Thursday, November 30, 2006

Yockey on Biogenesis

Was the origin of life so amazingly improbable only God could have done it? No.

Wasn't that a deliciously brief answer? Well, okay, first a quick recap, for those who don't like deliciously brief answers: I've argued this at dizzying length at the Secular Web in "Are the Odds Against the Origin of Life Too Great to Accept?" which led to a much briefer and considerably more rigorous paper, still my best known contribution to philosophy: "The Argument from Biogenesis: Probabilities against a Natural Origin of Life," published under peer review in Biology & Philosophy (19.5, November 2004, pp. 739-64). In that article I catalogued and analyzed seven frequently-repeated errors or fallacies deployed by creationists who try to argue that the origin of life ("biogenesis") could not have happened naturally. I even took the trouble of explaining exactly what they could do to avoid all seven errors and make their argument work. Unfortunately for them, actually doing this requires knowledge we don't yet have, which is their most fundamental mistake: they are arguing from ignorance, and arguments from ignorance are, well, ignorant. We simply don't know enough to say whether natural biogenesis is improbable.

Oh, you want to know what those seven typical errors are? Okay. Just a quick list:

  • Class I: Using Obsolete Sources

  • Class II: Omitting Important Context

  • Class III: Using the Wrong Math

  • Class IV: Confusing the Winner with the Players

  • Class V: Begging the Size of the Protobiont

  • Class VI: Confusing Evolved for Spontaneous Features

  • Class VII: Confusing Necessity with Accident
Using old data long since refuted, quoting sources out of context, and using the wrong math are reasonably self-explanatory. Confusing the winner with the players is what I call the lottery fallacy, which I'll talk about below. Begging the size of the "protobiont" (a fancy word for "the first life form on earth") means "making shit up" rather than legitimately calculating how small or simple the first life form was or could have been. As any old logic-curmudgeon will tell you, if you use a premise that you just pulled out of your ass, then your conclusion also comes out of your ass, and smells as bad.

Meanwhile, confusing evolved features for spontaneous is also a big no no, because the probability of something arising by chance is not at all the same as the probability of something evolving once life has already gotten started. And last but not least, we must distinguish when something is a product of random processes, rather than the necessary outcome of physical conditions, since the probability of the latter is always high, no matter how complex it appears. Certain aspects of protein structure fall into this category.

But that's enough about that. If you want to know more, you'll have to dig up the article. I'm not going to rehash the whole issue here. Instead I'm going to talk about just one book, which a patron of the Secular Web asked me to address. In my Secular Web article I already discuss the fallacious and misused work of physicist Hubert Yockey, who is now entering his nineties. Since then, Yockey has come out with yet another book, Information Theory, Evolution, and the Origin of Life (2005), which revamps the argument of his previous books and articles, now using information theory. Which is all very fancy and nice. But he still fails to address the central flaws of his earlier work, and simply repeats the same old fallacies. I'm a bit surprised that an academic press would publish it without addressing these issues. I'm not saying his book is complete poppycock, but a few key mistakes undermine Yockey's otherwise challenging work throughout.

Once again, Yockey reveals his true agenda in his closing chapter: he wants to destroy SETI and redistribute resources to other scientific projects (pp. 186-89). This has been the underlying theme of his past work against biogenesis research. As he said in a previous article long ago, it annoys him that so much time and money is being spent looking for "little green men." Yes, his words. Never mind what a straw man that remark is. What matters is that throughout his new book he is attempting to argue that we have no sound reason to expect to find life on other planets. Any other planets. Anywhere. Now, when it comes to finding life elsewhere in our own solar system (like, say, Mars, or Titan), I actually agree with him, but for entirely different reasons. I think life is much too rare to happen twice independently in one solar system. But even so, extending Yockey's argument to the whole SETI program (as well as the fields of xenobiology and protobiology) is not justified by anything he proposes in this book. The universe is much too big a place for all xenobiotic projects to be a stupid idea, even if some of them are impractical or unrealistic.

Still, I'm told creationists have started citing this new book of his in defense of their own argument that God must have started life on earth. I've not seen this abuse myself, but it wouldn't surprise me, and at any rate, forewarned is forearmed. If that's their plan, Yockey throws a bucket of cold water on it. "I have no doubt," he says, "that if the historic process leading to the origin of life were knowable it would be a process of physics and chemistry," hence his argument is that a natural "process of the origin of life is possible but unknowable" (p. 173). In fact, in a brief but amusing digression, Yockey pokes a stick at creationist Michael Behe, making the correct point that life is not analogous to mousetraps (one of Behe's favorite anaologies), because life reproduces by consuming what it traps, thus leveraging development over time (pp. 178-79).

