Pelmatza Eleutherza

Figure 2.17. A partial cladogram of the Echinodermata, including some fossil groups. HO = Holothuroids; EC = Echinoids; O = Ophiuroids; A = Asteroids; ED = Edrioasteroids; CY = Cystoids; CR = Crinoids; HE = Helicoplacoids; CA = Carpoids. • = extinct (after Smith 1984).

nation of character-state polarity, or, at most, that younger fossils are more derived than older ones and ancestry can therefore be established. Because we know that ancient groups often survive along with their descendants, or at least descendants of the ancient groups' close relatives, the latter assumption is often false. The existence of the living fossils Neopilina and Latimeria argues against the infallibility of strati-graphic position as an indicator of ancestor-descendant relationships. One can even imagine hypothetical cases in which the fossil occurrence of a given descendant taxon antedates its ancestors, because the latter group is missing from a portion of the fossil record (Eldredge and Cracraft 1980). The closer the origins of the two taxa and the longer the period of coexistence, the higher is the probability that sequence can be misinterpreted. This is not true, however, for direct phyletic sequences. If a descendant derives by transformation from an ancestor, it is highly probable that stratigraphic sequence indicates polarity (Paul 1982). The probability of two randomly collected specimens being preserved in the wrong order cannot be greater than one half.

The strong limitations of temporal fossil occurrence in evolutionary reconstruction are especially clear when paleontologists search for the ancestor of a large tax-onomic group. For example, the bivalves Babinka (McAlester 1965) and Fordilla troyensis (Pojeta, Runnegar, and Kriz 1973) have both been cited as transitional forms, which are ancestral to the mollusk class Bivalvia. In both cases, early strati-graphic occurrence is a principal part of the argument, though morphology also plays a role.

In the case of Babinka, a series of muscle scars were linked with the hypothetical ancestral states of the commonly cited likeness of Neopilina to the hypothetical ancestral mollusk (McAlester 1965). The anatomical claim was refuted by showing that the pedal muscle scar pattern in Babinka was not homologous to the serially repeated pedal muscle scars in Neopilina (see Stanley 1972, p. 166). The repeated "gill muscle scars" in Babinka most probably did not represent a transitional change between ancestors and later bivalves, where a single pair remains. Note that although the claim for ancestry depended on both stratigraphic position and character states, the refutation was based on an analysis of character states alone.

The case of Fordilla troyensis is more illuminating. This remarkable fossil is widespread in the Lower Cambrian rocks of North America and can also be found in the same Series in Denmark and perhaps England, Portugal, and Siberia (Pojeta and Runnegar 1974). Because nearly all of the bivalve superclasses were not found in rocks older than Middle Ordovician time, and only the Palaeotaxodonta appear in the Early Ordovician, Fordilla deserved its status in 1973 as the most ancient fossil bivalve yet discovered. There was a gap of some 40 million years between it and the beginning of the known Early Ordovician geographically widespread bivalve occurrences.

The unique fossil finds in New York State (Pojeta et al. 1973) permit reconstruction of internal scars, which have been used to establish its likely bivalve molluscan status. Individuals of the species appeared to have elongate, subequal adductors and a broadly inserted pallial line. The shape of the shell and position of muscle scars suggest a shallow-burrowing suspension feeder. These features were used by Pojeta and Runnegar (1974) to make a case for direct ancestry of the Bivalvia, via the group of Ordovician heteroconch families best represented by the Cycloconchidae. After speculating on the genealogical relationships between this group of families and the other bivalves, a tenuous link was even claimed between Fordilla and the univalved Cambrian rostroconchs.

The problem with this sort of reasoning is obvious. What if another lower Cambrian bivalve is discovered that harbors a set of character states completely different from Fordilla? Because the molluscan affinities of Fordilla troyensis were only recently appreciated (Pojeta et al. 1973), one can safely expect that some other group, now known too poorly to rise from the ranks of incertae cedis, will materialize soon as a competing ancestor. The preemptive claim made by Pojeta for this genus was based solely on stratigraphic position. There was no reason to believe, from any other evidence other than stratigraphic occurrence, that the character states borne by Fordilla troyensis were necessarily ancestral. There might have been a large and diverse bivalve fauna in Early Cambrian time that has gone unnoticed or unpreserved. This is not outlandish, given the 40-million-year span between Fordilla and later bivalve occurrences.

As it turns out, another, still older, bivalve mollusk was discovered in Early Cambrian rocks of South Australia by Peter Jell (1980) and was reverently named Pojetaia runnegari (Figure 2.18). A later morphological analysis with well-preserved specimens (Runnegar and Bentley 1983) established clear similarities between this form and the Palaeotaxodonta, a group often thought to be an ancestral bivalve subclass. This discovery only emphasizes the great potential for further discoveries in Early Cambrian rocks and the dangers of searching for ancestors by means of stratigraphic position.

