Do Fossils Show a Worldwide Record of Evolution?

The fossil record provides the acid test for evolutionary theory.  Everyone who walks a real dog by a poodle knows that small-scale variation occurs among living species, but non-evolutionists get understandably annoyed when Darwinians extrapolate the observed variations to encompass all of life: as if to say, because finch beaks vary, therefore humans had bacteria ancestors.  Darwin’s bold hypothesis connected all living things into a branching tree of life.  He claimed that, ultimately, whales and oaks and kangaroos and seashells could trace their ancestry to single-celled organisms.  The only way to connect this hypothesis to actual earth history is to examine the fossil record.  Does the record of the rocks show a sequence of life evolving from simple to complex?
    Those who assume so might be disturbed by a paper in the Annual Review of Earth and Planetary Sciences¹ by Peter M. Sadler (UC Riverside).  The annual reviews are a good place to catch up on the state of the art of this or that discipline.  Sadler’s review concerns quantitative biostratigraphy, the attempt to correlate global fossil data.  Things are looking up in this field; fossil data are becoming more available in large databases, and computers are making the number-crunching easier.  He takes the reader through the latest computer algorithms that attempt to correlate fossils from tens, hundreds, or thousands of sites around the world into a unified, global time sequence.  Though his lengthy paper never questions evolution (and hardly mentions it), and while written with a tone of scholarly confidence, it gives a distinct impression that biostratigraphy is more art than science.
    Imagine an ideal record where everything that had died left a fossil, and these fossils accumulated upward, layer upon layer, since the beginning of life.  If evolution had occurred, each species would have a first appearance in the record (a first-appearance datum, or FAD), and when it went extinct, it would exhibit a last-appearance datum, or LAD.  These “horizons” would form a vertical timeline for each species, which could be correlated with similar ones around the world.  Assume it were also possible to reliably date each layer.  Tracing the history of life, then, would be a piece of cake; actually, a layer cake, because the layers would preserve a clear sequence, from oldest at the bottom, to youngest at the top.  The fossils they contain, if evolution had occurred, would clearly exhibit increasing complexity as each new phylum, order, class, genus and species appeared through time.
    Alas, as with most things in life, the situation is far from being so simple.  Sadler points out a number of difficulties that make global correlation of fossil-bearing strata a challenge:

• Imbalance:  Most of the record consists of seashells.  “Richly fossiliferous sections are more common in the marine invertebrate record,” he notes.  (Marine invertebrates actually comprise about 95% of all known fossils.  That means all the large mammals, land plants and dinosaurs make up a tiny fraction of the record).  In a few studies, he claims, biostratigraphers can produce sequences of some marine invertebrates to resolutions of 10,000 to 50,000 years, though resolution is usually much lower.

• Gaps:  “Relative to marine Cenozoic correlation problems, nonmarine instances suffer from a lack of continuous sections,” he says.  Instead of showing a continuous record of evolution, the record is discontinuous or jerky, riddled with gaps.  (Stephen Jay Gould once remarked that the near universal presence of gaps in the fossil record is the “trade secret” of paleontology.)  Many sites display “isolated faunas” that do not overlap with other sites.  Sadler explains how the gaps affect biostratigraphy:

Biology demands that the global abundance of a species cannot fall to zero within its temporal range.  Unfortunately, species distributions are patchy, the patches may shift, few individuals are fossilized, and fossils may be overlooked.  Consequently, the local taxon ranges observed in single stratigraphic sections reflect local conditions and do include gaps.  More critically, and for the same reasons, gaps of unknown extent occur at the ends of observed ranges.  Thus, local horizons of highest and lowest finds of a species do not correspond to the global FAD and LAD.  The discrepancies vary from place to place, and locally observed taxon range charts contradict one another in detail concerning the sequence of range-end events.

For these and other reasons, Sadler warns that it is “crucial to acknowledge that local first and last appearances are also uniquely troublesome as recorders of calendar events: The local stratigraphic horizons at which they are observed do not reliably reproduce the true global sequence of origination and extinction events.  Discrepancies must be expected because local appearances and disappearances are likely to be migration events and probably displaced by lapses in fossilization.”

