it from the polymorphically defined species of FFR. Also unlike FFR, it is explicitly populational and proceeds from a tabulation and an analysis of the variation to be found within populations. Clearly, the definition of a 'population' is potentially as fraught with problems as that of a 'species', but in practice the definition used by Davis & Nixon (1992) of "local genealogical units … all individuals of a local population are regarded as belonging to the same phylogenetic species" is usable: more theoretically inclined workers (Brower 1999) have considered this a reference to 'background knowledge' sensu Popper (1992). For each population, individuals within it are scored for attributes, and then a profile is assembled for the population as a whole: in a simple system with an attribute that takes two possible states, 1 is scored for attributes fixed within a population, 0 for those wholly absent, and ± for attribute present in the population but not fixed. Populations are therefore regarded as conspecific provided they they lack fixed differences between them - i.e., that it is possible for any given individual to belong to any one of the populations assembled under that species. For polyallelic loci, PAA accepts as adequate for delimitation the presence of mutually exclusive sets of alleles (e.g. A or B in species 1, C or D in species 2). PAA therefore recognizes species on the basis of fixed, unique character combinations and is an appropriate method for consideration as an effective implementation of the phylogenetic species concept here adopted.

However, PAA does suffer from some problems. Some of these, such as the lack of resolution resultant from undersampling and incorrect homology assessment are problems general to systematic biology rather than particular to PAA. Others, such as the possibility of parallel fixation in different daughter populations of a single polymorphic 'mother' species, could lead to incorrect delimitation because of the way that PAA ignores information about the possible relationships of attributes.

PAA has been criticized on this count for not making the best possible use of the data by Brower (1999). Brower points out that PAA leaves open the question for sequence data of what exactly an allele is, and shows that there are two possible interpretations of PAA, either regarding the whole sequence as a single locus, or regarding each nucleotide as a separate locus. These different interpretations he calls respectively PAA1 and PAA2, and shows that their implementation can lead to different delimitations (Fig. 3 in Brower 1999). Furthermore, although PAA2 represents an apparently more rigorous exploration of the data by atomizing the character information into the smallest units (Brower 1999), and is therefore preferable, it has the major drawback of reducing the number of possible alleles to just four for each locus. For a system such as PAA which is avowedly non-topological (Davis & Nixon 1992), this means that much of the information used in species grouping may be undetectable homoplasy by chance alone, and the method may be little different in its effects from phenetic clustering analysis.

Brower's alternative is cladistic haplotype analysis (CHA), a technique he describes as a means of circumventing this shortcoming of PAA and the way in which it allows potentially homoplastic attribute fixations to shape species delimitation. CHA attempts to ensure that all members of a phylogenetic species are joined in a contiguous section of an unrooted network, separated from each other population by a single branch that represents parsimoniously inferred character state transformations, regardless of whether that difference is apomorphic or not (Brower 1999: 202). This corresponds to an effort to ensure that the implementation of the phylogenetic species concept includes only what have been called paraphyletic and monophyletic species (Olmstead 1995; Crisp & Chandler 1996), although it has been argued that such terms should not be used to refer to the relationships of phylogenetically basal units to one another (Nixon & Wheeler 1992). CHA uses cladistic analysis of attributes as understood under PAA2 to ensure that groups are not delimited using characters which are actually homoplastic, and that apparently homoplastic characters are not disregarded where they may in fact provide some form of hierarchic structure among several differing populations: a shortcoming of PAA which is implicit in Brower's critique is that as the number of putative species and individuals within them being examined rises, the total proportion of sites which are likely to show fixed unique differences among them falls.

