Bayes, evolutionary clocks, and biogeography 30 Mar 20122 Apr 2012 I just received a review by Gareth Nelson of Michael Heads’ book Molecular Panbiogeography of the Tropics (publishers’ site). I should have blogged this before, since I got a copy, being on the editorial board for this series (the same one I published with at Uni Calif Press), but I have been bogged down with personal troubles and other tasks. Also, I am not professionally competent to make much comment, so Gary’s review, forthcoming in Systematic Biology, is welcome. However, the book and the review set me thinking about how historical inferences in biology are made. We use dating to establish how evolution occurred – whether, for example, a particular species is the ancestor of another, or when a particular trait evolved. These inferences are ubiquitous and somewhat doctrinaire. But there is even a website and app (TimeTree) that uses techniques like molecular clocks to absolutely date divergence times, and which draws inferences from it (such as that a particular species is not an ancestral species because it is found too late or too early). Molecular clocks are not so reliable as is sometimes presumed, however. Gary quotes this episode: Here I am reminded of a scientific meeting in London (5 September 2001; published by Donoghue & Smith 2003). Francisco Ayala presented a paper on “Molecular clocks: whence and whither?” An audience member then asked: “Francisco, is there a molecular clock”? He gave a complex reply. The same member asked again: “Francisco, is there a molecular clock?” This time he gave a one-word answer: “No.” Stratigraphy is another example. The earliest date given for a species in the fossil record is usually the date of the oldest specimen found, but as a well known book, Systematics and the Fossil Record, notes, species may persist well outside of that record (“Lazarus taxa“), or be misidentified as persisting when they haven’t (giving rise to “Elvis taxa“, my favourite technical term). At best we have relative dates, with some degree of credence in the absolute dates, and little if any direct knowledge of the path of actual evolution. This sets up a problem or two. One of these is the problem of calibration: how do we match these relative dates with absolute time? Heads’ solution is to use plate tectonics: The method of dating used in this book does not assume an evolutionary clock, even a relaxed one. Instead it fits multiple tectonic events (rather than multiple fossils) to a phylogeny. This indicates a chronology in which rates can show extreme changes within and among lineages and genes at different times and places. [p71] Gary notes that this gives us at best a minimum age, and he and Heads both dispute that this will give us confidence in a maximum age of taxa. And he and Heads both note the problems with the “dispersal” method (or as it is better called in my view, a “presumption”) in trying to find the centres of origin, both in time and space, of taxa, instead adopting the “vicariance” method (“presumption”) in which the known distribution of taxa gives the “tracks” of a taxon, along which the origin may be said to be, either some location or across the whole track. Given the recent work on human origins and the multiple dispersal model interbreeding claims with existing species in Asia and Europe, this is a present concern. I leave readers to find Gary’s review and read it, and then to read at least the first two introductory discussion chapters in Heads’ book (which at nearly 100 pages is almost a book in itself). Right now, however, I want to make a point about historical inference in the natural sciences (and indeed in any domain). One of the besetting issues in the philosophy of history is “historicism”, which is the view that we know the natures of things when we know their origins. This is reflected in the title of a book, The Poverty of Historicism by Karl Popper. It is regarded by historians as a sin, along the lines of Whiggism and the Great Man theory. In this context we might ask whether we indeed do know biological taxa from knowing their origins at all. It seems to many that biological taxa are known independently of the explanations and origin stories given in evolutionary contexts; for example the discovery that the Leopard Frog (Rana pipiens) was several species was based, not on evolutionary considerations but on a knowledge of how frogs made mating calls. Evolution provided the explanation after that discovery. But the temptation towards historicism is inevitable (Ankersmit 2010); we do think things are known by their origins, for all that this is a fallacy of reasoning (the so-called genetic fallacy). And even more than that, we insist upon trying to find these origins, in time and space, and nowhere more than in evolutionary biology. In fact, one of the modern philosophical views on what species are relies entirely upon their origins (Griffiths’ 1999 historical essentialist account). Here’s the problem I have here, and it is one that was raised many times in the history of post-evolutionary biology, in particular by the so-called Pattern Cladists (e.g., Nelson and Platnick 1981): if we do not know the history, how can we evaluate claims made based on the evolutionary process? That is, if we don’t already know how evolution in general proceeds (such as assumptions of molecular clocks or rates of change), how do we evaluate the claim that a particular lineage evolved in the absence of direct evidence? And how do we learn how evolution proceeds in the first place? This circularity can be called the Creationist Objection, in a rather restricted sense: not that evolution does not happen, but that we do not know how it happens directly, but have to rely upon a “methodological creationism” (a term of Paul Griffiths’), in which the evidence takes priority over inferences, and in which inferences cannot be used as evidence. It’s an ironic take on the common creationist objection to anything that contradicts the account of Genesis: “were you there?” [Answer: well yes, in the relevant sense – observation – we were there.] The Creationist Objection is based on an extreme empiricism; the view that evidence is only measurement and observation. Consequently, assumptions of molecular clocks is not evidence (but radioactive decay rates are, because they are fundamental laws of nature, and this argument