Saturday, September 7, 2013

John Harshman's Technical Objections

Original post by: John Harshman

Note: Harshman's review of the other chapters of the book will be added later.

First, a take on the whole book that I don’t recall seeing in other reviews:

I've finished reading the book, and to my surprise it turns out not to be about the Cambrian explosion at all. Sure, the first few chapters are, but they and the explosion they discuss are irrelevant to the main point, which is that any significant amount of evolution is impossible. Meyer proves that no new protein can arise, and no developmental program can change, not even once in the entire history of life. So forget the Cambrian explosion, whose duration is by the way irrelevant. Humans and chimps can’t be related; they’re just too different for evolution to manage. Oddly enough, Meyer seems not to take his own message and doesn't draw the conclusions about the history of life that follow directly from his studies. Not even with a “much light will be thrown” sentence.

And, by the way, don’t go countering with anything like “we have conclusive evidence that it happened, so it must be possible”. Meyer rejects all historical evidence as mere conjecture.

Part 1: The mystery of the missing fossils

Chapter 1: 

Darwin’s Nemesis. Why Louis Agassiz was a great scientist and was perfectly right not to accept evolution. Agassiz was all for separate creation of each species, which Meyer conveniently elides into the inability of evolution to generate “wholly novel organisms”, without ever confronting the difference. And of course he raises the title problem that the book is ostensibly about: the sudden appearance of disparate animal taxa in the Cambrian explosion.

Chapter 2: 

The Burgess bestiary. All about the Burgess Shale and, eventually, the Chengjiang fauna, interpreted as weird wonders with no relatives. Hallucigenia, for example, is considered a bizarre, one-of-a-kind monster, which was certainly Conway Morris’s original notion; but that changed, and now we know it’s connected to a number of other Cambrian fossils and to modern onychophorans. Similarly, he can simultaneously claim there are no transitional forms while touting Anomalocaris as just an unusual arthropod.

Here we begin two major confusions that are repeated and amplified in succeeding chapters: first about when the Cambrian explosion happened, as any phylum with a first appearance in the Cambrian is counted in fig. 2.5 as part of the explosion, including phyla that appear in the 20+ million years of the Cambrian that he fails to mention before his explosion starts; second, confusing appearance in the fossil record with appearance on earth, as if the record were perfect.

And we also begin the habit of cognitive dissonance; Anomalocaris (above) is one such example. He also is capable of noticing (in fig. 2.5) that a dozen phyla have no or almost no fossil records while simultaneously proposing that the record is nearly perfect. I suppose you can reconcile that if you presume that some phyla have been created just recently, but Meyer seems not to notice, as will often be the case below, that his claims have implications.

A major claim in this chapter is the idea of “top-down” appearance: phyla appearing before families, families before species, etc. He dismisses the idea that this is an artifact of classification, but makes no real argument. But phyla were defined based on extant species as the broadest classifications, and so must arise earliest in the history of life, before lower-level groups that they contain. His counter is that these early taxa all have the distinctive features of their modern relatives. Oddly enough, he frequently cites one of my favorite papers, Budd & Jensen 2000, which shows that nearly all Cambrian taxa are at best stem-members of their respective groups. And he relegates potential transitional fossils (Anomalocaris, Opabinia, halkieriids, etc.) either to extant phyla or to new phyla, again unrelated to any others. Each transitional fossil, in other words, just creates another gap.

Chapter 3: 

Soft bodies and hard facts. Here we dismiss the idea that the sudden appearance of soft-bodied taxa in the Cambrian explosion can be a preservational artifact. For example, we have preserved fossils of bacteria in stromatolites billions of years old. If tiny little bacteria can be preserved, reasons Meyer, then no large animals should remain unpreserved. Can anyone be this naive about taphonomy? I suppose so. But different taphonomic conditions preserve different things; what preserves bacteria doesn’t necessarily preserve animals, and vice versa.

More cognitive dissonance: he takes pains to point out (in the previous chapter) that fossil deposits like the Burgess shale are extraordinarily rare, but here declares that if there were equivalent species before the Chengjiang, we would have found them. (I will note also that he doesn't say “before the Chengjiang”; he says “in the Precambrian”, again ignoring a 20-million-year stretch of early Cambrian time).

