Via Yann Klimentidis come this paper comparing patterns of polymorphism in mitochondrial DNA (mtDNA) and Y chromosomes from Native American populations. The premise is that mtDNA reveals the maternal demographic history (because mitochondria are only passed from mother to child) of a population and the Y chromosome reveals the paternal history (Y chromosomes are only passed from father to son). But we know that mtDNA is not a good marker for studying demographic history because there is no recombination in the mitochondrial genome. That means that evidence of demographic events will be swept away by natural selection anywhere on the molecule.
How about the Y chromosome? Most of the Y chromosome is non-recombining, although there is some genetic exchange with the X and some intra-Y chromosome recombination. The lack of recombination in the mitochondrial genome coupled with evidence for natural selection on that molecule make any mtDNA locus a poor neutral marker. The Y chromosome is thought to have 'decayed' do to recurrent fixations of deleterious mutations -- these alleles cannot be removed from a population because they are genetically linked to advantageous alleles elsewhere on the Y chromosome.
So, I ask you, is the Y chromosome a poor molecular marker for studying demographic history? If so, what do comparisons of polymorphism between the Y chromosome and mtDNA tell us? Note that the genealogical relationships (ie, the tree topology) should not be affected by the lack of recombination. But the shape of the tree will be dramatically altered every time natural selection alters the frequency of an allele. This shouldn't affect our ability to determine maternal and paternal relationships amongst close relatives; it will affect our ability to say whether there was a recent change in population size (ie, within the last
2N 1N generations1) and determine gene flow between populations.
1- It's 2N not 4N because it's a haploid genome, right? [Edit: As John McDonald points out in the comments, it's 1N because the Y-chromosome is only found in males. Duh! Looks like I've gotta retake popgen.]
It would be 1N, because there are 1/4 as many Y chromosomes as autosomes. If the variance in offspring number is higher for males than females (which seems plausible), effective population size would be even less for Y chromosomes.
And I agree with your main point--I don't know why anyone thinks a demographic analysis of mtDNA or Y-chromosome data that is dependent on complete neutrality of every mutation is worth doing. Even if someone is willing to assume that adaptive mutations are rare, there's got to be a ton of background selection.
Concerning mtDNA and Y chromsomes in human migration, sweeps of either marker would distort the relationships between lineages within haplotypes, thereby affecting interpretations of (1) age of initial migration/settlements, and (2) historical population size of the past (I expect the genetic variance based in these markers would be less due to sweeps). But how bad is this when dealing with human evolution/migrations? Since I deal with Melanesia, I am wondering if this is huge factor in outlining the genetic past of the region?