applies only to historical observations). It is my view that the historicism of modern evolutionary biology is based on the faulty assumption that we can treat some prior hypotheses as Bayesian priors to estimate the likelihoods of some process occurring and evaluate claims of origins. Now Bayesianism is a well-established theory of knowledge, and in one sense we are all Bayesians now, but that doesn’t imply we are actually using Bayes theorem; just that we know that we can’t make most inferences without assuming some prior hypotheses and likelihoods. However, when we do that, we are hostages to the correctness of those prior estimates. If we screw them up, all our subsequent reasoning will also screw up. So evidence should always trump our priors, and we need to do a kind of reality check from time to time, which is why Heads’ approach has merit; tectonics is a distinct domain from biology, and so the assumptions are relatively independent of our biases of evolutionary theory and narratives. I once asked Gary, as a callow youth (well, I was still just a doctoral student; I was aged even then), whether he thought we could get some idea of ancestors, a question motivated by the refrain among some pattern cladists that ancestors are unknowable, and his answer surprised me. He replied that it was something we could have various degrees of confidence in, but yes. We could know ancestors just to the extent that we had evidence of various kinds. I recall thinking this was a kind of consilience view. It’s just that we cannot observe ancestors, and so from a purely empirical perspective, one has to represent them as “sister taxa” to their putative descendants in a cladogram. Today this is universally understood (although at no time have the process cladists conceded that they agree with the pattern cladists now, so far as I know). The question is not how we represent them or what the evidence actually indicates, but how we take evidence to guide inference. Back in the 18th century, naturalists often debated what counted as “scientific reasoning” and many of them followed Lockean principles that knowledge is only ever got from observation, through the senses. This extreme empiricism led rather directly to positivism, and thence to logical positivism. The alternative view is something one might call a kind of Kantianism: one can only observe in the light of some theory. Empiricism and Kantianism are ever at war in the philosophy of science, and in science itself. Depending on what you require of inference, either we cannot know ancestors, events and centres of origin in the past, or we certainly (or almost certainly) do know these things although we have yet to uncover some events. Or, you are somewhere in the middle there. This mediate view is a more common position than the rhetoric of the debate might indicate. Generally in science, when one side makes claims at one extreme and their opponents take the other extreme, you can get them to qualify and qualify until they approach each other in the middle and only words separate them (a point once made by Stephen Toulmin in 1970). Most pattern cladists and panbiogeographers will admit that you can at least estimate the ancestors and origins; most evolutionary biologists and dispersalist biogeographers will admit that no, we don’t know for sure, but that these are our best guesses. Almost nobody is either a true methodological creationist or darwinian fundamentalist. And this is how it ought to be; science is about warrantable inferences, not doctrine. However, loose talk sinks inferences. We often assume that what we hypothesise is what we actually know, when instead we are multiplying uncertainty upon uncertainty, a problem with Bayesian inference. And yet… Bayes rules. It works well. We manage to navigate the world by assuming that what we hypothesise is true, so unless the world is simply bizarre and perverse, something about it must be right. I think that the solution is to presume there is no solution. Both approaches – Locke and Kant – are necessary in a population of scientists to ensure both that neither the opposition to theory so common in the 18th century nor the total theory-dependence of various language-centric philosophies is in control of science, as neither is sufficient for science to proceed. We do different things when we observe than when we make inferences based on hypotheses. The only problem arises when we confuse and conflate these two things. Theories are not evidence, and evidence is not explanation. The temptation of historicism is a kind of Kantianism, but Kant was not wrong, merely incomplete. Science requires that we are both empiricists ands theorists and that neither is dispensed with, nor takes on the role of the other in inference. The origins of species and traits in evolution are interesting topics, and we should try to work them out. But not by ignoring or trimming evidence. Ankersmit, Frank. 2010. The Necessity of Historicism. Journal of the Philosophy of History 4 (2):226-240. Griffiths, Paul E. 1999. Squaring the circle: Natural kinds with historical essences. In Species, New interdisciplinary essays, edited by R. A. Wilson. Cambridge, MA: Bradford/MIT Press:209-228. Nelson, Gareth J., and Norman I. Platnick. 1981. Systematics and biogeography: cladistics and vicariance. New York: Columbia University Press. Toulmin, Stephen. 1970. Does the distinction between normal and revolutionary science hold water? In Criticism and the Growth of Knowledge, edited by I. Lakatos and I. Musgrave. Cambridge UK: Cambridge University Press:39–48. Epistemology Evolution Natural Classification Philosophy Science Species and systematics Systematics
Epistemology Philosophy as forgetting, and index characters 13 Nov 2009 I was talking to a friend, Damian Cox, yesterday, and we were discussing how many of the ideas of, say, a Wittgenstein had been a rediscovery or reformulation of what had been commonly held over a century before. Damian made the comment that philosophy is a process of forgetting what… Read More
Administrative Competition: copy of my species book 4 Oct 201127 Oct 2011 My book Species: A History of the Idea is soon to be available in paperback. I have a few copies I’d like to share, so put your name in the comments and make sure your email is filled out (it doesn’t get shown publicly) and I will roll some virtual… Read More