This example is unfair, perhaps, as an indictment of the use of stratigraphic order to infer phylogenetic relationships. After all, the inferences here were clouded by poor preservation and the general difficulties of inferring relationships in the Cambrian, the time perhaps when the bivalve groups were beginning to diverge.

The study of evolutionary transitions between the fishes and the tetrapods has been similarly influenced by the fossil record and by overall similarity between puta-

Figure 2.18. A bivalve from the Early Cambrian of South Australia. (A) Pojetaia runnegari, phosphatic coat of right valve; (B) P. runnegari, near-sagittal section. (From Bengtson et al. 1990, with permission.)

Figure 2.18. A bivalve from the Early Cambrian of South Australia. (A) Pojetaia runnegari, phosphatic coat of right valve; (B) P. runnegari, near-sagittal section. (From Bengtson et al. 1990, with permission.)

tive ancestors and the descendant tetrapods. The Rhipidistia have been traditionally thought to be the tetrapod ancestors, on the basis of overall similarity in the skull roof, appendages, and appropriate stratigraphic position. In particular, the presence of paired internal nostrils (choanae) has been cited as a linking character. Rosen, Forey, Gardiner, and Patterson (1981) claimed that the interpretation of this character is incorrect and that the rhipidistian Eusthenopteron lacks choanae. By contrast, a restudy of a Devonian lungfish from Australia suggests the presence of choanae. Thus, the restudy of characters placed the Dipnoi (lungfish) as the sister group of the tetrapods and completely changed our conception of vertebrate phylogeny. Rosen et al. noted that overall similarity and the connection by stratigraphic proximity led us astray. Whether this interpretation is correct or not, it places the onus on paleontologists to avoid stratigraphic assessments of ancestry and classifications based on overall similarity, which might involve grouping with ancestral characters.

Ancestors aside, temporal sequence in fossil occurrence provides useful genealogical information (Fortey and Jefferies 1982; Harper 1976; Paul 1982). In some cases, as noted above, closely spaced samples reveal a gradational sequence of morphological change from ancestor to descendant, with no evidence of cladogenesis. Although one can never exclude the possibility of something happening "between the lines," such studies, common in deep-oceanic sediments in groups such as foraminifera (e.g., Bettenstaedt 1962; Grabert 1959; Malmgren and Kennett 1981), can rightfully justify temporal sequence as evidence for character polarity.

Temporal sequence may be a corroborative tool to strengthen a hypothesis of genealogy (Eldredge and Cracraft 1980, p. 58; Miyazaki and Mickevich 1982). Consider the following analysis. A systematist establishes a cladogram, which is rooted on the basis of an outgroup comparison or by an assumption of character state polarity using ontogenetic change. This analysis yields a cladogram showing an array of taxa that can be arranged from near the most ancestral state to most derived. If the strati-graphic order of the taxa occurs in the order "predicted" by the cladistic analysis, then the conclusion of character polarity based on character analysis alone is strengthened.

An example of this sort comes from the work of Miyazaki (Miyazaki and Mickevich 1982) on the evolution of the Miocene-Pliocene scallop genus, Chesapecten, preserved in basins in the eastern coastal plain of the United States. On the basis of ontogenetic change, the cladogram in Figure 2.19 infers a genealogy and roots the tree near the taxon with the most "juvenile" features as an adult. The cladogram is closely concordant with stratigraphic order. The cladogram and stratigraphic occurrence data can be properly considered as independent sources of evidence leading to a similar conclusion of descent.

Temporal sequence may also resolve vexing cases of convergence when other approaches fail. The two Cenozoic radiations of planktonic foraminifera demonstrate the difficulty of identifying particular morphs without good stratigraphic information (Cifelli 1969). A Paleocene radiation from globigerinid ancestors resulted in a morphologically diverse array of taxa, nearly all of which disappeared by the Oligocene. A second radiation in the Oligocene-Miocene repeated many of the species in such faithful similarity to those in the first radiation that a proper systematic assignment is impossible without the appropriate stratigraphic information.

Permission For Caminalcules
Figure 2.19. Cladogram for the Miocene-Pliocene Atlantic coastal plain scallop Chesapecten, superposed on the stratigraphic sequence. (From Miyazaki and Mickevich 1982, with permission.)

Stratigraphic and paleogeographic position may also be used to a degree to resolve problems in character reversal and parallel evolution. In some cases, identical derived character states may be acquired independently within several mono-phyletic groups, when independent stratigraphic and geographic evidence isolates these groups from each other. In the ancestral trigoniacean bivalves, stratigraphic position is essential in genealogical reconstructions (Newell and Boyd 1975). The hypothesis that parallel evolution has occurred produces the most corroborated hypothesis of genealogy. Although one can only speculate on the cause of such parallelism, commonality of ground plans might result in a similar response of independent groups to an environmental change. In any case, this approach parallels the use of gene duplication events (e.g., Goodman et al. 1982) as likely devices to increase the consistency of molecular evolutionary lineages.