CHA is comparable to PAA as a means of implementing the phylogenetic species concept: both rely on a degree of background knowledge in defining populations for primary study, and accept only fixed differences (or fixed differences in the identity of all alleles occurring at a locus polymorphic in both populations). They are therefore both possible means of identifying phylogenetic species in the sense adopted here. However, because CHA retains information in the data about the topology of allele trees that is discarded by both PAA and FFR, but which increases the probability that homologous alleles are correctly identified without introducing questionable assumptions of coalescence timing, it is a superior method in terms of its use of any given dataset and the ability to ensure that delimitations are not misled by the paucity of character states in molecular data, and is accordingly considered the most promising implementation for the phylogenetic species concept here accepted. Its applicability, however, is limited to certain kinds of data: with morphology, and with microsatellite and other molecular methods which do not generate sequence data capable of being resolved into an allele tree. For such data sources, CHA is inapplicable and PAA remains the most appropriate means of analysis. There are further issues with chloroplast DNA data, with a general consensus that cytoplasmic DNA does not show recombination (Hagelberg et al. 2000), and accordingly genetic drift, in the absence of tokogenesis, is likely to be the principal force shaping interspecific differences on this genome.

In conclusion, the adoption of an explicit phylogenetic species concept necessitates the evaluation of competing implementations for that concept. Debates over the value of different species concepts are separable from discussion over the most effective implementation of a given species concept, and the value and implications of different implementations can be understood both by a priori discussion and by application to real data. Morphological data available appears to be inadequate for species delimitation in the genus under study, both because of intrinsic limitations and the paucity of adequate collections, and consequently the use of molecular characters, potentially more numerous and more informative, offers a possible means of resolution of these problems. Whilst particular possible implementations such as FFR can be ruled out as inapplicable from first principles, the value of others can depend on circumstance, the nature of the data and of the genome, and may be evaluated by comparison with other similar implementations, by the rigour with which they explore the information in any given dataset (Brower 1999), and by the analysis of species groups to understand the effects of particular population structures and processes on population-based methods of implementation. There is, as Nelson once put it, "more than one way to skin the cat of systematic endeavour".

References

Avise, J. C. (2000). Phylogeography: the history and formation of species. Harvard University Press, London.

Avise, J. C. & Wollenberg, K. (1997). Phylogenetics and the origin of species. Proceedings of the National Academy of Sciences of the United States of America 94: 7748-7755.

Bailey, C. D. (2001). Systematics of Sphaerocardamum (Brassicaceae) and related genera. Unpublished PhD dissertation. Graduate School, Cornell University, New York.

Baum, D. A. (1998). Individuality and the existence of species through time. Systematic Biology 47: 641-653.

Baum, D. A. & Donoghue, M. J. (1995). Choosing among alternative "phylogenetic" species concepts. Systematic Botany 20: 560-573.

Baum, D. A. & Shaw, K. L. (1995). Genealogical perspectives on the species problem. In Hoch, P. C. & Stephenson, A. G. (eds). Experimental and molecular approaches to plant biosystematics. Missouri Botanical Garden, St Louis, pp. 289-303.

Brower, A. V. Z. (1999). Delimitation of phylogenetic species with DNA sequences: A critique of Davis and Nixon's population aggregation analysis. Systematic Biology 48: 199-213.

Chamberlain, J. R. (1998). Isozyme variation in Calliandra calothyrsus (Leguminosae): its implications for species delimitation and conservation. American Journal of Botany 85: 37-47.

Clark, A. G. & Kao, T. H. (1991). Excess nonsynonymous substitution at shared polymorphic sites among self-incompatibility alleles of Solanaceae. Proceedings of the National Academy of Sciences of the United States of America 88: 9823-9827.

Cracraft, J. (1997). Species concepts in systematics and conservation biology - an ornithological viewpoint. In Wilson, M. R. (ed.) Species: the units of biodiversity. Systematics Association Special Volume No. 54. Chapman and Hall, London.

Crisp, M. D. & Chandler, G. T. (1996). Paraphyletic species. Telopea 6: 813-844.

Davis, J. I. (1995). Species concepts and phylogenetic analysis - introduction. Systematic Botany 20: 555-559.

Davis, J. I. (1997). Evolution, evidence, and the role of species concepts in phylogenetics. Systematic Botany 22: 373-403.

