COHEN: ONTOGENY, SYSTEMATICS, PHYLOGENY 



phic) for the group being analyzed are discarded for purposes 

 of generating a phylogenetic classification; only derived char- 

 acters (apomorphic) are of value, and monophyletic groups are 

 defined by the degree to which they share such characters (syn- 

 apomorphy). The distribution of derived character states among 

 a monophyletic assemblage of taxa is analyzed and used to 

 generate an hierarchically arranged chart called a cladogram, in 

 which each node or branching point on the diagram gives rise 

 to two branches that are interpreted as genealogical lineages and 

 are called sister groups. In instances in which the data do not 

 allow the unambiguous definition of two branches, more are 

 often used. Each member of a monophyletic group is more 

 closely related genealogically to other members of its group than 

 it is to members of other groups. More than one cladogram can 

 be generated with the same data set, and the most parsimonious, 

 that is, the one requiring the fewest evolutionary steps, is taken 

 as the most natural or best. According to Panchen ( 1 982), prob- 

 lems in logic invalidate the use of parsimony in cladistics. Not 

 all cladists agree about precisely what a cladogram represents, 

 but some interpret it directly as a phylogenetic classification. 

 One of the greatest problems in using cladistics is the difficulty 

 in evaluating character states for primitiveness or degree of 

 derivation. Two methods have been used; one involves onto- 

 genetic stages and will be discussed later in this paper. A second 

 method, called out-group comparison (Wiley, 1981, gives a good 

 description), is the most subjective part of the entire cladistic 

 procedure and to a certain degree may involve circular reason- 

 ing. A practical problem that cladistics has not yet conquered 

 is that of naming, for classifications must be used by many who 

 have no interest in theory, and naming categories on a strictly 

 genealogical basis raises many problems, as does the practice 

 followed by some cladists of naming all branching points. Some 

 attributes of ELH stages that might be considered unsuitable 

 for use in evolutionary classification are available for use in 

 cladistics. One example concerns character stages that are in- 

 terpreted as being highly adaptive rather than conservative. If 

 polarity can be ascertained, then so-called adaptive characters 

 are available. Rates and sequences of ontogenetic change also 

 constitute potentially valuable character suites. 



The third method, presently called evolutionary classification, 

 is more difficult to define and discuss. It has a long history and 

 an extensive literature (Mayr, 1981). The methods of evolu- 

 tionary classification are eclectic and generally more subjective 

 than those of phenetics and cladistics. They do not easily lend 

 themselves to overall generalization. Characters are selected and 

 weighted by paying particular attention to homology and con- 

 vergence; to the extent that they are available, evidence from 

 embryology and palaeontology are also used. Primitive char- 

 acters are admitted to the system. Data are used from ecolog- 

 ically oriented facets of evolution such as selection, competition, 

 predation, and ecological biogeography. Historical biogeogra- 

 phy, rate of evolution, and genetics are also considered. An 

 hierarchical classification is derived, which has an inferred time 

 axis and which may generally reflect genealogical relationships. 

 However, degree of phenetic difference in selected characters, 

 which is interpreted as reflecting degree of genetic difference, 

 may be considered along with branching pattern in converting 

 a strict genealogy into a classification. Patterson (1981b) has 

 discussed and criticized such procedure. Whatever may be phy- 

 letic relationships, the definition of taxa is essentially subjective, 

 and each member of a group is not necessarily more closely 



related genealogically to other members of its group than it is 

 to members of a different group. The test for goodness of a 

 classification is pragmatic; if it has high predictive value it is 

 good. (By prediction is meant the degree to which a classification 

 encompasses additional data.) In commenting on evolutionary 

 systematics Panchen (1982) writes that it, "has always been 

 somewhat ad hoc in its procedure, yielding good results with 

 competent taxonomists and bad with incompetent ones. The 

 standard warks [sic] on procedure . . . are to some extent ra- 

 tionalizations of a tradition that is too largely intuitive." 



As a summary, I have tried to compare in Table 1 some of 

 the techniques, objectives, and assumptions of the three meth- 

 ods. Phenetics requires the fewest assumptions but would seem 

 to offer the systematist a classification with the least information 

 value. Cladistics has the most constraints, so many and so strin- 

 gent in fact, that they may limit its practical use, although the 

 method is particularly valuable in indicating areas for which 

 additional or more suitable data are required. Misuse of cla- 

 distics may soon rival the long-time abuse by systematists of 

 parametric statistics. Evolutionary classification tries to include 

 the most information from the most sources, but the methods 

 for doing so are not very well formalized. Cladists treat their 

 method of classification as a general theory of biology (Nelson 

 and Platnick, 1981), a forcing function among all evolutionary 

 phenomena, which must therefore comply with a parsimonious 

 model derived entirely from character state analysis. Evolu- 

 tionary classification, on the other hand, incorporates infor- 

 mation from a wide variety of biological phenomena and to 

 that extent is forced, rather than forcing. Predictability, as a test 

 of goodness for a classification, is more pragmatic and logically 

 less satisfying than is parsimony. Perhaps an important question 

 for theoretical systematists to consider is the formulation of 

 comparable definitions for replicability, parsimony, and pre- 

 dictability. 



Ontogeny and Fish Phylogeny 



Louis Agassiz, who fought the idea of organic evolution, pro- 

 posed a "threefold parallelism" of arranging organisms in a 

 series or classification. His three parallels were palaeontology, 

 what we would now consider to be homology, and ontogeny. 

 Even though he failed to interpret the parallels as evidence for 

 evolution, his keen perception of the fact that they do exist in 

 nature and are somehow interrelated has elicited extensive com- 

 ment and reinterpretation (see especially Gould, 1977) and is a 

 suitable point of departure for addressing the importance of 

 ontogeny as a source of information about homology, the bio- 

 genetic law, developmental stages as alternatives to outgroup 

 comparisons in cladistics, paedomorphosis, and the application 

 of life history stages to phylogenetic inquiry. 



If characters are the meat and muscle of classification, then 

 homology surely shapes the skeleton on which phylogenetic clas- 

 sifications are arranged. The worth of any allegedly phylogenetic 

 classification is no better than the degree to which homology 

 has been assessed, and how to do this is a major problem for 

 the systematist. Like the weather, everyone talks about homol- 

 ogy but does nothing about it— or almost nothing. The concept, 

 which is so pervasive in the study of phylogeny and in evolution, 

 has been with us since pre-Darwinian times, although not always 

 in the way that we understand it today. The great comparative 

 anatomist Owen defined it in 1866 as follows; "A 'homologue' 