Ultimately, Yockey says, "there are no 'irreducibly complex systems' in biology and there is no need for an Intelligent Designer in evolution" (p. 184). Although Yockey directs readers elsewhere for the details of his argument (his daughter provides a longer summary online), a different class of amusement can be found in what several engineers have done with Behe's analogy: see A Reducibly Complex Mousetrap by
John H. McDonald and The Evolution of the Mousetrap through Natural Selection by Alex Fidelibus (although I've seen even better evolutionary models from A.J Hill, starting from a basic deadfall trap). Yockey also points out (as I have done) that increasing entropy is not a barrier to life but actually "required for evolution to proceed" (p. 185). This point is articulated in John Avery's book on the same subject, Information Theory and Evolution (2003).

Okay. So much for Yockey being a friend of creationists. But even what Yockey wants to argue is, well, kind of lame. In making his case for the "unknowability" of the process that originated life, he admits (and even employs as part of his case) the fact that we cannot derive conclusions today about what's impossible based on what we don't know now, since science continually uncovers things we never knew before. Right on. But this same argument destroys his own case for "unknowability," which is essentially just another "it's impossible" argument, which he bases only on what we know now (even if Yockey is entirely correct, which I doubt). For that very reason his entire case is invalid. He has no justification for claiming the process of biogenesis will never be known, even if he can't yet see how we will discover it. Yockey of all people should know that any scientist worth his salt will see arguments like his not as a reason to give up (which is what Yockey wants them to do), but as a challenge, a sterling motivation to solve what Yockey thinks unsolvable. Scientists have done this sort of thing before. Quite a lot really. Is Yockey unaware of the lessons of history? "You'll never explain it" has been a losing bet so frequently, I don't see how Yockey got the idea of backing a lame horse.

Whatever. Let's get to some numbers. The only actual numbers Yockey generates that could be abused (by an unscrupulous creationist in some attempt to convert them into statistics) are not usable in such a way, nor valid even if they were. Of these, the most significant are the highest (roughly 10^70) and the lowest (roughly 10^-36). Already that lower number is well below the Borel threshold of 10^-50 and therefore can't produce a successful argument against natural biogenesis. What's a Borel thingy? Well, this French guy named Emile Borel (left) calculated that anything with a chance of happening less than 1 in 10^50 is basically impossible, but anything with better odds could have happened at least once in the whole universe by now. Even the creationist mathematician William Dembski, who updated Borel's work, found that his threshold should actually be set far higher, around 1 in 10^150 (a fact I discuss in Biology & Philosophy).

So what's the deal here? Yockey is implying that the prebiological production of cytochrome c (his favorite example of a protein) had about a 1 in 10^36 chance of happening on earth (p. 96). Yet even Dembski agrees this probability would entail that gazillions of events of exactly that improbability have already happened throughout the universe. And for all we know, one of those umteen kazillion cases might have been the appearance of this very protein. I don't think that's what happened (there are so many things wrong with Yockey's assumptions here I don't even know where to begin), but since the odds even Yockey gives against this are well under the Borel threshhold, Yockey must concede that such an argument as he implies doesn't hold water even if all his assumptions are correct (and I don't think they are). Goodbye statistic number one.

What about that other number? It, too, is well within the Dembski threshold, so we could dismiss it for the very same reason. But, to paraphrase Douglas Adams, just as some people think even coming down out of the trees was a bad move on humanity's part, so there are some who really insist on sticking with Borel's threshold. Hey, whatever man. I'll grant that such old-timey stick-to-itiveness would make Yockey's 10^70 an impossible event. Unfortunately for the creationists, though, it's also an event that has nothing to do with the odds of natural biogenesis.

As I've caught him doing before, Yockey calculates this number by invalidly "supposing" that exactly and only "twenty amino acids" have to be involved in any biogenesis event. This is a ridiculous assumption that I've already refuted, in both my works cited above. There are many more than twenty available, and twenty are not needed at the outset. Hell, only four are needed to compose DNA. But it doesn't end there. Yockey adds the even more implausible assumption that our DNA's "sixty four codons were all present at the origin of the genetic code," even though there is no reason to suppose this. As I've argued in both articles above, which is what most protobiologists maintain, life probably began with a simpler coding process that evolved into the current DNA code, and even if it didn't, ours isn't the only one available--there are, no doubt, a vast number of different coding systems that could have gotten life started. Basic math: We want to know the probability of any of these possible coding systems arising by chance, not just the probability of ours arising by chance.

