Hi Randall,

modern molecular divergence time estimation methods no longer make the strict clock assumption. Instead, rates are allowed to vary across branches (so-called relaxed clock models). Concerning the fossils, we need them because without external calibration it is impossible to estimate absolute ages from sequence divergence (because sequence divergence is the product of rate and time). How accurate are these methods? If you ask me, nobody really knows. For a definite answer, you would need a time machine (but if you had one you wouldn't have to rely on estimates...).

Hi Randall,

modern molecular divergence time estimation methods no longer make the strict clock assumption. Instead, rates are allowed to vary across branches (so-called relaxed clock models). Concerning the fossils, we need them because without external calibration it is impossible to estimate absolute ages from sequence divergence (because sequence divergence is the product of rate and time). How accurate are these methods? If you ask me, nobody really knows. For a definite answer, you would need a time machine (but if you had one you wouldn't have to rely on estimates...).

Thanks Martin. Your answer was helpful, yet expected. I have heard about relaxed clock models. Unfortunately, they don't improve my confidence in molecular estimates of divergences. Do you (or anybody on RG) recommend a paper or papers that candidly discuss the problems with node dating using sequence divergence in combination with fossil data?

Many thanks!

Maybe have a look at these recent reviews:

Bromham, L., Duchêne, S., Hua, X., Ritchie, A.M., Duchêne, D.A., Ho, S.Y.W. 2018. Bayesian molecular dating: opening up the black box. Biol. Rev. 93:1165–1191.

Donoghue, P.C.J., Yang, Z. 2016. The evolution of methods for establishing evolutionary timescales. Phil. Trans. R. Soc. B 371:20160020.

O’Reilly, J.E., Donoghue, P.C.J. 2016. Tips and nodes are complementary not competing approaches to the calibration of molecular clocks. Biol. Lett. 12:20150975.

Much of molecular dating methodology is nothing more than a big scam. There are two issues – whether there is a molecular clock across lineages, and whether molecular divergences are actual/maximal or minimal. The first is problematic and personally I do not regard the assumption as valid for dating taxa.

The second is the real scandal as molecular clocks are usually calibrated using fossils and to a lesser extent island ages. The calibrated ages presented as actual or maximal (often using priors). This presentation amounts to the Big Lie (große Lüge) because such calibrations are actually minimum estimates since the calibration points are themselves minimums. The setting of priors does not change that since priors are only a guess of the upper limits which have no empirical basis.

The result is a whole vast literature setting divergence ages and rejecting any possibility of earlier origins, to the extent of supposedly falsifying earlier origins. It's all quite vast and seemingly beyond question. But it all falls on the simple fact of minimums - a minimum is a minimum is a minimum.

The only other calibrations source that does not have the same weakness is tectonic calibration. Of course that may generate older dates which might be distasteful to many. The ages of some tectonic calibrations must also be treated with care depending on the degree of precisions involved and potential reactivation of old tectonic formations. But at least tectonic correlation is an estimate not a minimum.

Some papers:

Heads, M. (2005) Dating nodes on molecular phylogenies: a critique of molecular biogeography. Cladistics, 21, 62–78. https://doi.org/10.1111/j.1096-0031.2005.00052.x

Heads, M. (2012) Molecular Panbiogeography of the Tropics. University of California Press, Berkeley, 562 pp.

https://doi.org/10.1525/california/9780520271968.001.0001

Heads, M. (2017) Metapopulation vicariance explains old endemics on young volcanic islands. Cladistics, 2017, 1–20. https://doi.org/10.1111/cla.12204

I think also a difference should be made between accuracy and precision. Some divergence estimates (particularly older ones) are deceivingly precise, but not accurate (as the true divergence time does not correspond with it). The key is accuracy rather than precision - although we would like both. At least we need to be sure that the true divergence estimates lies within the range of estimates and i think that for groups with a reasonable fossil record and when methods are properly used (taking reasonable settings - not take default settings for granted), they are at least more accurate than just taking the fossil record or molecular clock estimates without critically choosing reasonable parameters at face value.

