I would describe what you’re talking about not in terms of parsimony, but in terms of the prior probability you’re assigning to different possibilities. (An assignment I think is reasonable – but someone with a strong belief in the literal truth of Genesis might assign a different prior.)
Agree with glipsnort that Bayesian priors are not the same as parsimony.
In general people say that when all else is equal one should choose the most parsimonious hypothesis. As soon as you consider a case where A is most parsimonious, but there is a little bit more evidence in favor of B than in favor of A, you can see the lack of clarity in all this. Exactly how much evidence does it take to override parsimony and conclude in favor of B? Chris Falter made it sound as if one requires a vast amount of evidence in favor of B to do that. But when people say “when all else is equal” they are implying that any tiny amount is enough to invalidate the use of parsimony.
I’m leaving aside the use of “parsimony” as a justification for minimum-steps methods for inferring phylogenies. There the declaration that the count of number of changes is a numerical definition of parsimony is, to my mind, arbitrary. It is a whole different discussion, which I spent years working on. Don’t get me started on that …
After this article: A First Couple? Here’s the Backstory | Evolution News. Here is another article from @Agauger responding to us here.
Why did most think a bottleneck of two was impossible? I can’t say for sure, but at least in Dennis Venema’s case it seemed to be his confidence in the many reported effective population sizes of 10,000. What he missed was the fact that averaging over millions of years would mask a sudden sharp bottle neckdown to two individuals if recovery is rapid.
She is absolutely right here.
Now over at Peaceful Science they are discussing this paper. I am pleased to see it. And I can even answer a few questions right away, such as why we didn’t use a more powerful tool such as ARGweaver .
The answer is simple: built-in evolutionary assumptions. We tried to keep our math as free as possible from assumptions about evolutionary history.
Well that is an odd objection. Argweaver doesn’t have a built in assumption like this.
That is also why we didn’t use the derived frequency spectrum, which would have entailed comparing chimp and human sequences to determine which human allele was likely to be ancestral and which mutant (using the chimp as the outgroup). If no evolutionary relationship is there, what is the point in using chimps as an outgroup? It is permitted, in any case, to use a folded allele frequency spectrum in cases where no appropriate outgroup exists.
@glipsnort, this explains why they did not use the derived frequency spectrum. It looks like they are really going for a de novo creation model here. Unlike the GAE, however, in this case, they do fall prey to the “appearance of common decent” problem. Why did God create AE such that it appears we share common ancestry with the great apes?
There are multiple directions we could go. We may use something like ARGweaver in future papers, but we would want to compare its results with what we have now. We could test the effect of population structures, migration, and subdivision on effective population size and the time required to match current genetic diversity.
I am all for this. I’m just not sure what they gained from this study in the first place.
For Joshua Swamidass: Actually, we mentioned HLA, which is home to many examples of putative trans-species polymorphisms. This an area I would like to take up because I think it is the way forward and may provide some answers to our origin that other regions of the genome can’t. More later.
Yes, I do think this is where things well get interesting.
As for ghost lineages, as far as I can tell they will shift things further in the past. No problem. Remember, I’m OK with 2 million years!
Yup, I think @Agauger’s willingness to go that far back really makes her position safe from falsification by everything but, possibly, trans-species variation.
@Joe_Felsenstein, she includes a discussion of Parsimony too. I think she wasn’t investing that word with much weight.
Well, as you point out, it’s not safe from falsification by any gene you care to name, since they all show that humans are related to other apes. A bottleneck of 2 (or perhaps a dozen or so, given those cross-species polymorphisms) may not be falsifiable, but de novo creation of genetic A&E certainly is, absent the auxiliary hypothesis of a deceptive deity.
Well yes. That “safe from falsification” was in a particular context.
I do think that the issue of “appearance of common descent” is not necessarily a deal breaker, but it raises serious theological questions. Those questions have to be articulated and engaged in a sensible way for them to make their case. As far as I can tell, they do not do this here, or elsewhere.