Evolution Myths 2: The origin of species 12 Feb 2009 So today, which is in the antipodes (we being so far ahead of you northern western types) the 200th birthday of an obscure British naturalist gentleman, we address this myth: Myth 2: Darwin did not explain the origin of species in The Origin of Species Here’s some folk claiming just… Read More
Your general points towards the end are good. But, I have to say that I don’t see much utility here in the 21st century of the pattern cladist approach and the application of similar approaches in biogeography. Pattern cladism was perhaps useful as a corrective to the old-guard “evolutionary systematics” school of Ernst Mayr & co, where classification was based in part on presumed fossil ancestors, on complex hypotheses about “adaptive zones” and the like, and where pleisomorphies and synapomorphies were often not distinguished. Clearly it is better to estimate relationships with as much data and as few unchecked assumptions as possible. But nowadays, I’m afraid, Heads, etc. are themselves the rigid old guard. They don’t like probablistic models (ML/Bayesian approaches), they don’t like branch lengths, they don’t like estimating dates on trees, they don’t like ancestors, and they don’t like dispersal. And they publish articles and books to this effect. Meanwhile, though, thousands of us are happily using the newer methods and actually *testing*, rather than just baldly assuming, ideas like whether or not a particular divergence event correlates with a plate tectonic event. The one thing that is pretty universal is the “no ancestors in phylogenies” dogma, but that one is also doomed sooner or later. It is easier to write software where you assume every specimen/OTU is a tip and all you have to estimate is the branches connecting them (instead of adding in the possibility that each specimen is an ancestor somewhere). But it is simple to simulate a realistic tree, simulate some plausible fossil sampling, and show that the chance of getting a fossil which is in a lineage directly ancestral to a modern group is actually quite good. Depending on assumptions, the odds can be 10% or higher. If so, it’s actually misleading to assume that you’ll NEVER have fossils that are ancestors, although whether not it is an error that matters very much is a different question (it might be quite important e.g. for accurate estimation of speciation rates; you don’t want fake “species” appearing for no reason other than an artefact of methodology). So it’s just a matter of time before someone comes up with a decent method for estimating P(specimen is a member of an ancestral lineage), and soon after that everyone except a few holdouts will adopt the new method, and we’ll all wonder how weird it was that people used to say you could never, ever, identify ancestors.
I have to disagree with both you and Nick. First, you: We use dating to establish how evolution occurred – whether, for example, a particular species is the ancestor of another, or when a particular trait evolved. Yes on the second, no on the first. We can rule out ancestors, but we have no way of ruling them in, and nobody does this. When Gary Nelson says we can know an ancestor, does he really mean we can assign a particular fossil population as ancestral to some subsequent population? Or does he just mean we can know quite a bit about the characteristics of that ancestor? Those are two quite different things. (I lodge a pro forma objection against the notion that we can identify species in the fossil record; we can identify morphotypes—and highly biased partial morphotypes at that—but if you want to call those species, you should carefully distinguish that idea from biological species. Unfortunately, the difference is almost always elided.) Further, these days we tend not to use molecular clocks, strictly speaking. Currently popular methods attempt to account for local differences in evolutionary rates, from non-parametric rate smoothing to Bayesian inference, and many methods assign error bars to node ages. We’ve gone beyond the whole clock issue. As to your central point, I’m not sure I understand it. I’m not sure there’s a live argument in biology that corresponds very well to the two sides you adduce, and I’m not sure just what you’re making of it. On to Nick: What value is there is saying that x fossil species is ancestral rather than sister group? Why not assign everything to a tip? What have we lost? If we’re looking at terminal branch lengths, the data can force that species arbitrarily close to the ancestral node, including a branch length of zero if it wants to. But does a positive branch length mean it isn’t ancestral, or a zero length mean it is? We can use that branch length as an estimator of probability that the taxon is ancestral, but I don’t see any gain in doing so. And the difficulty of identifying species briefly mentioned above, combined with the additional difficulty of extending species concepts in time and space, makes this even more suspect. What we can say is that some fossil sure looks a lot like we would expect an ancestor to look, and is around the age that we think an ancestor ought to be. And we could estimate the quality of our sampling. But I still don’t see a point. I also suspect that any method you develop will always place a fairly low probability on ancestral status. Or should.
As to what Gary said, you’d have to ask him, since I am relying on memory of over a decade ago. It shocked me then that he, the Ur-Pattern Cladist (well, along with Donn Rosen. Colin Patterson and Norman Platnick), would admit that we could think something was an ancestor. I don’t know why I was surprised, though, except that the caricature of pattern cladism was that it was antievolutionary. But context suggests that he meant we could know which taxon was an ancestor of some other taxa. So a plesiomorphic taxon might be regarded as likely to be an ancestral taxon for a group. I say “might” because the PC claim is that we do not know this, we infer it, defeasibly. Thanks for the update on molecular clocks. I guess the news is filtering out slowly. I agree there is no live debate in biology here. In fact the debate is one that has played out in the philosophy of history for a long time, going back to the nineteenth century (hence my appealing to “historicism”). But insofar as biology is a historical science, the same issues arise.
