It’s not really visible yet, but some of the more credentialed people on both sides of the evolutionary debate are starting to take notice of Probability Zero and MITTENS. And, of course, the usual sort of suspects are, as always, focused on the irrelevant minutia, so I’m having to further refine the argument. Which, of course, only makes it even stronger and more conclusive than before.
Here’s the problem. Biologists are stupid and innumerate, so even when they do the hard work required, they neither analyze it nor report it in a coherent, relevant, and meaningful manner. As I pointed out in Appendix F of the first edition, the 1,600 per generation number was both a) wrong and b) garbage. The problem is that every number reported in the 2009 and 2017 papers are also irrelevant garbage. The 2009 numbers are particularly bad because many of the fixations counted were not actual fixations.
Never mind, of course, that the actual number is almost entirely irrelevant, because there isn’t a bacterial fixation number that will ever make the numbers work. But because I value precision and I am not a biologist, I’ve downloaded the data for all 60,000 generations of the 80,000 generations published – 40k more than Barrick utilized in the 2009 study – and had Athos crunch the data. I’ll put everything in the next MITTENS 3.0 paper, but here are the highlights:
Kimura’s k = μ is mechanistically irrelevant
Kimura’s neutral substitution rate k = μ gives the correct long-run average but the MECHANISM producing neutral fixations on observable timescales is not drift — it’s hitchhiking. With Ne = 3 × 10⁷, drift takes ~30 million generations per neutral fixation. The LTEE ran for 60K generations and accumulated ~20 neutral fixations — all hitchhikers. The drift channel isn’t slow; it’s off.
Supermutation increases noise, not signal
Hypermutation increases clone mutation counts ~19× but total fixation throughput only ~11×, and BENEFICIAL fixation rates don’t increase at all, they actually tend to decrease clonal interference overload. One population (Ara-2) achieved zero fixations despite 100× elevated mutation rate. Supermutation is loading more passengers on the same number of buses. Also, it’s obviously not applicable to humans, because we have no supermutational gene-pairs.
6c. Deceleration is real and expected
Early adaptation is fast due to low-hanging fruit and big-effect mutations. The rate decelerates as easy gains are exhausted. The 30K–40K interval for non-mutators actually goes NEGATIVE. The faster rates from earlier generations cannot be extrapolated to long evolutionary timescales.
The breeding reality principle
Even at the LTEE’s bacterial throughput rate with no recombination costs, no mate-finding, no overlapping generations, and essentially infinite effective population size, the mechanism fails by six orders of magnitude. For organisms with 25-year generation times, small effective populations, sexual reproduction, and limited offpsring, the situation is considerably worse.
The real, updated 60-generation LTTE numbers are:
4,615 generations per beneficial fixation
1,322 generations per all-cause fixation