Heads raises some good points, but just using alternative calibrations like tectonic events is just as problematic or not more precise or accurate than using fossil calibrations:

http://rstb.royalsocietypublishing.org/content/371/1699/20160098

Furthermore, most paleontologists and evolutionary biologists are well aware of the issues of using fossils as maximum constraints - therefore they speak of soft maximum constraints and hard minimum constraints. Nevertheless, in groups with a reasonable fossil record using older sites with similar environmental and preservations where this particular group should be preserved, but is not preserved could still be informative. It is true that it is a bit strange that a method which was originally thought to be able to work without fossils, is so critically dependent on them. However, even here progress has been made to test the impact of fossil preservation on molecular clock estimates:

http://rspb.royalsocietypublishing.org/content/284/1857/20170227

Kenneth,

No one ever said that fossils should not be taken into account. Certainly not by me or any other panbiogeographer. So not sure where that came from.

Point is that a minimum is a minimum is a minimum. Maximum constraints whether soft or hard is still just inspired guesswork.

Heads (2017) noted the following regarding De Baets et al:

How should priors be selected? This is controversial, and “judgement of the degree to which fossil minima approximate divergence timing . . . could be considered a dark art . . . ” (De Baets et al., 2016, p. 1).

De Baets et al. (2016, p. 1) discussed the use of tectonic features to date clades and wrote: “Fossils only really provide minimum clade age constraints. In their place, phylogenetic trees can be calibrated by precisely dated geological events that have shaped biogeography. . . Biogeographic calibrations are no panacea for the shortcomings of fossil calibrations, but their associated uncertainties can be accommodated . . . Biogeographic and fossil calibrations are complementary, not competing, approaches to constraining molecular clock analyses."

The DeBaets paper does not invalidate tectonic approaches, although it does make some problematic assertions and so should be read critically. For example they assert "Advocates of biogeographic calibration argue that tectonic calibrations are more accurate and reliable than fossil-based calibrations because they can directly evidence both maximum and minimum constraints on the age of lineage divergence events." They don't say who makes that claim (about accuracy), but it is not necessarily true for all who have used the method of tectonic correlation. The difference is between assessing an actual divergence date (whether narrow or prolonged) or a minimum.

Similarly, no citations are given for the assertion that "most studies assume implicitly that geological constraints on the timing of lineage divergence are equivalent in all organisms, ignoring taxon-specific differences in their environmental tolerance, ecological requirements and dispersal ability." Panbiogeography has never ignored any other evidence. Interesting that the last point is resurrected - the gold standard of 'dispersal ability', even though purported dispersal ability has little correlation with the range of taxa (i.e. good and poor dispersers show the same kinds of biogeographic patterns).

Their main point, with which I would agree, is that some tectonic processes have an extended time of activity, certainly such as the opening of the Atlantic basin. In such cases there is definitely greater potential uncertainty with a tectonic correlation, but this is about approximating actual dates rather than just misrepresenting minimums as maximums (whether 'hard' or 'soft'). In a study of Macaronesia I pointed to the range of ages involved with the Atlantic and for that reason focused on the date of final separation, although event hat is an approximation of about 100 Ma. In the north Atlantic there are later estimates that would certainly be relevant for separated taxa in that region (and final dates for that separation have also been contentious).

So, to repeat, presenting minimums as maximums is voodoo science. I am in full agreement with De Baets et al that tectonic correlation is not without its challenges with respect to the precision of dating. But the essential point is that fossil calibrated ages do not falsify the possibility of earlier dates suggested by older tectonic dated calibrations. Assertions to the contrary also represent another example of the große Lüge.