It might be argued that forams and mollusks are so simple that constraints will often lead to parallel or iterative evolution. One might expect repeated morphologies here but not in higher organisms such as mammals. The widespread occurrence of atavistic character states in vertebrates (e.g., Riedl 1978) makes this claim highly unlikely. In vertebrate jaws, developmental fields can be defined where correlations among characters are stronger than with characters in other putative fields (Kurten 1953). Thus, ancestral character states often reappear in lineages as coordinated complexes. Consider the case of Lynx lynx, whose fossil record has been studied extensively (e.g., Werdelin 1981). In some Pleistocene and Recent specimens, a coordinated appearance of the M2 molar and a postcarnassial element derives from ancestors where the condition is completely absent (Kurten 1963). Indeed, the M2 molar has been lost in the Felidae since the Miocene! For an unknown reason, the characters have reappeared with noticeable frequency and with sufficient morphological complexity that they can be regarded as a character reversal toward an ancestral condition.

Despite the caveats concerning the influence of errors of stratigraphic position, evolutionary radiations, and incomplete preservation, it may surprise the skeptical reader just how good stratigraphic position can be in recording correctly the order of stratigraphic relationships. The simplest test, node-order correlation, would be to calculate a correlation between the order of first appearances of fossil groups with the order of nodes in a cladogram, deduced from morphological data alone. To do this in a simple way, complex cladograms are usually collapsed to a pectinate form so that order of nodes in a cladogram and geological occurrence can be related directly. A test of this sort demonstrates for many groups a significant correlation between node position and order of first appearance in the fossil record (Gauthier, Kluge, and Rowe 1988; Norell and Novacek 1992a,b; Sereno et al. 1999).

Stratigraphic consistency is another and perhaps more powerful means of comparing the record of fossil appearances with the order in a cladogram (Clyde and Fisher 1997; Fisher 1994, Huelsenbeck 1994). A node in a cladogram is consistent with the rock record if the stratigraphic first occurrences of the taxa above it are younger or equal in age to the node below. Therefore, an inconsistent node has stratigraphically older taxa that are placed in more derived nodes. This comparison can be readily done with intact cladograms, which need not be reduced to pectinate form. The index is simply the number of stratigraphically consistent nodes divided by the total number of relevant nodes (the root node cannot be tested). Hitchin and Benton (1997) tested hundreds of cladograms of fossil echinoderms, fishes, and continental tetrapods and found a high degree of stratigraphic consistency for all groups, with a mode at about 0.75% consistency. Echinoderms appear to be the best. Stratigraphic consistency was well correlated with the node-order correlation measure mentioned above. Interestingly, neither the node-order correlation nor the stratigraphic consistency measures were significantly correlated with the degree of stratigraphic completeness, measured by discrepancies of first appearances of sister taxa (e.g., Benton and Hitchin 1996).

This apparent success has led to the new field of stratocladistics. Like conventional cladistics, stratocladistics relies on parsimony and attempts to minimize ad hoc hypotheses of homoplasy and failure of preservation of fossil lineages in inter vals that contain fossils. An interesting advantage of using temporal data is the possibility of identifying ancestors and even connecting lineages of fossil taxa. As we have discussed above, cladistics logically excludes such a possibility. Stratocladistic hypotheses of relationships attempt to minimize two types of ad hoc hypotheses: homoplasy and temporal order. The former is discussed above. The latter simply involves minimizing cases in which a more ancestral taxon appears in the record after the first appearance of a derived taxon.

A simple analysis is illustrated in Figure 2.20. The phylogenetic tree on the left is consistent with the character data, but the derived taxon A "skips" two time intervals before appearing at stratigraphic level four, even though B and C are preserved in those levels (two and three). The tree on the right requires no such ad hoc hypotheses of nonpreservation, but taxon A must have two character reversals relative to its immediate ancestor, taxon B. Thus, these two trees are equivalent.

Does stratigraphic order perform significantly worse than characters in constructing a cladogram? To test this question, we need a measure that applies to both trees. Clyde and Fisher (1997) used a derivative of the retention index (Farris 1989) to compare trees from 29 published data sets, mainly from fossil vertebrates. The retention index measures the degree to which homoplasy must be invoked for a given cladogram of relationships based on parsimony. Therefore, a cladogram whose characters are completely consistent (all characters would individually produce the same tree) has a perfect index of 1.0. Clyde and Fisher (1997) compared

Minimizing Homoplasy

Minimizing ad hoc Preservation Gaps s ms ar rg o

"O

+1 0

Post a comment