On to Nick: What value is there is saying that x fossil species is ancestral rather than sister group? Why not assign everything to a tip? What have we lost? First of all, and most importantly, it would be Really Cool to say Yep, This Is An Ancestor* with high probability. Second, it would matter in at least some practical cases. E.g. 1. Currently many people are engaged in taking molecular phylogenies and estimating speciation and extinction rates, and how they change, for different groups. It is quite interesting if one can infer that, say, the rate of speciation in a group went up after the KT extinction or whatever. However, the estimates (particularly for extinction) can be quite poor if all you’ve got is a tree of modern taxa. Presumably we would do better with fossils. But how should we put the fossils in? There are variety of answers, most of them ad hoc or non-phylogenetic, but the best way would presumably be to put the fossils and their characters into the phylogenetic analysis, and estimate the speciation and extinction parameters from the resulting tree. But if you do that, then you have some problems to face. If you put in every specimen, or an OTU from each geological zones, as a tip, then you’re have a problem wherein you take what is one named species extending through several time horizons, and represent it as a lineage with a short “side branch” in each time zone. But if you used the resulting tree to estimate speciation rates, this would clearly Be Really Dumb to do. But if you don’t do that, here are your options: (1) take only one OTU for each named species, which makes the existence of the species, and the lineal ancestry of each specimen in each time zone, an untested assumption; and you are making the ancestry assumption which most cladists deny making; (2) just say that anything with a (morphological) branch length shorter than X represents the “same lineage” for the purposes of your analysis; (3) have a probabilistic method that makes the estimation of ancestry explicit. Any way you shake it, the no-ancestors-allowed position loses, I think. This is a Real Problem, at least for things like North American Cenozoic mammals where we have quite a good fossil record and a lot of characters. If we’re looking at terminal branch lengths, the data can force that species arbitrarily close to the ancestral node, including a branch length of zero if it wants to. But does a positive branch length mean it isn’t ancestral, or a zero length mean it is? In the ideal situation, it would be some kind of probablistic function which depends on the number of characters and their rate of change, and some estimated parameter about the within-species or within-population diversity in these characters. If you had a thousand characters, and the average morphological branch-length distance between members of the same species was, say, 10, and you had a fossil with a distance of 2, you might be warranted in saying it is more probable-than-not that the fossil represents an ancestor*. We can use that branch length as an estimator of probability that the taxon is ancestral, but I don’t see any gain in doing so. The current method seems to be to Rely On An Expert To Eyeball It And Tell You if fossil X is a member of the “same species” as some living species (to take the simplest case). Somewhat better (and quite commonly done by paleo-types) is to measure a bunch of things and do a PCA analysis and see if the fossil groups with the living species — basically a phenetic analysis. Presumably it would be better if this were all part of the joint phylogenetic analysis. And the difficulty of identifying species briefly mentioned above, combined with the additional difficulty of extending species concepts in time and space, makes this even more suspect. What we can say is that some fossil sure looks a lot like we would expect an ancestor to look, and is around the age that we think an ancestor ought to be. And we could estimate the quality of our sampling. But I still don’t see a point. I also suspect that any method you develop will always place a fairly low probability on ancestral status. Or should. I touched on most of this above. The only thing I would add is that we need to think more carefully about what “ancestor” means. Above, my “ancestor*” means “member of a lineage ancestral to Y”. This is a different thing than, e.g. being an ancestor in population genetics terms. Any particular human dug up from 5,000 years ago has, I believe, only a tiny chance of being a direct lineal ancestor of anyone living today. Wilkins — if you want to coauthor something and cause a ruckus in the literature, I think this topic would be a great one, email me if so.
Nick, Not wanting to hijack John’s blog, I will content myself with a very short reply. But if you and John want to collaborate, I offer my services any time in arguing why your ideas are wrong. (Assuming I think so, which I probably will.) I agree that there is a coolness factor in assigning ancestry, which is why press releases do it so often. But that’s all. No scientific value. The problems you try to solve by assigning ancestry are really problems of deciding how to delimit species in the fossil record. Delimiting OTUs is necessary even if we don’t assign ancestry.
Please, hijack the blog. This is so much more interesting than the usual. Nick, my life is a mess, but let’s discuss a possible paper later…
The book review, in the form of uncorrected page proof, is at Advanced Access on the website for Systematic Biology. The web address, regarding the list of citations mentioned in the review, is now defunct. For the moment, the list of citations can be had from Michael Heads or from me.
Oddly topical for me. I only look at fictive narratives, in particular the development of one particular migratory legend in Europe, the wild man. I am running into ancestor issues in the way I classify at the moment in relation to Asian and European variants. Whilst I am not familiar with the terms used in biology in regard to tax and classification the problems seem very familiar. Hope this gets discussed further its very useful. Asian examples lack two traits found in later European versions. This is however not the case with regard to the early European version of the tale, which also lacks the traits, a point missed as nearly all literature on the subject focuses on later historical evidence. Its not beyond reason that their may be some direct early relationship between Asian and European examples although will require some work to determine exactly what the case may be. Not particularly interested in searching for origin or ancestors but the suggestion by scientists, archaeologists and anthropologists that Asian examples of the wild man narrative actual reflect biological historical and an ‘ancestral memory’ of contact with earlier hominids I find a somewhat unusual and surprising phenomena I did not expect to encounter in an academic setting. The view that you can reconstruct an Ur text establish what motivates the telling and then view that as the factor that allows these things to repeat through time is not a perspective I hold. Its interesting how the issues I face tracking such ghosts of imagination are not entirely dissimilar to the issues faced by biologists with regard to taxonomy. A lot of what John has discussed is directly applicable to my subject area would be interesting to understand further how directly applicable it is to issues surrounding taxonomy in biology. The use of Folklore by some folks researching early hominids is a different but not unrelated subject.
John, I can hardly follow your eloquent words as they move about with quick deliberateness. Can’t follow taxonomist words either. Here be a plaice where causality flounders for the halibut
“I agree that there is a coolness factor in assigning ancestry, which is why press releases do it so often. ” I certainly wish the things that interest me were less of a feature of such events.