Just to be fair and not restrict myself to criticism of molecular clock applications, I will mention that I have a forthcoming article in Zootaxa which illustrates an application of tectonic correlation for a group of moths in New Zealand and Australia as well as some regional perspectives. The article addresses issues concerning molecular calibration, but does not ignore them. Instead, molecular calibrations are explicitly recognized as minimums - for example "Molecular divergence ages calculated for the Lepidoptera are [at least] 200–208 Ma (Wahlberg et al.2013; Misof et al. 2014), and for the Prototheoridae and Hepialidae [at least] 65–40 Ma (Wahlberg et al. 2013), or [at least] 110 Ma for Hepialidae as represented by the genus Triodia (Misof et al. 2014). These dates allow for existence of extant exoporian genera by at least the end of the early Cretaceous and do not preclude the possibility of earlier origins."

I have a proof copy of the article which is near enough OK to distribute at this point should anyone be interested - send me an email at [email protected]

A copy of the abstract is below:

The biogeographic history of Exoporia (Lepidoptera) in the Southwest Pacific is reconstructed for genera and species that show distributional boundaries corresponding to tectonic structures in the region. Correlations with tectonic formations of Mesozoic origin such as the Whitsunday Volcanic Province and Otway-Bass-Gippsland Basin system in Australia, the Vitiaz Fracture Zone in northern Melanesia, and the Western Province Eastern Province boundary, Waitaki Fault Zone, and Waihemo Fault Zone of New Zealand are presented as evidence of an East Gondwana origin for genera and species before the geological separation of Australia and New Zealand. The correlated boundaries also suggest that many extant species retain at least parts of their original East Gondwana distribution ranges. The presence of Exoporia on the northern Melanesian Arc, New Caledonia, and New Zealand is attributed to the tectonic isolation of these areas when East Gondwana expanded into the Pacific following retreat of the Pacific Plate subduction zone. Local endemism of Mnesarchaeidae in New Zealand is interpreted as the result of an original vicariance from a widespread ancestor (‘Exoporia’) resulting in two allopatric descendants —a narrowly distributed Mnesarchoidea and a widely distributed Hepialoidea. The current overlap of these two groups in New Zealand is explained as the result of subsequent range expansion by the Hepialoidea prior to geological fragmentation of East Gondwana. The potential impact of Cretaceous geography on modern distributions is also considered for Exoporia in southern Africa and northern America. Along with lateral displacement of Exoporia, tectonic processes also contributed to the origin of high elevation endemics through a process of passive tectonic uplift.

Thank you to all who have been contributing to this thread. The discussion has greatly assisted me in searching for patterns among phylogeographic analyses.

Dear John,

Thanks for clearing this up and so we principly agree - i misunderstood your original comments concerning the fossils. I having nothing against Pangeography, so not sure where that came from either.

As for the use of fossils, there have papers who advocate the use of tectonic calibrations in favor of fossil calibrations (these are listed in our paper). In De Baets et al. (2016) we wanted to demonstrate the biogeographic or tectonic calibrations also underly uncertainties which need to be properly considered - some of similar to fossils, other are different. The claim that no citations are given for the assertion that most studies implicitly assume that geological constraints on the timing of lineage divergence are equivalent in all organisms, ignoring taxon-specific differences in their environmental tolerance, ecological requirements and dispersal ability is far-fetched. Our assesment comes from a spreadsheet we compiled from studies using biogeographic calibrations. Most of the ones we found implemented tectonic calibrations in molecular clocks in the same way for flying birds, plants or frogs so implicitly ignoring their differences in environmental tolerance, ecological requirements and dispersal ability which i found quite shocking. Instead we opted to cite some of these studies in our paper in the respective boxes as the point was not criticize this work, but to highlight difference in the implementation which need to be considered.

As for the maxima, De Baets et al. (2016) never stated that fossil provide hard maxima (the problem is that you need to provide a maximum to run a all molecular clock analyses in the first place) nor that fossil calibrations provide no uncertainties as the paper explicitly states that both underly uncertainty - some similar, others not and was the main point of our analysis.

I agree with you that finding proper use of calibrations without running into circularity when testing hypotheses is necessary and not a straightforward endeavor and using fancy methods without informed knowledge on potential accuracy will also not help.

