Well, I might have to eat my hat on this one. It looks like Schlafly has finally read Lenski’s study, and he still doesn’t like it. Okay, I don’t actually know much about science. Unlike Schlafly, I know my limitations (to see Schlafly countermand a doctor on a medical issue, click here). So could someone who does know science rebut Schlafly’s attempted rebuttal?
That said, I’m already noticing one error. Schlafly notes, in point #2, that Lenski may have erroneously included in his studies a bacteria group already known to possess the Cit+ ability, thus (accordingly to Schlafly) stacking the deck in his favor by putting the desired result in the initial sample group. That’s a hell of an allegation! Of course, it doesn’t bear out: if you read Lenski’s paper, in the very first page, Lenski explains that the bacteria group that Schlafly thinks is a polluter (in generation 31,500) was the result of Lenski’s experiment, not a preexisting sample inserted to deceive. Schlafly might be confusing retesting with the initial experiment.
In any event, we now know that Schlafly’s not going to let this one go. He’s even willing to READ to win this argument, thus disobeying the Conservapedia commandment, “don’t read a book to learn, write a book to learn” (I’m serious about that one). But I think I know the secret to his persistence Aside from his earnest desire to please mommy, Schlafly is a recent convert to young-earth creationism. Though this shouldn’t come as a shock, Schlafly’s mind is impressionable. If Schlafly were to lose this debate… and allow himself to recognize and try to cope with the loss… he might begin to question young-earth creationism itself.
A side point. Why is Lenski getting all the name recognition and play on this issue? There were two other scientists, you know, on this paper…. Blount and Borland. Poor guys. I guess they’re the Collins and Aldrin to Lenski’s Armstrong.
Another side point. Conservapedia is dying. If the health of a wiki can be measured in the time it takes for 50 changes to accrue on its “Recent Changes” list, Conservapedia is varying between 6 and 11 hours. In contrast, it takes RationalWiki 5 hours at a slow time of day, CreationWiki 2 days (read: mostly dead), and Wikipedia 1 minute. I’ve predicted it before, but I think Schlafly has alienated or blocked most of the people he cares for exploits.
Submitted to a Candid World 
Schlafly thinks he’s qualified to discuss Lenski Blount & Borland. Schlafly’s a lawyer with a BS in electrical engineering. He’s also is an ignorant blowhard for whom ideological point-scoring ranks much higher than the dispassionate search for truth.
Martin, at the Lay Scientist blog, suggested that Schlafly’s tactic was to spread mistrust of science. Ames pointed out that Schlafly doesn’t even have respect from other creationists. Schlafly has a long, hard job ahead of him to keep from being pwned.
Disclaimer: I’m not a biologist, though I am doing graduate work in engineering right now which involves close work with some geo- and astrophysics people and a lot of reading through papers like this one, and I have a reasonably strong background in biology.
Taking Schlafly’s points in order:
1. Presumably this is a reference to the fact that 2 independent Cit+ mutants appeared in generation 20,000 in the third experiment (and probably also to the 2 that appeared from the 27,000 set). Lenski’s historical contingency claim is that something happened shortly around 31,500 to cause this. Schlafly’s point is probably that we shouldn’t have expected half of the Cit+ mutants in this experiment to have shown up before 31,500.
Lenski was interested in comparing the rare mutation hypothesis with a 1-step historical contingency hypothesis. And it’s incredibly obvious from the data that it’s not a rare mutation. The complete lack of any mutants before gen 20,000 in any experiment, and 17 after, would be mind-boggling if there weren’t some sort of contingency at work. What this suggests to me is that the historical contingency at work may not be limited to a single stage. It’s interesting to note that the 2 Cit+ mutants that appeared at gen 20,000 in the third experiment came from replicants of the same clone (they made 10 replicants of each of 20 clones at each generation to get their sample of 200). Unprofessional speculation aside, nothing that Schlafly is pointing to here contradicts historical contingency in general or makes the novelty of the mutation less likely.
2. You covered this.
3. I can’t easily get access to the paper they cite for their method of testing their collection of experiments for statistical significance. Schlafly seems to think that they ought to just weight the experiments according to their initial sample. I’d assume that there are papers on this Z-method for a reason but it looks like they did the individual P values correctly – I’m only set up to calculate shuffling with replacement, but I get very close agreement on the third experiment (the one with the largest samples). Further, it simply makes no sense to weight the P values given by sample size, if that’s what Schlafly is suggesting. I did a very rough approximation of just lumping all of the results together into a single experiment (I used the sample size and allocation of the third experiment because it was by far the largest) and I found that the collection is easily statistically significant. Contrary to what Schlafly may be thinking, the fact that the first two experiments showed just about the same number of mutants with much smaller samples makes them much weightier in the final analysis.