Meyer also perpetuates the claim that many body plans are impossible without mineralized skeletons; he consistently confuses “hard” with “mineralized”, despite the evidence of the commonly preserved, mineralized trilobites vs. rarely preserved, non-mineralized arthropods of the Burgess and Chengjiang. Clearly, a tough, organic exoskeleton or shell can make a body plan possible without readily preservable mineralization. So, what we have in the Cambrian explosion is the sudden appearance in the fossil record of a host of phyla, but what that means is that they all appear in a single deposit, the Chengjian fauna. There are no earlier deposits with a similar type of preservation. Meyer, looking through a narrow window into a meadow, sees a horse, and therefore concludes that there are no other horses in that meadow to left or right of his view.

Chapter 4: 

The *not* missing fossils? This chapter is all about the Ediacaran fauna, with the purpose of dismissing Ediacaran life as transitional. And indeed much of it isn’t. Much of what he says here is true. Spriggina probably isn’t bilaterian at all, since it isn’t bilaterally symmetrical. However, he also dismisses other potential intermediates on the basis that they lack derived characters when in fact we can’t know whether they had them or not; preservation quality just isn’t good enough to tell. At the end, he mentions Kimberella, a fossil he had earlier accepted as a mollusk, but here he does all he can to cast doubt on its nature. Note again: Meyer goes straight from the Chengjiang (about 520ma) to the Precambrian (ending about 543ma) and never talks about the 20+ million years in between. God of the gaps, indeed.

Chapter 5: 

The genes tell the story? Meyer starts by attacking the molecular clock, which is admittedly an easy target. Estimates of the age of the bilaterian common ancestor vary widely depending on data and methods. To Mayr this means that all such estimates are meaningless, but that isn't necessarily true. We might, in fact be learning more about how to do it right. But I will agree that error bars should generally be wide. Meyer does however compound the problem by failing to clearly distinguish at least three separate nodes: Metazoa, Bilateria + diploblasts, and Bilateria. These all presumably have different ages, so randomly listing dates as if they all estimate the same thing is a problem. We now also believe there are multiple clades of both sponges and diploblasts, so lumping them conceals further nodes. 

This, however, is perhaps the weirdest claim, which bears quoting: "Histones exhibit little variation from one species to the next. They are never used as molecular clocks. Why? Because the sequence differences between histones, assuming a mutation rate comparable to that of other proteins, would generate a divergence time at significant variance with those in studies of many other proteins. Specifically, the small differences between histones yield an extremely recent divergence, contrary to other studies. Evolutionary biologists typically exclude histones from consideration, because the times do not confirm preconceived ideas about what the Preambrian tree of life ought to look like." In other words, he's accusing biologists of cherry-picking data to fit (the irony of which escapes him). No that isn't why. It's because histones have an evolutionary rate (not, incidentally, equivalent to mutation rate) much slower than that of other proteins, and this can easily be shown by comparing divergences much more recent than the Cambrian. Though it may be that Meyer doesn't believe in different evolutionary rates, because he doesn't seem to believe in those recent divergences either, or in evolution of pretty much any sort. 

In another part of the chapter, Meyer begins to doubt that there is such a thing as homology or phylogenetic relationships. While it's true that tree-building methods assume that there is a tree to build, there are also ways of testing whether the tree built is a better fit to the data than some other tree, or in fact than no tree at all (e.g. Theobald 2010). But to Meyer, phylogenetic analyses do not count as evidence of common ancestry. Conveniently. 

And finally there is an attempt at Catch-22. A bilaterian ancestor must lack the special characters of descendant groups, so those characters must arise later. And he thinks that there can't be time for such characters to arise (because, as he tells us later, no amount of time, including the entire history of the earth, would be sufficient for even one of those characters to evolve). 

The question of whether there was a bilaterian ancestor is of course separate from the question of its age. We end with a shameless quote-mine from Simon Conway Morris that doesn't at all say what Meyer wants to make it say, much less mean what he wants it to mean. "A deep history extending to an origination in excess of 1000 Myr is very unlikely", which Meyer takes to mean that Conway Morris thinks metazoan evolution must begin very close to the Atdabanian. How 1000 became close to 520 is unclear.

Chapter 6: 

The animal tree of life. Although Meyer doubted common ancestry in the previous chapter, it's necessary here to drive a stake through its heart by showing that phylogenetic analyses are invalid. And we do that the same way we dealt with the molecular clock: different analyses disagree! For this he goes as far back as the 1940s, never acknowledging that significant consensus has emerged more recently. Meyer falsely claims, though I'm not sure he realizes what he's saying, that phylogenetic analyses assume a molecular clock. 