I am therefore very interested in work which addressed them or at least discusses them. Could you send me your articles which relate to it?

Kenneth,

Many thanks for your additional comments and insights. Most appreciated.

I mentioned panbiogeography since it is by far the most extensive application of tectonic correlation in the literature. But no worries.

Thanks for your explanation about whether or not articles ‘ignore’ environmental tolerance, ecological requirements and dispersal ability. My response was in the context of such qualities being empirically demonstrated as effecting a distribution and then being ignored. I know that there are a LOT of theoretical propositions about environmental tolerance, ecological requirements, and dispersal ability going back to Darwin's time, but in their application to the origin of biogeographic differentiation [which is what I work on] there is nothing of empirical biogeographic content to be addressed (at least from the literature I have seen).

Years ago Joseph Hooker proposed geological connections to account for biogeographic connections but then questioned this perspective when it was shown that snails can survive for considerable time in sea water. Of course this demonstration actually said nothing empirical about biogeographic history and rather it might be the other way around, that biogeography informs about the historical significance of ecological parameters (e.g. how is theorized dispersal ability to be assessed?).

In my recent ghost moth study I noted that the group evidently had ‘poor’ means of dispersal. But that was just an observation of current ecology. It was not used to say that this was evidence that the moths present in Australia and New Zealand had to originate by vicariance. That someone might believe that pregnant moths made their way from Australia to New Zealand (the usual story) is something that I could not directly falsify. But the tectonic correlations are evidence that such ecological events do not need to be invoked. They are simply irrelevant (or course in my opinion for whatever any opinion is worth).

I was referring to hard and soft maxima being used in molecular approaches, but OK that you did not take that view.

A common theme I find in reaction to tectonic calibrations is the idea that certain dates are just too old to be believable. For example, the idea that the major primate groups were already present in the Mesozoic. People say “oh no, that cannot be true. There are no fossils”. Same for angiosperms. And yet there are hundreds of documented patterns in these groups that correlate with Mesozoic tectonics. I have no problem with this being questioned through analysis, but if it is just “oh no” then not much to say.

Another aspect of tectonic correlation is that these do not necessarily rest on a single tectonic feature or on a single part of the distribution. For example, one may consider breaks across the Atlantic or Indian Ocean for primates, but also the details of distribution within.

I find your paper important for drawing attention to tectonic correlations to be considered with care. I especially agree with some of the geological models – such as whether the Panama Isthmus is relevant or not to differentiation of taxa either side (Heads discusses that and other examples). And of course stratigraphic correlations are highly problematic as it seems that old life can occupy young landscapes (this model being introduced by Heads 1990).

The most extensive global applications of tectonic correlation are to be found in the books by Michael Heads. I see that you have seen his first book and so well aware of that material. But he has since published two more books also listed below.

Heads, M. 2012. Molecular panbiogeography of the tropics. University of California Press, Berkeley.

Heads, M. 2014. Biogeography of Australasia: A molecular analysis. Cambridge University Press

Heads, M. 2017. Biogeography and Evolution in New Zealand. CRC Press.

A while back I published individual application of tectonic correlation for the origin of hominid distribution and a copy is here appended. This paper aroused great hostility for both its phylogenetic reconstruction (which questioned molecular wisdom) and the biogeography. I will send you a copy of the Zootaxa ghost moth paper separately. Also attached here, my study of the biogeography of Macaronesia. This is an example where tectonic correlations are more open ended given the complexity of the tectonic history of the region, but at least gets away from the common assumption of chance dispersal to explain allopatry and phylogeny.

Thanks for your interest in the relationship between tectonic correlation and molecular methods. Most molecular systematists (at least those works I have seen) dismiss the method as lunacy (or words to that effect).