Davis, J. I. & Manos, P. S. (1991). Isozyme variation and species delimitation in the Puccinellia nuttalliana complex (Poaceae) - an application of the phylogenetic species concept. Systematic Botany 16: 431-445.

Davis, J. I. & Nixon, K. C. (1992). Populations, genetic variation, and the delimitation of phylogenetic species. Systematic Biology 41: 421-435.

Doyle, J. J. (1992). Gene trees and species trees: molecular systematics as one-character taxonomy. Systematic Botany 17: 144-163.

Doyle, J. J. (1995). The irrelevance of allele tree topologies for species delimitation, and a nontopological alternative. Systematic Botany 20: 574-588.

Du Rietz, G. E. (1930). The fundamental units of biological taxonomy. Svensk Botanisk Tidskrift 24: 333-428.

Dupré, J. (1999). On the impossibility of a monistic account of species. In Wilson, R. A. (ed.) Species: new interdisciplinary essays. MIT Press, London, pp. 3-22.

Ereshefsky, M. (1992). Eliminative pluralism. Philosophy of Science 59: 671-690.

Farris, J. S. (1985). The pattern of cladistics. Cladistics 1: 190-201.

Flannery, T. F. (1994). The Future Eaters: an ecological history of the Australasian lands and people. Reed New Holland, Sydney.

Gaur, L. K., Hughes, A. L., Heise, E. R. & Gutknecht, J. (1992). Maintenance of DQB1 polymorphisms in primates. Molecular Biology and Evolution 9: 599-609.

Hagelberg, E., Goldman, N., Lio, P., Whelan, S., Schiefenhovel, W., Clegg, J. B. & Bowden, D. K. (2000). Evidence for mitochondrial DNA recombination in a human population of island Melanesia: correction. Proceedings of the Royal Society of London, Series B: Biological Sciences 267: 1595-1596.

Hull, D. L. (1964). The effect of essentialism on taxonomy - two thousand years of stasis (I). British Journal for the Philosophy of Science 15: 314-326.

Hull, D. L. (1965). The effect of essentialism on taxonomy - two thousand years of stasis (II). British Journal for the Philosophy of Science 16: 1-18.

Hull, D. L. (1997). The ideal species concept - and why we can't get it. In Claridge, M. F., Dawah, H. A. & Wilson, M. R. (eds). Species: the units of biodiversity. Systematics Association Special Volume No. 54. Chapman & Hall, London, pp. 357-380.

de Laubenfels, D. J. (1969). Diagnoses de nouvelles espèces d'Araucariacées de Nouvelle-Calédonie. Travaux du Laboratoire Forestier de Toulouse t. I vol. VIII art. V: 1-2.

de Laubenfels, D. J. (1972). Gymnospermes. In Aubréville, A. & Leroy, J.-F. (eds). Flore de la Nouvelle-Calédonie et Dépendances. 4. Muséum National d'Histoire Naturelle, Paris.

de Laubenfels, D. J. (1979). The species of Agathis (Araucariaceae) of Borneo. Blumea 25: 531-541.

de Laubenfels, D. J. (1988). Araucariaceae. Flora Malesiana I, 10 (3): 419-442.

Levin, D. A. (1979). The nature of plant species. Science 204: 381-384.

Lherminer, P. & Solignac, M. (2000). L'espèce: définitions d'auteurs. Comptes Rendus de l'Académie des Sciences, Paris, Sciences de la vie/Life Sciences 323: 153-165.

Luckow, M. (1995). Species concepts: assumptions, methods, and applications. Systematic Botany 20: 589-605.

Marshall, D. F. & Abbott, R. J. (1982). Polymorphism for outcrossing frequency at the ray floret locus in Senecio vulgaris L. 1. Evidence. Heredity 48: 227-235.

Mayden, R. L. (1997). A hierarchy of species concepts: the denouement in the saga of the species problem. In Claridge, M. F., Dawah, H. A. & Wilson, M. R. (eds). Species: the units of biodiversity. Systematics Association Special Volume No. 54. Chapman & Hall, London, pp. 381-424.