By the way, either the quote-mining is thicker in this chapter than in previous ones, or I've just read more of the papers. He cites a paper about conflicts among gene trees due to lineage sorting to claim that phylogenetic analyses are spurious. Of course it means nothing of the sort, only that the histories of genes may differ slightly from the histories of the species in which they are embedded. And he uses studies that claim extensive horizontal transfer to make the same point. Finally, he uses other studies that point to the possibility of very short branches that would be hard to resolve. In other words, if history is more complicated than a simple, single, obvious tree, it therefore doesn't exist. Oddly enough, though Meyer rejects the tree, he accepts affirmations based on it that the Cambrian radiation was quick. 

Next he attacks the agreement between molecular and morphological phylogenies by pointing out that there are disagreements. Should have actually read Theobald's "29+ Evidences" instead of merely quote-mining it. 

After that, we discover that morphological characters are not always in agreement with each other, and that some are quite labile. Therefore, of course, there is no real phylogeny. 

We finish with repetition of an earlier point, that phylogenetic algorithms assume a tree; again, no mention of statistical tests. And anyway, convergent evolution (or, to Meyer, a hypothesis of convergent evolution, since he never accepts that evolution really happens) makes phylogenies invalid. Because hey, if there's any homoplasy at all, we can't trust anything, right?

Chapter 7: 

The existence of this chapter is inexplicable, since PE was never intended, by Eldredge, Gould, or anyone else, to account for the Cambrian explosion. Nevertheless, Meyer triumphantly and at length proves what everyone knew from the start. Meyer indulges in two main confusions (neither limited to creationists). 

First, confusion of time scales. The difference between PE and "gradualism" isn't about the mechanism of evolution -- natural selection in each case -- but about whether change has a constant rate over geological time or is episodic; but while PE episodes are rapid in geological time, they are gradual on the human scale. 

Second, confusion of magnitude. The lack of transitions PE is intended to explain are those between closely related, similar species, not those between higher groups. Nothing at all to do with the Cambrian explosion. Meyer, despite a long explanation of its genesis, doesn't seem to know that PE was originally intended as an exploration of the consequences of Ernst Mayr's ideas of speciation for the fossil record. 

Here's another fine quote mine, again just one I happened to notice because I've read the relevant paper. Meyer says "As Foote explained (writing with Gould in fact), the adequacy of punctuated equilibrium as an account of the fossil record depends on the existence of a mechanism 'of unusual speed and flexibility'". But in fact Foote & Gould weren't writing about PE at all, or even about a mechanism. Here's the actual sentence in which that fragmentary quote is embedded: "Moreover, even if their conclusion were correct, it would support the idea of unusual speed and flexibility in Cambrian evolution followed by constraint upon fundamental anatomical change." In that sentence, "They" are Briggs et al., who attempted to show that arthropod disparity in the Cambrian was about the same as that of present-day arthropods, and the Foote & Gould letter is a methodological critique. The underlying issue is whether rates of evolution were unusually fast in the Cambrian explosion because of increasing developmental canalization toward the present. Again, exactly nothing to do with PE or any evolutionary mechanisms. 

OK, that's all the paleontology. The rest of the book is just about why evolution to any significant degree is impossible. So all the argument about how long the Cambrian explosion lasted, or just how much happened, is irrelevant, since no matter how much time is available, it isn't enough, and all of the other radiations in the history of life are impossible too. 

Part 2: How to build an animal 

Chapter 8: 

We begin with the stunning revelation that without new variation, natural selection will eventually grind to a halt. A tedious review of the history of genetics eventually arrives at mutations to DNA sequences as a potential source of variation. From there we make a leap to the assumption that the Cambrian explosion required vast amounts of new biological information. But of course that's an undefined term, and Meyer realizes that a definition is necessary. 

First, a digression into complexity. We intuitively suppose that a sponge is more complex than a choanoflagellate, and a trilobite more complex than a sponge. But how could that be quantified? Meyer starts with the number of cell types, which gives him a convenient ladder of life. He is even able to estimate, by making it up on the spot, the number of cell types in various extinct taxa. (Once again, by the way, we jump instantly from 555ma, without intervening events, to 530ma. Just saying.) How true is that? Hard to tell, as cell types are difficult to quantify, and tend to expand in numbers as you look for more fine distinctions. The closer to humans we get, the more cell types we tend to see. But does that reflect anything more than our obsession with ourselves? 