I guess nonsensical is in the mind of the beholder! Is it more nonsensical than to mis-represent minimum molecular dates as actual or maximal? Is it more nonsensical than to misrepresent priors as some sort of objective boundary? Is it more nonsensical than to assert that DNA sequence similarity is the absolute arbiter of phylogenetic relationship? Whether or not the orangutan perspective is considered nonsensical or not is neither here nor there. What is important, as in any science is the nature of the evidence presented and how that evidence may be assessed by whatever criteria, whether for or against. I would remind readers of the lesson of Barbara McClintock .

Artur,

Saying that DNA sequence similarity is the "best arbiter" pretty well makes it absolute. But that aside, the point is that claiming that molecular sequence is "the best"s simply propaganda. Maybe it is in many or most instances, but that does not mean always. Saying that you "don't see how" we can question and insist that Pongo and Homo have a more recent common ancestor than Homo and Pan is not informative scientifically. We lay out our reasoning in the paper. If there is anything in that you wish to dispute I am happy to discuss further.

Propaganda in the sense that just claiming that DNA trees are necessarily better does not necessarily mean that they are.

You are free to prefer to trust the sequence tree. I have no problem with that. Everyone is free to make their choice. I never criticize personal preferences.

But just saying that one prefers a molecular tree in no way gives any empirical credence to the morphogenetic tree necessarily being wrong.

The funniest thing about all this is that the earliest unambiguous fossil hominds not only look more like orangutans than chimps, but share a number of apomorphies as well.

Not so, since we do not claim that morphogenetic evidence is necessarily better. We express a preference for the morphogenetic evidence, but that is just our preference.

Understood. You follow what has been called by some in systematics 'the law of large numbers" - a bit like the principle of phenetics, that more characters is more important than the kind of characters.

Sampling of ancestors (now fossil) would not change my outlook since I consider that there are potential issues in molecular approaches that can (but not always) generate erroneous results.

What I do conclude is that when molecular and morphogenetic trees are incongruent there is an anomaly, particularly where the morphogenetic evidence is very robust (unequivocal characters, and especially lots of them [at least 28 or so for humans and orangutans, a bit like law of large numbers again, but restricted to apomorphies]) for which there is currently no objective criterion for making a choice.

I appreciate your level headed responses. I do not say that to be patronizing. Over the years I have become used to quite nasty responses from some molecular and non molecular biologists.

One of the classic paradoxes of the molecular/morphogenetic dilemma is illustrated by biologists such as David Pilbeam who on one hand decided that morphology was totally unreliable phylogenetically (since molecules proved it wrong for human-hominoid relationships) but on the other hand it is totally reliable when he works on fossils and makes declarations about their phylogenetic status. Even the Red Queen could not have done better.

Cheers, John

This gets a bit complicated. Not sure how to answer the question as it seems to have some assumptions that I might not fully understand. We have suggested that the nature of derived character states in the Hennigian sense are problematic in molecular studies. There are a couple of inherent issues. One is alignment (which is used all the time in hominoid molecular systematics - at least when I last worked on this) which is a phenetic measure of similarity that creates matches that do not exist in nature. The other is the the informative states appear to be created as a result of the preferred tree. In morphological cladistics the derived states are individually identified (by outgroup comparison) and the tree building only involves finding the best matching cluster for those derived states. In this understanding there is no necessary reason why the morphological (morphogenetic) trees will necessarily match all the time, and potentially why molecular trees may at least sometimes given an answer that is more phenetic than cladistic, even if using cladistic terminology and techniques.

Hi Artur,

Comments below

"assuming no problems with alignment"

Since alignment is used so often it is a real problem. I take it that you do regard these approaches as invalid (or there would be no need to remove them)

“ so lets assume…..that we do have a perfect set of genes which align without any doubt”

Certainly would be a better situation.

“ How come it is OK for morphogenetic tree building to use a priori outgroup but it is not OK for molecular tree to infer informative states along with the tree?”

In the morphogenetic approach (at least the one I used) one is restricting analysis to putative apomorphies. In the molecular approach one is not doing that when using the tree building to identify apomorphies after the tree is built (as a result of the tree). In my view it is problematic for molecular approaches that cannot individually identify apomorphic states before the analysis. Of course that is just my personal opinion and carries no authority.

