The Multiregional Stipulation Society23 Sep 2005
This morning a good friend wrote me to ask about the "multiregional stipulation of random mating and constant population size." What an odd thing for anyone to say, I thought. No anthropologist that I know of thinks that humans mated randomly in the past, or that the population size has been constant. But, as usual, I felt a sinking feeling: if somebody was sending along this weird sentence, it sure must have been published somewhere. And Science comes out on Friday....
Then I found the source for the quote. It was written by Vincent Macaulay and colleagues in this week's Science. The issue includes a letter from Henry Harpending and Vinayak Eswaran (2005), regarding two papers that appeared earlier this year on mtDNA migration through southern Asia. The authors of each of those two papers have a response along with the letter, and Macaulay et al. (2005b) is one of these responses.
Reading these letters is what set off my thoughts about pork barrel paleoanthropology, which became long enough I split them off into a separate post. But the letters certainly provide a good illustration: one asks a very simple question; the others respond in ways that somehow miss the key issues.
Harpending and Eswaran (2005) raise a very simple point: the data aren't all consistent with the preferred explanation; some account must be found that will explain all the data; the papers ignored this central issue in their interpretations.
I think their letter is pretty subtle, and certainly more so than what I wrote about the papers after they came out, where I pointed out some of the data they ignored. And I noted the complete lack of confidence intervals on their estimates. Heck, rereading what I wrote about these papers reminds me just how many problems they had, especially Macaulay et al. (2005). That paper tried to construct an overarching theory -- based only on mtDNA -- about how humans sailed along the coast of the Indian Ocean to settle Southeast Asia, ignored any piece of information that might contradict the story, and suggested that the archaeological evidence that could support their story must have been buried beneath the waves.
Here is part of the response by Macaulay et al. (2005b) to Harpending and Eswaran's letter:
Existing autosomal data do not, in most cases, provide strong evidence for either replacement or hybridization, despite claims to the contrary. The high coalescence time of autosomal loci is not relevant, since in itself this tells us almost nothing about more recent settlement events: A small founder gene pool could well have either a deep or shallow ancestry within a replacement perspective. Given the limited amount of variation in nonrecombined stretches of the autosomes, there is typically little power to distinguish different demographic models. This is as true of the autosomal data used by Templeton (3) as it is of the data in the papers cited by Harpending and Eswaran. The authors of the most recent of these (4) are entirely open about this, but their (frequency-based) suggestion that the root of their tree lies in Asia is mistaken; there is simply insufficient branching structure to fix the geographical location of the root with any confidence. Moreover, supposedly ancient Asian-specific singlenucleotide polymorphisms such as those cited by Harpending and Eswaran are associated with age estimates of enormous uncertainty.
Bold conclusions of ancient Asian ancestry also suffer from limited sampling. Non-African mtDNAs most likely evolved in the Horn of Africa and dispersed from there, but none of the cited papers on autosomal loci include data from this region. Even if such data were available, identifying non-African founder lineages in such low-resolution systems is deeply problematic because of recent back-migration across the Red Sea (5). In cases where autosomal loci do have the necessary resolution, they suggest the replacement model (6-8). The discordant population-size estimates referred to by Harpending and Eswaran are likely more apparent than real, since these long-term values are usually obtained with the multiregional stipulation of random mating and constant population size. The analysis of overly simplistic models with methods that throw away what little information there is in most of these loci throws up straw men, such as the apparent lack of "strong signals of expansion" in some autosomal loci (9) (Macaulay et al. 2005b:1996).
This strikes me as an extraordinary amount of Rube Goldberg-like complexity. Yes, a replacement is true if:
- (a) we disregard entirely the lack of massive population expansion from nuclear genomic data [this is the part of Harpending and Eswaran's letter that there is no answer to in the response];
- (b) the evidence from autosomal data that suggests a long phylogeographic history for non-African populations mysteriously (i.e., without reference) has "little power";
- (c) "frequency-based" suggestions that the root of evolutionary trees lie in Eurasia are mistaken;
- (d) if only the Horn of Africa were sampled better;
- (e) even when you find "founder mutations" from Eurasia, they are really African because of recent back-migration across the Red Sea;
- (f) you avoid any "multiregional stipulations".
Aside from the stipulation problem, I found one other thing worthy of mention. The response cites three papers in support of the idea that nuclear and mitochondrial genetics sometimes show the same result. It apparently escaped their notice that Harpending is a coauthor of one of these papers, and edited another. He might be expected to be aware of their contents, in other words! Could it be, perhaps, that there are some newer papers that show a different pattern? Where the evidence might have more than a "little power"?
Alonso S, Armour JAL. 2001. A highly variable segment of human subterminal 16p reveals a history of population growth for modern humans outside Africa. Proc Nat Acad Sci USA 98:864-869. Full text online
Harpending HC, Eswaran V. 2005. Tracing modern human origins. Science 309:1995-1997. Full text (subscription)
Macaulay V, et al. 2005a. Single, rapid coastal settlement of Asia revealed by analysis of complete mitochondrial genomes. Science 308:1034-1036. Science Online
Macaulay V et al. 2005b. Tracing modern human origins. Science 309:1995-1997. Full text (subscription)
Thangaraj K, et al. 2005a. Reconstructing the origin of Andaman Islanders. Science 308:996. Science Online
Thangaraj K et al. 2005b. Tracing modern human origins. Science 309:1995-1997. Full text (subscription)
Tishkoff SA et al. 1996. Global Patterns of Linkage Disequilibrium at the CD4 Locus and Modern Human Origins. Science 271:1380-1387. Summary
Watkins WS et al. 2001. Patterns of Ancestral Human Diversity: An Analysis of Alu-Insertion and Restriction-Site Polymorphisms. Am J Hum Genet 68:738-752.