Nick writes: “But nowadays, I’m afraid, Heads, etc. are themselves the rigid old guard. They don’t like probablistic models (ML/Bayesian approaches), they don’t like branch lengths, they don’t like estimating dates on trees, they don’t like ancestors, and they don’t like dispersal. And they publish articles and books to this effect. Meanwhile, though, thousands of us are happily using the newer methods and actually *testing*, rather than just baldly assuming, ideas like whether or not a particular divergence event correlates with a plate tectonic event.” My, my! It sounds like Nick is telling us what, and whom, we can safely ignore.
” they don’t like ancestors, and they don’t like dispersal.” Me I will read anything that looks like it may be helpful dealing with the rigidity of Ur origin and the linear dispersal patterns the historical geographic method used in my own very different subject area likes to draw. It leads to a very neat and rigidly ordered filing system in an archive where things never meet and are studied in isolation as separate and apart. Such places have no life and are somewhat sterile.
This started as a response to Jonathan and Johan on John Wilkinson’s earlier post, “What warrant is there for a belief in God?” – I am a slow writer – it now seems appropriate here. Jonathan – Johan – John’s reply. Clearly we are related – members of the same evolving species – also by language through our names – which have their own evolution – going back to the Greek Yehochanan meaning ‘Yahweh is gracious’. This relating to the first syllables of Jehovah – Yeho – then Io the god of Ancient Syrian religion, who was female apparently or with a husband under her thumb. – We may thus argue better than many folk, that we are directly related to God. – We could also argue that we are related to Walt Disney – remember the Seven Dwarfs singing, Hi Ho, Hi Ho its off to work we go. – I’ll bet the ancient Syrians sang something similar. Clearly, between us, there are gaps of various kinds, oceans for instance – but there are connections, such as language. – Some connections are more tenuous than others, some little better than figments of the imagination. The same might be said about the interpretation of the fossil record. Relying on the fossil record, I can make justifiable argument for my being directly related to Apes, a chimpanzee in particular. – However, because of our increasing knowledge and a few new fossil finds, putting a few more pointers into the gaps in the fossil record of humanity, it seems that whilst I can still use the vernacular term Ape, proper scientists have now to be a bit careful. (See John S Wilkins recent post on this site, “Are humans, apes, monkeys, primates, or hominids?” – With the name John, he too then is directly related to God, though clearly he thinks God is a Gorilla. – As to whether he then still qualifies as a human hominid is down to how you interpret the fossil record and the aforementioned post.) My point is that there are many ways to interpret both the gaps and the connections in the flow of events – some warranted, some rather preposterous – some relative to our knowledge, some to our culture. The prevalent cultural convention of Darwinian theorists is to interpret evolution as a continuum – in accord with Darwin’s one mechanism which only has provision for continuum – so the only interpretation of the fossil record allowed is that any gaps must be imagined to be filled. Jonathan says, “Indeed there are gaps, and it would be surprising if there weren’t.“ – that gaps are explained by the paucity of the fossil record – clearly happy that the 65 million year gap between the Coelacanth and its morphologically close living relative discovered in 1938, the Latimeria, has in theory no missing links – clearly believing that the gap MUST be imagined to be filled. There is justification for this belief – many fossil lineages have lesser gaps than 65 million years – the ape, simian, hominid, human lineage, whatever name is put to it, has lesser gaps. Amongst living species we see virtual joins in general trends. In the northern hemisphere, equatorial type hominids have darker skin than those nearer the poles. Genetics and migratory patterns allows us to study the nature of the joins within such trends. – Of course there are still gaps between individual extant organisms we study and between the species we name. So – imagining joins and ignoring the gaps as if filled, we justify a Darwinian continuum. At the end of the day though the only thing we can call fact, is that the gaps are there. The question of whether it is correct to fill them in our minds or not might for ever remain a matter of speculation. Confucius said something like, ‘True knowledge is knowing that we know what we know and knowing we don’t know what we don’t.’ What do we know of the Coelacanth? – It existed in one form 65 Million years ago. Nothing more is known of it – no fossil of later dating or living example has been found so it was justifiably deemed extinct. In 1938 a living ‘Coelacanth’ was found but it was realised that this was a close family relative, Latimeria. Since then only 200 or so living examples have been found – divided into two species, possibly a third. – This tells us that the Latimeria is a rare bird (sorry fish – sorry tetrapod). Did the Coelacanth become extinct 65 million years ago or 6 million years ago – we don’t know. – When did the first Latimeria appear on the planet? – 1938. We know little much of the gap between Coelacanth and Latimeria between those dates. The fossil record of the Coelacanth lineage stretches from 350 Million years ago to 65 Million years ago – the last of the line was named Macropoma, and that was assumed to be the end of it. – Out of the blue (the sea off the south coast of Africa) appeared in 1938 the first living or fossilised Latimeria ever seen by a biologist. – The local fishermen were familiar with them however and knew they don’t taste good. (It’s not on record if a biologist has confirmed this – or for that matter if South African Fishermen have been throwing them back in for thousands of years.) – What we have then, as fact, is a gap 65 million year gap in our knowledge – with little to fill it. I feel warranted, having studied evolution for forty years, in imagining that the precursor of the Latimeria lived in the past, and that precursor was related, by however many links, to the Macropoma and so on back to the precursor of the first Coelacanth. I feel I have little knowledge to justify any idea that the re-appearance