“Besides, there are many ways to reconstruct phylogenetic tree from an alignment and while they all have weaknesses I don't think this particular critique applies to all of them.”

OK. But I recall that it did to most if not all the hominid studies I examined, but it has been some years and so might be wrong about that.

“ It is possible to build a molecular tree with a fixed outgroup for example”

Yes, but I refer to use of outgroup to individually identify apomorphic states prior to analysis.

“ Moreover, you can always take the same algorithm you use for morphology and apply it to molecular characters.”

Meaning one can apply parsimony and other tree building algorithms. True, one can.

“ assume we …. avoid all known pitfalls …. the molecular tree … fully supported …Would you expect such a tree to be congruent with your morphogenetic tree.”

One could have that expectation. Whether or not that works out is another matter. If it does not then there is still an unresolved anomaly.

Anomalies are always interesting. Calls for deeper investigation. At least in theory. In practice, with hominid systematics, of course the choice is to prefer the molecular results, problems notwithstanding and further, regard the human-chimp clade as proven and even have the status of a scientific fact.

Cheers, John

Yes, that is one of alternatives. Either the morphology has generated homoplasies or the sequences have generated homoplasies.

Interesting, to me at least, is that even if the derived morphological similarities of humans and orangs are homoplasies, I have noticed a great reluctance among primatologists to even talk about them (every time in the mirror - face has four - receded hairline from birth, forward facing cranial hair, fully developed beard and mustache in males, same external ear structure). Even our reproductive biology shares much in common including mating biology and physiology. And of course again to fossil australopiths that have uniquely orang features of the skull. All very intriguing wherever one stands on this matter.

Sometimes these discussions are transformed into a dialog between two respondents going off on tangents. Perhaps these should be directed to another forum within RG.Practical answers are lost in this sea of side-issues.

As Lindell Bronham pointed out years ago, molecular clocks have tremendous limitations but they have been the best that we have and generally do give insights into evolutionary theory.

And since Bronhm made that statement, relaxed clocks and new fossils and new dating techniques have provided far better estimates of dates.

Part of the problem is that the statistical distribution best describing nucleotide substitution is the Poisson (actually even a compound Poisson). The trouble is that the Poisson mean equals the Poisson variance so that as we go back in time, the confidence interval is increasingly large. So we can have high confidence in a date but the precision becomes so minimal that the CI - which seldom includes the model variance - is rarely reported.

The point is that even if all other variables fit perfectly around the mean (no variation around the mean), the very nature of the evolutionary history - a Poisson process - means that the further one goes back in time, the wider the interval means that precision is not possible.

Nevertheless, even though tree inference is fraught with departures from model assumptions they are usually pretty darn good at describing an epistemologically unknowable past. The same is true of divergence dates. One simply must avoid the hubris of certainty in reporting these dates.

Patrice,

Some comments on your observations below.

“And since Bronhm made that statement, relaxed clocks and new fossils and new dating techniques have provided far better estimates of dates.”

I would argue that estimates are not 'far better' if they (authors) still fail to admit that that such estimates are minimums only.

“they are usually pretty darn good at describing an epistemologically unknowable past.”

How does one know that they are usually ‘pretty darn good’?

“One simply must avoid the hubris of certainty in reporting these dates.”

More than that, one much avoid the Große Lüge that the estimates are anything other than minimums – whether or not supposed distribution probabilities, priors etc are tacked on.

John Grehan

I have provided a link to an article I wrote addressing this issue among others. Admittedly the relevant methods have advanced considerably in the decade since I wrote the article. Nonetheless, some of the points I make still stand. One can measure mutation rates directly with deep-rooted pedigrees, and (lo and behold) they seldom match the deep-time estimates informed by a variety of ad hoc historical assumptions. When particular migration scenarios are taken as given, then the assumptions themselves are used to calibrate the rates (to some extent), and tautologies are possible, where expectations inform results and validate the expectations...