• Reworking:  Many fossils have been transported or reworked, destroying the temporal sequence information.  (Some of the best-known fossil sites, such as Dinosaur National Monument, La Brea Tar Pits, and Petrified Forest present this difficulty.)  The biostratigrapher cannot assume the apparent FAD-LAD horizons represent the true history of the fossils, because many processes can disrupt the correlation of fossils with strata: floods can transport fossils from one location to another, burrowing animals can rework the deposits, or deposits can fall into a cave or be washed down well bores into older strata.  Moreoever, it is not always easy to tell when or how much reworking has occurred.  “Severe caving may require abandoning FADs altogether,” he says.  Marine microfossils are especially subject to reworking.  The sometimes “cryptic” signatures of reworking may go “unrecognized,” and their impact on the record may be significant.  Yet the biostratigrapher needs to rely on databases that are contaminated by this problem: “Large integrated databases will combine taxa that are prone to reworking with those that are not.  Decisions about the likelihood of reworking, or the most palatable assumptions concerning reworking, currently force a dichotomous choice between methods that seek maximal ranges and those that seek probable ranges.  No method yet embodies a satisfactory theory of reworking that can obviate this unfortunate choice,” he laments, yet the computer models often assume that little or no reworking occurred.

• Decreasing Information with Age:  The farther back in time, the less reliable the inputs: for instance, “Paleozoic instances include less radiometric, paleomagnetic, and stable isotopic data.”  The known instances usually do not overlap.  “The large Paleozoic correlation problem in Table 1 includes many pairs of sections that do not overlap in age.  They must be stacked in the correct order and impart to the problem a significant component of seriation.  Seriation is the essence of the problem when the data are isolated faunas.”

Considering these difficulties, is it even possible to produce a global correlation of fossils into a time sequence?  Sadler apparently feels the problem is tractable and current work is promising, but the use of simplifying assumptions is unavoidable.  Some are reasonable (e.g., a FAD must precede its LAD, and proven coexistences must be honored).  Also, certain geological events provide a means of independently correlating fossiliferous strata.  A volcanic ash fall, for instance, might be traceable across a large region, or magnetic reversals or global climate changes can provide clues.  In addition, paleontologists try to hitch the data to milestones obtained via radiometric dating (although these are usually not applicable to the sedimentary strata that contain fossils).  Putting it all together is easier said than done:

The way to improve the resolving power of the geologic calendar is obvious but not easy—increase the number of events and thus reduce the average time intervals between them.  There is no shortage of species to add.  The real problem is to keep all the appearance and extinction events in their correct sequence.  The difficulty increases dramatically with the number of species for three reasons: First, the number of possible sequences of appearance and extinction events grows faster than exponentially as a function of the number of species (Figure 1).  Also, events that are separated by smaller time intervals are more likely to be preserved in contradictory order from place to place.  Finally, as the list of species grows it must include more provincial organisms that will be missing from many locations.

    The bulk of Sadler’s paper concerns various clever mathematical algorithms biostratigraphers have developed to approach this huge puzzle.  Some make use of the principles of operations research.  Some employ heuristic algorithms or manipulate matrices with iterative processes to try to converge on a solution.  Each method is best suited to its own data type, each makes its own assumptions, and each has its shortcomings.  Consequently, he cautions the reader not to expect too much:

The true global sequence of FADs and LADs is not knowable in detail and the locally preserved sequences of highest and lowest finds are incomplete and contradictory.  The practical and tractable problem is to find a hypothetical sequence of FADs and LADs that enjoys the lowest net misfit with all observations in local range charts and isolated faunas, or requires the smallest net adjustment of all observed ranges.  It is an optimization problem.

Sadler freely admits that contradictions are inevitable.  Much of his paper concerns dealing with misfits: how to measure misfits, and how to minimize them.  Some of these misfits are those that contradict the expectations of evolution.  One of the criteria for success seems to be how well the result of an algorithm agrees with the “correct” phylogenetic sequence: “Procedures for fitting the best LOC [line of correlation on the graph] include deterministic regression techniques … and heuristic search algorithms from evolutionary programming,” he explains.  Congruence with evolutionary phylogeny seems to define Sadler’s “best-fit” or “optimal” sequences.  In the opening, he indicates that evolutionary sequence information takes priority over geological dating information:

Geologic time correlation proceeds by constructing a global calendar of past events in which the appearances and extinctions of fossil species dominate the entries.  Other events include changes in ocean chemistry, reversals of Earth’s magnetic field, and the deposition of volcanic ash beds, some of them dated by radiometric methods.  The challenge is to merge incomplete inventories of physical events and partly contradictory faunal successions from many local thickness scales (measured stratigraphic sections) onto a single calendar that correctly sequences all the events and scales the time intervals between them.  Because correctly sequenced events serve the purpose of correlation, with or without knowledge of their numerical ages, sequencing is the fundamental task and the focus of this review.  Numerical estimates of age are available for very few events, especially in the older periods of the Phanerozoic.  Furthermore, estimates of the relative size of time intervals between events rest largely upon questionable assumptions about rates of sediment accumulation and biological turnover.  Consequently, scaling and calibration tasks are best attempted after the optimal sequence of events has been determined.