of this once thought extinct lineage was divinely planted in the sea off South Africa some time prior to 1938 – though because of the concept, “don’t know” the possibility cannot wholly be ruled out – only forgotten one day perhaps, in the same way we have forgotten that Poseidon rules the waves. The idea does promote an interesting question though. If the Latimeria is the result of a miracle – and the Latimeria is clearly a deep sea fish (sorry tetrapod) – then would such a miracle occur by pushing up a Latimeria into the sea from the bottom, or dropping it from a great height above the sea? – An answer to this question might give us guidance to where we might seek a divinity responsible. We can seek better justification for our imaginings – in gaps though we can find little fact other than that there is a gap – to be filled or not. With only one Darwinian mechanism for continuum, it must be imagined that the gaps in evolution are filled. – I believe that to fill all the gaps may prove an impossible task – not because the fossils are hard to find – but because science has missed a second natural mechanism which facilitates leaps into (or across) gaps. I believe I have found such a mechanism. I ask you for a just moment to consider such a mechanism exists. Then some, though not all the gaps, are real and factual gaps, to be included in the evolution process. – I see little to warrant a denial that this might be so. If it is, then interesting possibilities come to light. – One is that the timescale of evolution can be shortened. I understand that radiometric dating is only accurate concerning relatively late periods – that further back in time we have relied on estimates of the length of time taken for Darwinian evolution to occur – then used these estimates to date geological layers – then used geological layers to date the fossils found to show evolution rates in an interesting circle. Obviously, as long as adjustments are made along the way, the picture should become progressively more accurate. As I think I have discovered a mechanism for occasional leaps, I would expect that any discrepancies which the ‘time experts’ might have found would be of a type better resolved with a shorter not longer timescale. (Is that the case please?) One possibility my mechanism has led me to, relates to change at the basic level – of interest Democritus. – It relates to the subject of of continuity and discontinuity. Darwinian evolution has only the ‘continuity’ of slow ‘joined-up’ change (Eldredge and Gould have led us to the acceptance of change so slow as to appear stasis interspersed with times of quicker change) and the ‘discontinuity’ of extinction. My mechanism suggests a third order of change which I call broken continuity.
John, this is inappropriately long for a comment, and should appear on your own blog. Moreover, many of the points you make are dealt with in some pretty standard places like the Index to Creationist Claims, which you should check before posting these claims. However, note that whether or not evolution is continuous, we should expect to find gaps, because of the spotty nature of paleontological preservation. That is about what we can know, not what we can infer from what we know.
Apologies for length of previous. Please consider my ‘claims’ more as straight and long considered questions, with nothing whatsoever to do with Creationism. I genuinely believe that evolution lacks what might be considered the small tweak of a slight addition I accidentally discovered 40yrs ago. That addition to the theory though, if valid, could dramatically alter many aspects of scientific understanding. This in itself though might be seen as a heady claim. I just wonder how I got lumbered with it. I try to test my ideas at the levels they will be tested if I find the means to publish them. The fossil gap bit is fundamental. I simply ask am I right that as yet there has been no absolute continuum of fossil evidence found, even in rich strata, between any two species – or have I missed something.
What would the fossil record look like is gradualism were true? What would it look like if abrupt (punctuated) evolution were true? What would it look like if Cuvier-style special creation were true? The answer is, it would all look the same if sampling were rare enough. The fossil record is like pictures from the attic of a family’s past. It doesn’t tell you how fast members grew, who is the daughter of whom, or who survived long enough to breed. All it tells you is that at those dates (if they are dated at all) there were members of the family who looked like that. Fossilisation is very rare. Consider, if you will, how many human beings who have lived over the past few thousand years are likely to be fossilised. Almost none of them will, and the likelihood they will be discovered who are, is extremely low. If a speciation event occurred now (say, Republicans become a species due to reproductive isolation with reality-based humans), we’d leave no record of it. At some time in the future, there’d be, maybe, two distinct species, assuming they both live in environments that allow fossilisation. Darwinian evolution does not require a monotonic and gradual evolution of new species; we have known that for a very long time, arguably since Darwin proposed it in the fourth edition of the Origin, and more arguably since Pierre Trémaux proposed it a few years earlier than that. But as any statistician knows, if you only have a few data points, you can fit them to a large number of curves, depending on your regression function. It is not the case that we have no transitions between species, or at least forms (since we cannot tell from the fossil record what are and are not distinct species), but that we have so few is not a surprise.
Yet the fossil record can be used to make inferences about how evolution has worked; Eldredge and Gould have made their point quite clear in that regard. This is illustrated by the fact that in a typical stratigraphic section, you don’t see morphological evolution, even though there should have been natural selection all the time, going by Darwin. Inferences about specific ancestors can in rare cases be made, for example when the geography is constrained, as in the case of island evolution. Also, in the case of microfossils, fossilisation can be so common to allow for true comparison of models; i.e. a strong case can be made that the fossils found are a good representation of all populations in a certain period of time, rather than a biased (bayesed?) ‘snapshot’.
