Time as arbiter
A model is only as good as the assumptions and data on which it is based. Here Dr Peter Forey takes issue with earlier contributors who wish to use incomplete stratigraphic data when constructing phylogenies. He argues that these data are better employed as independent checks once a phylogenetic theory has been proposed.
Smith, in his introduction to this debate, asked participants to direct their attention to two questions:
The contributors have so far covered a wide spectrum, from those believing the record is accurate and should be used to reconstruct phylogeny, to those, like me, who totally disagree, believing the record to be distorted and of no use in reconstructing phylogenies.
Stratophenetics
Pearson argues that the stratigraphic record is paramount in attempts to reconstruct phylogeny at any taxonomic level. He studies planktonic foraminifera within deep sea sediments, the cause celebre of the stratophenetic method1 which accepts that phylogeny can be read directly from the rocks and legitimises the recognition of ancestor-descendant lineages.
Stratophenetic studies are traditionally confined to low taxonomic levels (that is species or genera) and to limited depositional basins and stratigraphic time periods2,3 , but both Pearson and Pawlowski/de Vargas set no bounds to the direct use of the fossil record. Central to their arguments is the assertion that the foram record is sufficient in both quantity and quality that we will not be misled. Their answer to Smith's first question is an unqualified 'yes'.
Other authors (for example refs 4,5) have pointed out that even deep sea foram-bearing sediments contain frequent sedimentary hiatuses caused by changing sea levels. Varying amounts of time and species have been winnowed away distorting our ability to recognise extinctions, or discriminate between gradual evolution and punctuated equilibria.
Logical leaps
More generally, and as Siddall and Hedges point out, fossil sampling is markedly affected by rarity6, facies changes, and the environmental preference of animals (and plants). Even acknowledging sampling bias, the simple reading of ancestor-descendent relationships from sequences deemed to contain even a fair sample is flawed because ancestors can never be identified with certainty (see Siddall).
Consider the following thought experiment: given five successively younger strata which bear distinct species, the lowest contains a regular echinoid, the intermediate three each contain irregular echinoids and the youngest a regular. Literal reading would put these together in a simple ancestor-descendant lineage (or perhaps a hypothetical lineage with five cladogenetic events) and explain the reinvention of a regular echinoid as iterative evolution which, according to Pearson, is to be expected. How might we test such a proposition?
Cladistic analysis of morphological and/or molecular data is the obvious way. Pearson and Pawlowski/de Vargas both claim that molecular phylogenies confirm the majority of stratophenetic hypotheses, so it is a little surprising to learn that the latter commentators question the confidence we may have in the use of molecular data!
Likelihood
Other contributors have approached the debate from within a cladistic framework. Marshall and Wagner both argue that stratigraphic data should be admitted to reconstruct phylogeny, but then attempt to define conditions under which that might be possible; they try to answer Smith's second question. They both acknowledge that the morphology/molecular-based, most parsimonious tree, may not be the solution which minimises implied stratigraphic gap.
Marshall invites us to choose the tree that maximises the product of the probabilities from morphological/molecular trees and stratigraphic occurrence. But probabilities of recovering a fair representation of the stratigraphic record are, as yet very poorly developed.
The majority of methods are dependent on theories of relationship7, 8, but some estimate preservation potential9. These latter methods are dependent upon the stratigraphic sections that have actually been sampled and we can have no idea of how representative that sample is of reality. At root, probabilities are just that, estimations of deviation from some perceived truth (the model). They immediately raise the question, what value of probability justifies the inclusion or exclusion of stratigraphic data in phylogenetic analysis?
For Wagner, the truth is centred in models of morphological and molecular evolution as well as models of the fidelity of the stratigraphic record. Wagner emphasises that the reason for reconstructing phylogenies is to understand the tempo and mode of evolution. But model-driven likelihood methods of character change, origination and extinction rates, or models for preservation potential7 , will surely only satisfy the assumptions under which the phylogeny or stratigraphic range was reconstructed in the first place.
Stratigraphy as arbiter
If stratigraphic data, with10 or without11 the baggage of probability, is used at the outset to construct the phylogeny, as advocated here by Marshall and Wagner, then it cannot be used to 'test' the phylogeny. However, if stratigraphy is kept independent from the process of phylogenetic reconstruction then it could be used to arbitrate between competing, equally parsimonious topologies based on morphology and/or molecules12.
Just how effective an arbiter stratigraphy is may well depend on the confidence we have that the occurrence of a taxon in the fossil record represents its real range7,13,14 in both time and space. In the end, stratigraphic information may be all we have left to arbitrate between different equally parsimonious theories based on morphology/molecules.
Resetting the clock
One of the reasons for wishing to incorporate stratigraphic data with phylogenies is to calibrate molecular clocks, although this is by no means necessary15. However, I suggest this can be done more effectively by mapping the stratigraphic occurrence on to a phylogeny and using confidence intervals to specify lower bounds7.
Time is a tricky thing. Unlike morphological (or even molecular) characters, time is not hierarchical, and it is difficult to see how time can provide any grouping or optimality criteria necessary to construct a phylogenetic tree. Time can be surreptitiously excised from a depositional sequence and represented by different amounts of fossil evidence in different parts of the world. Such evidence can itself easily distort the calibration of time by migrations, lack of preservation, or differential extinction. My advice is to leave time as the arbiter.
Peter Forey
References
1. At what level of taxonomic and temporal resolution does the available fossil record provide a true and accurate record?
2. When, if ever, should stratigraphic information be brought into play when reconstructing phylogenetic history?
Natural History Museum, London, UK