Mayr, E. (1942). Systematics and the origin of species. Columbia University Press, New York.

Mayr, E. (1992). A local flora and the biological species concept. American Journal of Botany 79: 222-238.

McDade, L. A. (1995). Species concepts and problems in practice: insight from botanical monographs. Systematic Botany 20: 606-622.

Nelson, G. & Platnick, N. I. (1981). Systematics and biogeography: cladistics and vicariance. Columbia University Press, New York.

Nixon, K. C. & Wheeler, Q. D. (1990). An amplification of the phylogenetic species concept. Cladistics 6: 211-223.

Nixon, K. C. & Wheeler, Q. D. (1992). Extinction and the origin of species. In Novacek, M. J. & Wheeler, Q. D. (eds). Extinction and phylogeny. Columbia University Press, New York, pp. 119-143. n.v.

Olmstead, R. G. (1995). Species concepts and plesiomorphic species. Systematic Botany 20: 623-630.

Paterson, H. (1985). The recognition species concept. In Vrba, E. S. (ed.) Species and speciation. Transvaal Museum, Pretoria, pp. 21-29.

Platnick, N. I. (2001). From Cladograms to Classifications: the Road to DePhylocode. Oral presentation at Systematics Association AGM, Linnean Society of London, organized by the Systematics Association.

Pleijel, F. (1999). Phylogenetic taxonomy, a farewell to species, and a revision of Heteropodarke (Hesionidae, Polychaeta, Annelida). Systematic Biology 48: 755-789.

Popper, K. R. (1992). The Logic of Scientific Discovery. Routledge, London.

Psillos, S. (1999). Scientific realism: how science tracks truth. Routledge, London.

de Queiroz, K. & Donoghue, M. J. (1990). Phylogenetic systematics, or Nelson's version of cladistics? Cladistics 6: 61-76.

Rieppel, O. (1986). Species as individuals: a review and critique of the argument. Evolutionary Biology 20: 283-317.

Rieppel, O. (1994). Species and history. In Scotland, R. W., Siebert, D. J. & Williams, D. M. (eds). Models in Phylogeny Reconstruction. Systematics Association Special Volume no. 52. Clarendon Press, Oxford, pp. 31-50.

Rosen, D. E. (1979). Fishes from the uplands and intermontane basins of Guatemala: revisionary studies and comparative geography. Bulletin of the American Museum of Natural History 162: 267-376. n.v.

Stace, C. A. (1986). The present and future infraspecific classification of wild plants. In Styles, B. T. (ed.) Infraspecific classification of wild and cultivated plants. The Systematics Association Special Volume No. 29. Clarendon Press, Oxford, pp. 9-20.

van Steenis, C. G. G. J. (1957). Specific and infraspecific delimitation. Flora Malesiana I, 5: clxvii-ccxxxiv.

Templeton, A. R. (1989). The meaning of species and speciation. In Otte, D. & Endler, J. A. (eds). Speciation and Its Consequences. Sinauer, Sunderland, Massachusetts, pp. 3-27.

Van Valen, L. M. (1976). Ecological species, multispecies, and oaks. Taxon 25: 233-239.

Vrana, P. & Wheeler, W. (1992). Individual organisms as terminal entities: laying the species problem to rest. Cladistics 8: 67-72.

Vrba, E. S. (1984). Evolutionary pattern and process in the sister-groups Alcelaphini-Aepycerotini (Mammalia: Bovidae). In Eldredge, N. & Stanley, S. M. (eds). Living Fossils. Springer-Verlag, New York, pp. 62-79 n.v.

Whitmore, T. C. (1998). An introduction to Tropical Rain Forests. 2nd ed. Oxford University Press, Oxford.

Whittemore, A. T. (1993). Species concepts: a reply to Ernst Mayr. Taxon 42: 573-583.