I've called this the lottery fallacy. It's like arguing "it's almost impossible that I'll win the lottery, therefore it's almost impossible that anyone will win the lottery." Sorry, that's not how it works. In actual fact, it's almost certain someone will win the lottery. The person who wins can't claim he's hallucinating simply because "it's so improbable I won that it can't really have happened." I've also called this the "configuration of the stars" fallacy: that the stars would be arranged exactly as they are when seen from earth is astronomically improbable. Pun intended. But that the stars would be arranged in some comparably improbable way is absolutely guaranteed. It has a probability of 100%. So you can't argue "that this arrangement would occur is so improbable, God must have arranged the stars!" That's just silly. No matter how the stars were arranged, that arrangement would be just as mind bogglingly improbable, whether random or planned.

That's what's going on here. Let's say there are a billion different ways life could begin, each leading to a completely different history of life, each completely alien to the other, but all just as amazing and complex. We would call each a unique "biosphere," encompassing all the strange life that evolves from that first organism. And let's say there is only enough time for one of these biospheres to appear by chance somewhere in the universe, but plenty of time and room for that to be certain to happen at least the one time. Yes, the probability of any specific kind of biosphere occurring is then one in a billion, but since one in a billion of them is almost certain to appear in the time available, the probability that one of these biospheres would appear is not one in a billion, but practically 100%.

Yockey completely misses the boat on this. He thinks the probability that our biosphere would be the one selected by chance is the same thing as the probability that life would originate by chance on earth. Wrong. Since some biosphere will be selected somewhere, the fact that this one occurred here might just be that one random occurrence that was already guaranteed to happen in the universe, just like the configuration of the stars. Now, that's just a hypothetical example with probabilities I just made up--the real math is much more complex--but the principle is the same. It doesn't matter how improbable our specific biosphere is. All that matters is how improbable it is that there would be any biosphere. Yockey never even attempts to figure that.

Moreover, this 10^70 of his is nonsense to begin with. Where does that number come from? It's "the number of genetic codes" of the sixty-four codon, twenty-amino-acid type that are possible. From this number, Yockey argues our code could not have been "selected by evolution." But for all we know, all of those 10^70 coding arrangements could be usable to generate life (along with uncountable other coding systems which are not of the sixty-four codon, twenty-amino-acid type, a contingency Yockey never includes in his calculations). In other words, Yockey is assuming that evolution "had" to hit on "our" coding system, and that alone. But he never explains why anyone should believe that. This is just another lottery fallacy. If the universe rolled a 10^70 sided die, any outcome of which would be a life-producing genetic code of some kind, then the probability of a life-producing genetic code appearing would be 100%, not 1 in 10^70. Or suppose only a tenth of all possible outcomes would be a
life-producing genetic code of some kind, then the probability would be 10%, still not 1 in 10^70. Yockey never even tries to work out what percentage of all the possible codes would effect a viable biosphere, and yet that's the only number that matters. Goodbye statistic number two.

Yockey repeats this same fallacy by again reviving his use of the example of cytochrome c (pp. 27-29), which I already dismantle in my discussion of his past work. This time he derives a viability space for a certain form of cytochrome c, which is actually below the Borel threshhold (his smallest number, above), so it's moot anyway, but he still ignores the fact that we are not looking for the probability of landing in the viability space for cytochrome c, but of landing in any viability space for any self-replicating protein (which cytochrome c is not). I won't bore you by discussing what the hell a viability space is (if you're a glutton for punishment, see my discussion of Murray Eden at the Wistar Conference). All you need to know for now is that the size and complexity of spontaneous self-replicators can easily be smaller, and their number and viability space larger, than Yockey estimates for cytochrome c, so his analysis of cytochrome c is a complete waste of time. He produces no evidence to justify his estimate of viability space even for cytochrome c anyway. For all we know, every version of cytochrome c is viable in some biosphere. In fact, does any biosphere even require cytochrome c in the first place? Yockey doesn't even address that question.

Like he has done in his previous books and articles, Yockey actually backtracks by admitting that simpler codes could have gotten life started, and that in fact the number of possible codes for a doublet codon system is not 10^70 but 10^4 or even as low as 10^3 (pp. 97-101), which entails that biogenesis was probably accomplished with such a simpler coding system--not because there is any merit to Yockey's fallacious argument (which invalidly assumes only one of those coding systems will work), but for the simple reason that simpler coding systems are far more probable to arrive at by chance and therefore will jumpstart life far more frequently, while a slightly more complex code can subsequently evolve from the simpler one, vastly expanding the evolutionary resources of the biosphere. Yockey admits this is possible and even what many scientists propose. Yet this seems to destroy the entire argument of Yockey's book. Once again, he refutes himself. Dude, what's the deal?