In Darwin’s words to Falconer, “There is not a single word I would object to;… .” On reflection I’d change Republican to politician – it’s not fair on the rest to leave them out. Pierre Tremaux – knew nothing of him – I do now thanks to your 2008 paper. Maybe it was Cuvier to Pictet on one side – with Lamarck to Darwin (use and disuse in his final paragraph summing The Origin) on the other. – with Pierre Tremaux in no-mans land between, ostracised by both camps, absquatulating into obscurity. (I do know one or two long words.) Hugh Falconer perhaps was lucky – dealing more with observation than hypothesis and his observations still valid. As it’s now snowing here on the other side of the Globe in Bacup, a town as famous Tremaux – after a hard day’s plumbing I’m going to bed – not burning the midnight oil tonight. – Your blog refuses adamantly to accept that according to my clock I reply to your posts a dozen hours or more before you post them – it’s a shame really – its refusal spoils a whole new dimension. I am sure nobody comes up with anything wholly new – generally somebody has said some of it before – anything new comes simply by someone later putting the odd bits together differently, then we work on from that. I want to digest your Tremaux translation – the body of text of your paper has wet my appetite. – Then I might email you details of a very simple mechanism everyone seems to have missed and Tremaux might have been clutching for – if you might be interested that is.
Look at how the bright new moon is rising above the land of black and green Listen to the rebel voices calling I’ll not die though you bury me The Aunt upstairs in the bed she’s calling Why has he forsaken me? Faded pictures in the hallway which of them brown ghosts is he? Aisling, Shane McGowan
John, to answer your question: you are not right, because such fossil evidence has been found, particularly in microfossils and a few invertebrate lineages. Ofcourse these are not as ‘sexy’ as whales, birds, or early tetrapods, but any cursory inspection of the rarity of such specimens will make it clear that the transition will never be made seamless by fossils. But do a google papers search on ‘foraminifera phyletic gradualism’ for example, and you’ll find your continuum before you can say Neogloboquadrina pachyderma left-coiling.
But Jeb you’ve touched a chord on the strings of a hardened old heart – old enough to have stopped a beat listening to Ewan McColl singing in his and my dirty old town – and old enough to have beat to a terrible sad rythm as young girls in green danced a jig down a lane near Enniskillin mixed with black mortuary men in white gloves nearly forty years ago – eating ice-cream from an ice-cream van as the parade of history and troubled belief passed by with the pipes skirling – someone giving me ‘the man from england’ another whiskey in the sunshine – that’s just pulled me a tear or two..
It is a good one. I have the same family background with the same political history as Shane, though Glasgow Irish rather than London Irish. My strong sense of being Irish and different from the rest of society should have died generations ago, my family moved here in 1890 I’ll not die though you bury me Nature of politics, dispute and identity. Identity cannot move under such circumstances as it should and communities become deeply divided under the relentless thunder of futile and divisive rhetoric from the extreme sections of both sides. Things have to change can’t be allowed to remain fixed and rigid.
Hi john, Latimeria is not a tetrapod. See J. Clack, ‘Gaining Ground’. It’s not even a tetrapodomorph (p. 67). Geological layers are not dated based on the estimated time for evolution to occur. The comment that radiometric datin of ancient layers is relatively inaccurate is meaningless without reference to specific methods, standards, dates and their resulting uncertainties. I’ve explained why in paleontology, the combination of evolution and deposition of sediment satisfies the current body of observations. If you think a new theory can do a better job, by all means publish your theory in the appropriate channels.
I’ve read this post through a few times now. It seemed to me that clock calibration can and will be refined using whole genome data (say Sayres et al 2011) so that it will be valid for inference about recent evolution/ speciation/population genetics. But that most recent issue of Systematic Biology also contained the article by Saquet et al where calibration differences lead to 100 My shifts in dating, which is to say there is not enough data even from a “rich and well-studied fossil record” combined with limited sequence data when going further back (the actual tree topology doesn’t shift around much though). WRT Bayesianism and Bayesian modelling, there are still plenty of us frequentists. who happily use Bayesian methodologies with uninformative priors. I don’t think we yet have a formalism that combines scientific model generation and model criticism. The latter is what people do in a more qualitative reasoning kind of a way, eg is a worse-fitting mechanistic model superior to a better-fitting atheoretical statistical model, and how can I improve on either? Why does a particular mathematical distribution fit data? And yes, I think you are right that there is another of those strange loops going on in science, between theory and data, or evaluation and application.
I am very uninterested in, or perhaps I had better say incompetent to discuss, the Bayes v. frequency v. likelihoods debate. Whatever happens, though, you have to have some background to assess probabilities against in cases where you have no direct observational data, and in history, where every event is singular even if each event participates in some number of classes (which ones?), the problem of inference exists no matter the methodology adopted. Clearly historical sciences will attempt to infer what happened, or there would be no point to do history in biology or anywhere else. But observational data only gives at best some reference points, and so the temptation is great to assume more than we know as evidence on which to base inferences. The problem of circularity of justification arises almost immediately. I agree that we do not have a good model of scientific reasoning and model building. I think the reason is, as Feyerabend said, the only universal scientific method is that anything goes. You can’t model something that is not, itself, a natural kind (that is, which is not driven by regularities across the board). Nevertheless there are many commonalities shared by most science, and one of them, almost without exception, is that evidence must be empirical (that is, quantitatively measured) when you can get it. The choice of model problem, which some (such as Elliot Sober) think is resolved by Akaike Information Criterion theory, seems to me intractable. AIC works in some cases but not all (as you note, much of it is qualitative, which is another way of saying that our reasoning is often semantic or verbal, or even preverbal). I doubt there is a necessary theory of model choice, myself. But then I think science is an evolutionary process in its own right, and these things change over time and disciplines.