In the conclusion, titled “The Remaining Challenges,” Sadler reveals his discipline’s dependence on evolutionary theory, and drops hints that it needs to be more of a two-way street:

Paleobiologists can extract considerable information about the phylogenetic sequence of taxa by analyzing the morphology of fossils, without recourse to stratigraphic information.  But these insights do not yet aid the correlation task as much as they might.  To date, more effort has been committed to questions concerning the place of stratigraphic information in cladistic analyses of morphology than to the possibility that the resulting cladograms provide independent evidence of sequence that can improve biostratigraphy.

How this avoids circular reasoning he does not explain.  Instead, he suggests how evolutionary systematists can help – by revealing, for instance, “the order of FADs that best fits the morphologic information.”  But even with their assistance, he sees three “looming challenges” posed by modern stratigraphic databases:

1. Deciding on a single method:  “First it is desirable to integrate more data types into a single method.  Every method, regardless of the data to which it is suited, must seek a sequence of events.  Consequently, the best way to suit all the data is to invert the problem, working through a suite of permutable sequences and achieving iterative improvements as judged by the fit between the sequences and the data.”
2. Speed vs. Completeness:  “But the second challenge is to manage considerably larger data sets without loss of speed.  The flexibility of the inverse approach sacrifices speed.  The fastest algorithms are those that are tailored to specific data types and work forward from the data to the best solution.
3. Reworked fossils.  As quoted above, “No method yet embodies a satisfactory theory of reworking that can obviate this unfortunate choice” between maximal ranges and probability ranges (that is, choosing between incorporating all the data into the model vs. using the data that produce the expected result).

.  Are biostratigraphers stuck in a rut?  He ends, “As in the past, answers to all these challenges might be discovered by recognizing analogies with problems in other disciplines and adapting their numerical methods.”

¹Peter M. Sadler, “Quantitative Biostratigraphy: Achieving Finer Resolution in Global Correlation,” Annual Review of Earth and Planetary Sciences, May 2004, Vol. 32, pp. 187-213 (doi:10.1146/annurev.earth.32.101802.120428).

It must be acknowledged that Sadler neither doubts evolution nor intended to cast doubt on evolution in this paper.  A casual reading would lead one to think that everything is fine and the Darwinians are making great progress.  But, if read perceptively, without evolutionary assumptions, it is quite revealing.  Where is the proof of the pudding?  Where is the evidence in the fossil record to prove Charlie right?  Sadler exposes to view what a huge “optimization” problem he has on his hands.  The best he can do is try to keep the “contradictions” and “misfits” to a minimum.
    As with everything else in evolutionary theory, the tweak space is greater than the data space.  Only massive inputs of questionable assumptions keep the story intact.  A story of evolution clearly doesn’t jump out of the data, as if it were an intuitively obvious fact that only an obscurantist would deny.  No; instead of supplying the Darwin Party with the proof they desire, he needs to ask them for help as he stumbles through a contradictory, unmanageable, confusing, formidable task.  It’s reminiscent of the impossible dream the molecular phylogenists face trying to keep Charlie’s imagined tree of life connected to reality (see 07/25/2002 and 06/13/2003 headlines).  In the end, they must assume evolution to prove evolution.  Instead of taking the evidence where it leads, they apply similar heuristic “optimization” approaches to handling overwhelming and contradictory inputs, where “optimal” means “mostly agrees with Charlie, if we neglect the misfits.”
    Notice that “gap” is a loaded word.  What if it is a brute fact that the data are discontinuous?  Then that is the true sequence; there are no gaps.  A gap is only a gap if you assume evolution.  Why not face the evidence squarely: living taxa are discontinuous, and fossil taxa are discontinuous.  They appeared abruptly, and some died abruptly.  If it weren’t that such an admission destroys Darwinism, that would be what the textbooks would matter-of-factly present.
    Skeptical readers are encouraged to put aside “questionable assumptions” about “rates of sediment accumulation and biological turnover,” and to study this article without Darwin-tinted glasses on.  Look at the fossil data as objectively as possible.  What is found?  Multitudes of non-overlapping “isolated faunas” without clear “seriation” information.  A preponderance of seashells.  Unknown effects of reworking.  Fossil graveyards.  Myriads of dead organisms buried in water-laid rock strata all over the world.  Sadler suggests a solution in his ending sentence; biostratigraphers might have better success by looking outside the box and adapting the techniques of other disciplines.  Most likely he did not intend to consider some disciplines that the ruling Darwin Party has placed off limits.  Too bad; what if that’s where the true solution is waiting to be found?