Article A New Perspective on Later Migration(s)The Possible Recent O...

Hello again Randall!

The relevant section of my 2008 article linked above is footnote 3, pp. 274-275, dealing with a single example pertaining to NRY divergence estimates. What I called RPS4Y-T(using Lell et. al's 2002 nomenclature) is now conventionally called C3b. I think the point still stands. Whether or not that particular Y chromosome is considered "paleo-American" or a product of a later Na-Dene language migration from Asia l depends largely on a slew of historical assumptions constraining our interpretations. Thus this conventional wisdom might be derived directly from molecular divergence estimates, but it looks to me that the estimates are more often selected in an ad hoc way to support the author's favored view (single vs. multiple migration scenarios for the peopling of the Americas).

Really all we can do is "the best that we can do". Nevertheless any "best" estimate should acknowledge the large confidence interval around the estimate. Even if the data really fits a perfect clock - an it rarely does - and supposing that the fossil date is known perfectly (impossible), the CI will still be increasingly large as the tree goes back in time. This is simply because the stochastic process describing substitutions is, at best, Poisson distributed so that the variance equals the mean (as the the mean age goesf the tree increases, the variance increases as well). Lots and lots of useful lterature has presented methods for relaxing the clock assumptions and much of it shows that a simple Poisson process is a poor model and that a compound Poisson is more realistic although the variance is even larger. So do date a tree but always realize that precision is unobtainable even with perfect data and a perfect model.

Keeping in mind that the ranges are just guesses around a minimum fossil calibrated date (i.e. the confidence interval itself can only provide an upper minimum since it is constructed around a minimum).

Use as many fossil calibration dates as possible. All methods seem to require both old and young calibration dates which lie both inside and outside the clade of interest; if these conditions aren't met, the dates will be systematically distorted.

Unfortunately, this requires knowing the fossil record of the group in question quite well, to make sure you aren't falling prey to outdated identifications or simply overlooking newly discovered fossils. You cannot just copy calibration dates out of a table. Many molecular timetree studies contain errors with a large impact of just this kind.

The fossil record cannot provide hard maxima. Sometimes it's good enough to provide soft maxima, however. These should be used whenever possible, not only because many methods require a maximum at the root, but also because including only minima systematically makes all dates too old.

For references, my own practice, a few calibration dates I have proposed and discussion of them, please see my 2007 paper and my 2013 Hist. Biol. paper. Not included there are the effects of body size and mass extinctions on molecular divergence-date estimates, on which I recommend the papers attached here.

John Grehan, what do you mean by "alignment is used so often"? If you don't align molecular data (whether before your phylogenetic analysis or during it, as POY does), you don't know what your characters are – which states belong to which character.

And why do you claim that molecular dates are minima? They are actual dates, not minima, unless of course there's something wrong with a particular analysis. In fact, they're much more often too old than too young unless you're willing to make extremely implausible assumptions about the fossil record and about paleobiogeography.

In the morphogenetic approach (at least the one I used) one is restricting analysis to putative apomorphies.

What is this approach, then? In all phylogenetic analyses I've seen, all characters have at least two states, not just one; which state is plesiomorphic is determined by the outgroup unless you've got an asymmetric stepmatrix in some character.

Oh, another comment for John Grehan: you wrote "Sampling of ancestors (now fossil)" – very, very few fossils are ancestors of anything alive today, or even of each other!

I do agree with that in the sense that once cannot say an individual fossil is ta 'ancestor', only its inclusion or relationship to an extant clade at some particular taxonomic level. So when one finds, for example, a fossil that is said to have all the features that would cladistically define it as a placental mammal, this does not mean that it is 'ancestral' in any direct sense, just that it is part of that clade. And of course the age of that fossil, if the oldest age known for any placental mammal fossil, does not directly indicate whether it is the oldest possible fossil or the upper limit for the possible age of origin of the placental mammal clade.