Suddenly we're back to genetic information, which we are now estimating by looking at genome sizes. Using exactly three data points (minimal prokaryote genome at c. 500,000bp, unspecified protist at "upwards of a million", and Drosophila melanogaster at 140 million) we see that genome size is nicely proportional to either complexity or information content, not sure which. At this point I wonder if Meyer has ever in his life encountered the term "C-value paradox". 

Though we still haven't defined "information", Meyer now asserts that the Cambrian explosion must have required oodles of it in the form of new cell types, proteins, and genetic information, the third of which is apparently in addition to the other two. 

Finally, Meyer promises to define "information". He starts with two sorts: Shannon information and functional information. He spends much time explaining Shannon information, after which he tells us that isn't what he's talking about. No, he's talking about that functional information. But unfortunately, he never bothers to define that. It's like, you know, meaning and/or specification. 

So, in sum, pretty much a useless chapter, in which Meyer alleges without real evidence or argument that a huge increase happened in a quantity he is unable either to quantify or define.

Chapter 9:

This is mostly about the Wistar Conference,  an attempt by engineers and computer scientists to help evolutionary biologists by showing that they were wrong about everything and introducing a little mathematical rigor into the field founded in part by R. A. Fisher. It develops that it's hard to assemble a specific protein sequence by chance. And that only 1 in 10^90 of all 100-residue proteins is a functional cytochrome c. There is much time spent proving that a very big number can still be much, much smaller than an even bigger number. But don't worry: Doug Axe will make everything clear in the next chapter. This chapter is merely preliminary to the meat of it all. 

Chapter 10:

First, Doug Axe has epiphanies. He realizes that artificial selection is intelligent design, and so is Dawkins' "Me thinks it is like a weasel" program. And he realizes that building a new organism requires building new proteins, apparently forgetting his claimed expertise in gene regulation, or perhaps only forgetting what promoters are. Oddly enough, this is followed fairly quickly by a citation of one of Ohno's papers, of which the major point was that most Cambrian explosion animals had the same genetic tool kit, "nearly identical genomes, with differential usage of the same set of genes accounting for the extreme diversities of body forms." But of course some of those nearly identical genes were new before the explosion, and that's what Meyer wants to notice. I know we were supposedly talking about the Cambrian explosion itself, not its prologue, but bear with him. He's making a point. 

Next we discuss protein folds. How different does a tertiary structure have to become before it can be called a new fold? I have no idea, and Meyer doesn't say. Here's another thing Axe knew, because he was a protein scientist and they know stuff: " protein folds could be viewed as the smallest unit of structural innovation in the history of life." And "...the ability to produce new protein folds represents a sine qua non of macroevolutionary innovation." I guess Ohno was just kidding. So, having reduced macroevolution to the evolution of new protein folds, Axe finds that randomly replacing 1/5 of the exterior amino acids in a protein make it no longer functional, at least in its old role. We don't know if it had a new function, but of course it's hard to test for some unknown function. Axe also finds, surprisingly, that changes to a protein resting on an adaptive peak tend not to be selectively advantageous. 

But here's the important bit: Axe's experiments show that it's impossible (that is, so improbable as to have a low chance of ever happening, anywhere, during the entire history of life) for one functional protein fold to evolve into another, either gradually through selection or drift, or by macromutation. Thus the duration of the Cambrian explosion is irrelevant. The smallest unit of structural innovation is unable to emerge no matter how much time you give it. No new proteins can evolve. And macroevolution is all about new proteins. Oh, and "new function" is synonymous with "new fold", so no new functions, ever. Bumblebees can't fly, so don't bother pointing out that bee in the garden. 

Again, Meyer seems uninterested in grasping the implications of what he's just proven. If no new protein can arise, it isn't just the Cambrian explosion that's in trouble. It's the entire history of life. The relationship of humans to chimps and many other comparatively recent divergences are also problematic. Why no mention? 


Budd, G. E., and S. Jensen. 2000. A critical reappraisal of the fossil record of the bilaterian phyla. Biological Reviews 75:253-295.
Foote, M. H., and Gould, S. J. 1992. Cambrian and Recent morphological disparity. Science 258:1816-1817. 
Theobald, D. L. 2010. A formal test of the theory of universal common ancestry. Nature 465:219-222.

1 comment:

  1. Note: no reviews of other chapters will be added later, as I started to lose interest as soon as it became apparent that the Cambrian explosion, my main interest here, was just the bait to get me to read about something else altogether.