How would one even start?
A post was split to a new topic: A YLC is Bemused At Others Engaging Evidence
I assumed they were going for de novo creation. That’s no reason to use the folded spectrum – you don’t have to accept common descent to do so. Just call one allele the ‘non-chimp allele’ if you don’t want to call it ‘derived’. Whatever you call it, it’s an empirical fact that the frequency distribution of the non-chimp allele continues to decline past 50% and therefore imposes a constraint on how much initial diversity there can have been in this model. (Or maybe it imposes a constraint – hard to know without looking.)
Along those lines. . . In their model(s), primordial diversity (diversity appearing in the first couple) contributes negligibly to low frequency alleles but represent more than half of alleles with frequencies around 50%. One might guess, therefore, that the genetic variants in the two frequency ranges would have different characteristics. Of course, they don’t. Here’s the breakdown of the different classes of single-base substitutions for high frequency (40-50%) and low frequency (<1%) alleles.
The all look like mutations.
Sorry, I got the labels wrong on those plots. The classes in order are
- transition, non-CpG
- transition, CpG
- G-C, non-CpG
- G-C, CpG
- AC/GT, non-CpG
- AC/GT, CpG
@glipsnort, that is a very good and interesting test of the genetic-bottleneck-de-novo-creation model they are proposing. That is worth writing up more carefully so people can understand it better. Looking forward to seeing the numbers in a table and the revised graphs.
Does this paper shows that the Mathematical Conjecture is false and that there wasn’t a single pair of humans for over a million years?
Very nice study. It also provides no support for the so called Venema conjecture. He equivocated average with minimum. Effective population size, it turns out, is not a minimum.
But doesn’t the study lead one to accept that the mathematical conjecture of a human bottleneck of precisely two is highly unlikely to such an extend that it can be ruled out as not historically accurate?
No, it does not tell us it is unlikely unless the prior says it is unlikely. The evidence does tell us one way or another.
The motivation for these plots is that different kinds of mutation happen at different rates, which means that we should see more of some kinds of genetic variant than other kinds, if the genetic variants are the result of previous mutations. What’s shown in the plots are different classes of single-base differences and how frequently they occur in a human population. The details of the different classes are explained in a blog post by me, here.
I have made the plots for two different frequency bins; that is, one plot show variants that are rare in the human population (0.5% to 1.0%) and the other variants that are common (40% to 50%). (I pulled the variants for the different frequencies when @evograd was preparing this post.) According to the model that is the subject of this thread, rare variants are all the result of recent mutations, and should show the characteristic pattern expected of mutations – which they do. In the same model, however, most high frequency variants were created along with the original couple – but they also show exactly the same pattern, which is what we would expect if the human lineage is older than 500,000 years and high frequency variants are also the result of mutations.
This is the part that I would like to hear more about. Can you elaborate some on this statement and this kind of reasoning?
Can I? Sure. I’ll be happy to do so as soon as you explain why genetic differences between humans and baboons show the same pattern as genetic differences between humans and chimps, despite there being five times as many of them. (I wouldn’t be a very good steward of my time if I spent it on explaining things to someone who didn’t want to engage with the issues, would I?)
I think I missed these point from @Agauger:
So why make “parsimonious” a threatening word? Maybe I should have said “simple” instead? I think they all missed the point of the title. Our whole point was not that the single-couple model was true or to be preferred, but merely possible. Just possible. 500,000 years ago, or 2 million years ago, with ad-mixture, without, still possible. Unique or from a bottleneck, take your pick, the model doesn’t distinguish. All we set out to show was that it was possible to account for current diversity starting from two. Which, by the way, Josh Swamidass has kindly reminded me, he has already done. Now we have done it a second way.
First off, thanks to @Agauger for acknowledging the TMR4A work. Also she substitutes “simple” for “parsimony” which I have to think about.
Neither word seems like what she’s trying to say. Why not just “possible” if that’s what she means?
That makes sense to me. Or “not ruled out” by the genetic data?