Fossil calibration

A question for the list:

The rapidly increasing number of sequenced genomes allows an
increasingly precise determination of the shape of the tree of life.
Dating the deep divergence times between higher taxonomic levels,
however, still seems to be a serious problem, as the "molecular clock"
almost never seems to run very clock-like. A calibration by means of
well-dated fossils is required for every evolutionary divergence point,
with the big problem being a run-away propagation of statistical and
systematic errors.

In a recent paper, Graur D., Martin W., "Reading the entrails of
chickens: molecular timescales of evolution and the illusion of
precision", Trends in Genetics 20 (2004), 80-86, wrote (their abstract):

"For almost a decade now, a team of molecular evolutionists has produced
a plethora of seemingly precise molecular clock estimates for divergence
events ranging from the speciation of cats and dogs to lineage
separations that might have occurred ~4 billion years ago. Because the
appearance of accuracy has an irresistible allure, non-specialists
frequently treat these estimates as factual. In this article, we show
that all of these divergence-time estimates were generated through
improper methodology on the basis of a single calibration point that has
been unjustly denuded of error. The illusion of precision was achieved
mainly through the conversion of statistical estimates (which by
definition possess standard errors, ranges and confidence intervals)
into errorless numbers. By employing such techniques successively, the
time estimates of even the most ancient divergence events were made to
look deceptively precise. For example, on the basis of just 15 genes,
the arthropod-nematode divergence event was 'calculated' to have
occurred 1167 +- 83 million years ago (i.e. with a 95% confidence
interval of ~350 million years). Were calibration and derivation
uncertainties taken into proper consideration, the 95% confidence
interval would have turned out to be at least 40 times longer (~14.2
billion years)."

One of the papers they reference is Hedges S.B., Kumar S., "Genomic
clocks and evolutionary timescales", Trends in Genetics 19 (2003),
200-206, which displays, in Fig.6, 35 divergence times, "calibrated" by
fossils, between 5.4 and 3970 Myr ago. Among these, only 3 are given as
ranges: 82-83 Myr for cattle vs. carnivores, 88-90 Myr for primates vs.
rabbits, and 90-98 Myr for primates vs. cattle. All the other numbers
are apparently errorless.

I had come across some interesting data earlier in a paper by
Aris-Brosou S., Yang Z., "Bayesian models of episodic evolution support
a Late Precambrian explosive diversification of the metazoa", Molecular
Biology and Evolution 20 (2003), 1947-1954. They estimate the
protostome-deuterostome divergence at 582 +- 112 MYA (million years
ago), "broadly consistent with the fossil records". Under the heading
"Materials and Methods - Calibration Points", they write (in part): "To
reduce errors associated with calibration points, only fossil-based
dates were used, as in Bromham et al. (1998) (in) (??):
Collembola-Pterygota, 390 MYA; Aranaea-Scorpionida, 405 MYA;
Coelacanth-Dipnoi/Tetrapoda, 418 MYA; Osteichthyes-Dipnoi/Tetrapoda, 428
MYA; Asteroidea-Echinoidea, 500 MYA; Agnatha-Gnathostoma, 510 MYA;
Arachnida-Merostomata, 520 MYA; Cephalochordata-Chordata, 530 MYA."
Their reference is:

Bromham L., Rambaut A., Fortey R., Cooper A., Penny D., "Testing the
Cambrian explosion hypothesis by using a molecular dating technique",
Proceedings of the National Academy of Sciences USA 95 (1998),
12386-12389. Quite conventionally, Bromham et al.'s "Methods" section
starts as follows:

"We used two data sets: (i) protein-coding mitochondrial DNA, taken from
whole mitochondrial genome sequences (5,676 bp) and (ii) nuclear-encoded
18S rRNA (1,710 bp). Sequences were aligned by eye and any saturated
regions that could not be confidently aligned were removed. The list of
sequences and alignments are (sic) available on:
http://evolve.zoo.ox.ac.uk./Alignments/Cambrian.html. These sequences
were then combined into dated pairs using fossil dates taken from the
literature."

At this point, the expected literature reference is replaced by a
footnote reading in part: "The fossil dates used to form pairs of
sequences were (in millions of years before present):
Diptera-Hymenoptera, 240; Diptera/Hymenoptera-Orthoptera, 300; ..." (10
more range- and error-less numbers up to 530 for
Cephalochordate-Chordate follow).

As I am interested in such divergence date estimates, I wrote to
Bromham, the corresponding author, in an email, asking him for the
missing references (using his more recent email address found in Trends
Ecol.Evol. 18 (2003), 2). That was 3 weeks ago; to date, I haven't yet
received any answer.

I wonder whether Graur and Martin describe a general problem, or whether
they are just pointing to some excesses. How do you evaluate the
situation?

Even more, however, I would be interested in some reliable fossil dates
(including realistic error margins and references to the primary
literature) which could reasonably be used as bounds for divergence time
estimates with respect to higher taxonomic levels.

Peter

-- 
Dr. Peter Ruest, CH-3148 Lanzenhaeusern, Switzerland
<pruest@dplanet.ch> - Biochemistry - Creation and evolution
"..the work which God created to evolve it" (Genesis 2:3)