“You can’t model something that is not, itself, a natural kind (that is, which is not driven by regularities across the board).” I feel at times I am on a fools errand and indeed have not engaged fully with increasing my knowledge of issues with taxonomy in biology. Issues do however seem to have some relationship and I do have clear regularities. Enough to suggest that while I may not be able to apply things directly it may offer some hope and stimulate ideas. But often in my thoughts the distinction between natural kinds and the cultural objects I look at do not look hugely different and the forces that shape them seem broadly comparable. I may be completely wrong. But so what, serious error is the best teacher and I am not trying to impose my view on anyone else that’s a habit intellectuals have that I find particularly distasteful.
Jeb: Look at how the bright new moon is risingabove the land of black and greenListen to the rebel voices callingI’ll not die though you bury meThe Aunt upstairs in the bed she’s callingWhy has he forsaken me?Faded pictures in the hallway which of them brown ghosts is he?Aisling, Shane McGowan One upon a midnight dreary as I pondered weak on weary …. only this and nothing more. : )
I would never claim that the ‘Theory of Evolution’ (microscopic sense) is wrong. Rather that it is often not relevant. (Example: Paul Ryan budget plan slammed by Obama as “social Darwinism”) How would a miss-emphasized representation look? … See Why is there only one human species? (BBC) for a plausible instance. For biological evolution, the network-of-causal-sequencing is complex and mufti-threaded. The focal point is confused. Perhaps the Hominidae radiation is explained in the context of marginal survivorship, for the sake of ToE confirmation bias, rather than as a study of causal veracity. The misrepresented distortion indicated as: … to prove Darwin rather than to explore what occurred.
Can’t say I lie awake at night staring at the stars hoping to prove Darwin correct. I lie awake at night staring at the stars thinking I am wrong and hoping a new thought will come. Knowing something, it becomes boring and uninteresting. Ignorance is bliss, its the exciting bit of it all, not knowing, not having a clue, making mistakes, exploring new subjects, learning. The fools errand.
The fools errand. …. and was allowed to roam the Brisbane botanical gardens, but ignorant and nasty folk would put their cigarettes out on her carapace, which is covered in skin and nerve endings, …
I’m unsure where my views fit in to all of this, but I see flaws in the so-called molecular clock while I also see that the data provides important information about evolutionary relationships. For example, I look at the time estimates with a grain of salt while I typically agree with conclusions from the data about cladistics.
Makes sense. A clock assumption, or any dating for that matter, has not generally been used to determine relationships for at least the last 20 years. And though there are methods that take the ages of fossils into account, they are not commonly used. Dates are added to a settled topology after the fact. So you are free to accept the tree and reject the dating. It might, however, be better practice to attempt to place error bars around assigned dates. There are many possible sources of error: in dates of calibration points, in placement on the tree of those points, in lengths of branches both in time and in units of change, in reliability of topology, in methods that reconcile differences between time and change, and probably more that I’m not thinking of. I don’t know that any method has considered all these sources at once.
I accept a molecular clock with the assumptions of neutral molecular evolution and constant rate of neutral mutations. I also assume that the majority of genome evolution is neutral, but my criticism focuses on changes in mutation rates. For example, everything from proto-monkeys to extant monkeys probably probably had the same mutation rate, but the human lineage had an increase in the time of generation cycles that decreased the mutation rate in female germlines. But if we find some genes with the appearance of a molecular clock in all animals, then we are talking about a lucky correlation.
I need to clarify that I’m not talking about the mutation rate being the same in every region of the genome but the mutation rate being the same in a particular molecular unit across respective generations and species.
John Harshman wrote: “I don’t know that any method has considered all these sources at once.” Lartillot and Delsuc (2012) looks like an example of such a joint analysis: ” The most recent common ancestor of placentals is inferred to be a rather small animal (mode 1 kg, 95% credibility interval: 0.4-10.7 kg)…The female age at maturity of this ancestor is estimated to be between 208 and 824 days, and its longevity between 11 and 25 years.”
That doesn’t seem to have addressed any of the sources of error in the dating of phylogenetic trees, or to have any relevance to the question. It’s something else entirely, the reconstruction of character states.
Have I got this right? Noah (in theory) counted species one by one (for clarity) in an unchanging world with a gap between each species. In that tradition, Linnean classification, with each new discovery, finds more extinct and extant species, and so more gaps, estimated now in hundreds of millions, to add to Noah’s lading list. – Cladistics, plotting the genetic relationships between species, joins the gaps with each new genetic discovery, so successively reducing the number of gaps between species till there is only one species. The success of one approach confounds the other. Should the focus of evolution study be on the species, the gaps, or the joins? Evolution study spans the time between whoever wrote the Ark story, to the man or woman poking in a Petrie-dish this very minute. – Darwin sits in the middle of that time, Newton before him, Einstein after him. Thus to ponder the question of gaps or joins in Darwinian terms is the same as pondering the question, if apples did fall sideways would they fall to the left or the right. In the time span – Noah to now – where should we sit to see the process of nature pass by. – Einstein sat on a beam of light to study the universe. Our study of evolution, sitting on a tangled bank, is a bit parochial in comparison. What is the problem we are trying to solve? – then – How do we address it to find a solution?