For all these reasons, when Yockey concludes "it is quite out of the question that life emerged by chance" from any primordial soup, and that "all speculation on the origin of life on earth by chance can not survive" (pp. 118-19), this conclusion rests on fallacious premises, and therefore is itself fallacious. Goodbye Yockey's entire book. He even repeats the old canard about "homochirality" being improbable, even though we have already discovered numerous possible natural sources of homochirality that make its probability very high, not very low. I won't bore you with a digression here about what "homochirality" is or why Yockey is such a dufuss for assuming it had to be arrived at by random chance, but if you are really keen to know stuff like that, I discuss it enough in my two articles.

In the end, none of Yockey's math even leads to his conclusion that "an unbelievably enormous number of sequences must be formed in order to find the ones that form the minimal essential genome of the first living organism" (p. 133). He never produces a valid calculation for how small "the minimal essential genome of the first living organism" could have been, and doesn't even suggest a number for how many different "minimal essential genomes" there are (or how many viable codes there are), and yet we must know both even to begin to work out whether "an unbelievably enormous number of sequences must be formed" in order to hit upon life. Thus his book's conclusion that "it is impossible that the origin of life was 'proteins first'" is completely unwarranted by anything in his book.

Now, again, Yockey is at pains to emphasize that all he has argued is "that the origin of life, like the origin of the universe, is unknowable." Not that it couldn't have been a natural process. And of course he emphasizes that natural law fully explains evolution "once life has appeared" and thus does not need "assistance from an Intelligent Designer" (p. 181, 184). He even points out the important (and, by creationists, often neglected) phenomenon of gene duplication and redundancy in the evolutionary process. As Yockey says, the evidence "proves that the theory of evolution is as well established as any in science." It is equally true that we haven't yet "solved" the question of biogenesis. But we have many good theories on the table and many worthwhile avenues of research to explore, and much is getting done in that field. Calling for all this to stop makes no sense. Even so, Yockey is right to remind us of an important truth: "the fact that there are many things unavailable to human knowledge and reasoning, even in mathematics" (math nerds take note: he is referring to Gödel's Incompleteness Theorem) "does not mean that there must be an Intelligent Designer."

Indeed. Ignorance is not a sound basis for a conclusion. If only Yockey took his own advice.


Andrew said...

And yet scientists are still unable to create life.

But what I would really like to see them do is accompish that feat by a mindless, undirected, process.

Richard Carrier said...

That's not a reasonable request, Andrew.

It's very likely that the smallest self-replicating protein will have a probability of spontaneous formation in random mixing of around 1 in 10^20 trials or possibly as much as 1 in 10^40 trials. Even if we redirected the entire world economy to doing nothing but running the random trials to test this prediction, it would still take thousands of years before the predicted result would occur.

If every human being on the planet (or six billion of them) ran one random trial every minute of every day, nonstop, 24-7, without ever sleeping, we would not expect to see a naturally-occurring 1-in-10^20 self-replicating peptide for several tens of thousands of years. If we are looking for a 1-in-10^40 target, we'd have to wait ten trillion-trillion years.

Since you won't live that long (even if any civilization were so insane as to actually run such an experiment), you can have no reasonable expectation of what you are asking to see.

Indeed, if a naturally-occurring self-replicator isn't enough for you, we already know that it takes many millions more years for natural processes to turn a simple self-replicator (via natural selection) into a complete organism (a single-celled life form with a self-sustained metabolism). Because the evidence shows that was the time scale involved the first time around, and some simple math corroborates this.

However, the universe is vastly larger and older than you will ever be or human civilization has ever been. We could marshal only six billion lab assistants. The cosmos houses over one hundred billion-trillion stars, collectively sporting at least as many planets as potential lab assistants running trials. And we can't even imagine running an experiment for tens of thousands of years, yet the universe has been running its experiment for billions of years.

If it would take 6 billion people 10^25 years to get a 1-in-10^40 self-replicator by random natural processes (i.e. by, exactly as you say, an entirely "mindless, undirected process"), running one trial a minute each, it would take the 10^22 planets that the universe probably contains about one million years to see a successful result (if the average number of trials is one per minute per planet--since even assuming some planets run zero the whole time, others will run trillions, etc.). The universe has hit that time-mark fifteen thousand times already.

Of course, since there are over a trillion galaxies, and these self-replicators will arise randomly and thus be scattered about randomly, there will only be one per every million galaxies or so--thus we're probably quite alone in our own galaxy, unless self-replicators in the 1-in-10^20 range are possible, which is not unlikely. But since we've already discovered some in the 1-in-10^40 range, there is no reason to believe the thing you ask for (life forming by a mindless, undirected process) hasn't occurred many thousands of times already--we're just too far away to see those other results.

Thus, science doesn't have to do what you ask, in order to demonstrate that life could have originated naturally on earth by a "mindless, undirected process." We can do the math from observed facts instead. And that's all anyone should expect.