4. See 3, and I don’t know what PNAS convention is in this area or how serious this is. It looks like nitpicking to me.
To clarify what’s going on for those without the background in statistics to follow, what Lenski and friends are doing in Table 2 is providing the probability that the observed value is a result of the expected process.
Assuming the rare mutation hypothesis, you can find the average generation of the mutant population by simply averaging the generations you’re sampling from (they’re evenly distributed). You can then find an average mutation rate by dividing the number of independent Cit+ mutants by the total sample size (4/72 for the first experiment). It’s then possible in theory to calculate the probability that the random mutation rate you’ve calculated would act on your sample in such a way as to produce an average mutant generation value at least as great as what was observed. That is, there is some chance that an evenly distributed random process would result in a difference at least as large as the one we see between what was observed and what would have been expected. This is difficult to do, and so they’ve resorted to a monte carlo method, which is basically just simulating the experiment over and over again. The P value shown is equal to the number of simulations that had an average mutant generation equal to or higher than the observed one divided by the total number of simulations.
Schlafly seems to be suggesting (it’s hard to tell – he’s incredibly vague) that we find an overall P value by weighting the individual values by the sample size, but this is absurd. That assumes that the analytical expression you could find for P would be linear with respect to sample size, and I see no reason to suppose that that’s the case. Intuitively, it looks to me like you could get something like an ‘overall P’ by just smashing all the results together into one experiment of sample size (72+340+2800) and with 17 total mutants. Consider that when you do this, the small size of the first two samples relative to the third makes them almost negligible. However, the size of the mutant populations was pretty close among all three. Effectively, then, this ‘combined experiment’ is similar to the third experiment but with three times as many mutants in the high end of the distribution. The average mutant generation goes up significantly, as does the hypothesized rare mutation rate. A higher mutation rate is going to tend to push observed monte carlo values towards the average, and our actual observation is now farther from the average, so the P value shrinks drastically.
I have liked the blog very much and have seemed to me to be very interesting.
A greeting.
Thanks! And thanks for the analysis, Gotchaye. I think that’s right on.
Also, I added a new “side point” at the bottom – I think Conservapedia is dying.
He’s updated his list of objections, it looks like. What I was referring to as #4 is now #5. His new #4 seems to be an attack on Lenski for not including “God did it” as an hypothesis, and #6 displays an obvious misunderstanding of how these experiments were actually set up. What he’s saying just doesn’t make any sense – the mutation at 20,000 was never claimed to be the historical contingency point, and those bacteria were entirely separate from the bacteria in later generations in that experiment.
Interesting about its page statistics. Does that represent a significant drop since a relatively recent point in time? I wonder how much of its activity was only ever due to fighting over pages between Schlafly and page vandals. I’ve heard rumors that unsavory individuals from rationalwiki and the like would try to work their way up through the conservapedian ranks in order to bring it down from the inside.
Correction – the abstract apparently claims it at 20,000, but that’s an obvious typo.
Am trying to avoid Conservapedia, for a variety of reasons, but am verty tempted for this.
More to the point, though, the thought that we should include “God did it” as an alternative hypothesis is hilarious. Since it’s inherently untestable, it’s logically excluded from the set of possible hypotheses in any experiment. Andy apparently needs to go back to basic stats and methods classes…
Did break down and read the criticisms, and based on what they’ve come up with so far, not only do I doubt their ability to do basic statistics, but to read and write in the English language. That list is non-sensical at best. And nothing there addresses the underlying finding–Cit+ mutants did present themselves. Attacking the research in this way effectively cedes that point, which would seem to be at odds with Conservapedia’s position on all of this.
Disclaimer: I have a background in stats but not in biology. In this case, it’s not really all that important as they’re not doing any biology, but are instead critiquing their evaluation of hypotheses.
Dammit, Ames, as a molecular biologist I want to help out, but that means wading through Schlafly’s nonsense! I’d much rather see Lenski administer another brilliant literary smackdown.
On your question about Zachary Blount and Christina Borland, Blount is a grad student and Borland was a postdoc in Lenski’s lab at the time of publication. A passing reference to any publication in biology will always cite the lab’s principal investigator, although a good report will provide a literature citation and mention the other authors (or at least the first author).
Gotchaeye – the paper states that the potentiating mutation upon which the later mutation to Cit+ was contingent had occurred prior to 20,000 generations. The final switch to Cit+ likely occurred sometime between 31,000 and 31,500 generations. I think you are mixing up those two. The abstract is accurate.
You’re right, of course – I guess I misread it.
I started a rebuttal at Rationalwiki.com. search for “alleged flaws”
[…] Schlafly’s first experimental foray into the field of science (reported here), an attempt to discredit a real scientist (Lenski) on his own turf, has ended poorly, mainly in […]