Read on Twitter
Up now @JEvBio, @BenGFreeman1 & @CASBirdman & I show that elevational series of New Guinea kingfishers in replicate isolated mountain ranges did NOT undergo parallel ecological—but that divergent selection & gene flow may be playing a role anyway (1/n): https://onlinelibrary.wiley.com/doi/epdf/10.1111/jeb.13698
The correlation between series of congeneric species with adjacent but nonoverlapping elevational ranges and strong climatic stratification in tropical mountainsides has long attracted interest as a possible stage for that slipperiest of processes, parapatric speciation (2/n)
Dating at least to Chapman (see Kattan et al. 2016: https://doi.org/10.1111/bij.12720), this compelling idea has undergone recurring bouts of enthusiasm and skepticism over the past century (3/n)
For example, Archbold expedition ornithologist A.L. Rand thought the process was unlikely to explain much of anything in the New Guinea avifauna, something his slightly more famous contemporary Ernst Mayr agreed with (though note the exceptions!) (4/n)
Fast forward to the 1990s, when the advent of molecular phylogenetics provided an opportunity to more rigorously test mechanisms behind the origin of elevational series (5/n)
In genera (/taxa) w/ 3 or more lineages, a basic prediction of a casual role of adaptation across mountainsides in driving speciation is that sister relationships should reflect adjacent elevational zones (6/n)
In other words, given species A found from 0-1000m, species B found from 1000-2000m, species C found from 2000-3000m, and species D found from 3000-4000m, a topology of (A,(B,(C,D))) would provide some degree of support for parapatric speciation...(7/n)
...while a topology of (D,(B,(A,C))) would seem more consistent with a model of allopatric speciation followed by chaotic range shifts and secondary contact (8/n)
There are limits to this approach, of course, as anyone who has read @JLosos' essential review on phylogenetic comparative methods and the geography of speciation is well aware ( https://doi.org/10.1016/S0169-5347(03)00037-5)—namely, that ranges shift, a lot! (9/n)
So to REALLY be confident that elevational series formed via parapatric speciation, you'd probably want evidence on a level w/ that for sympatric speciation in cichlids—sister relationships between lowland & highland spp. in *replicate*, geographically isolated mountains (10/n)
To my knowledge the first application of this framework in tropical mountains was in Patton & Smith 1992 ( https://doi.org/10.1111/j.1558-5646.1992.tb01992.x), who rejected parapatric speciation (or the "gradient model") in mice in the northern vs. southern Andes (11/n)
Since their paper, both this "parallel speciation" test and its simpler (A,(B,(C,D))) cousin have been applied to many plants and animals in the tropics worldwide, including by many talented and inspiring colleagues (12/n)
Of particular note is a recent book chapter by @lcespedesarias and @cdanielcadena ( https://evolvert.weebly.com/uploads/8/3/3/2/83324532/neotropical_diversification.pdf), which not only does an excellent job of summarizing many of these studies...(13/n)
...but also spotlights relationships among elevational replacements in the magnificent Scytalopus tapaculo "radiation" of tiny, nearly flightless birds. And what did they find? Nearly universal support for allopatric speciation and range shifts (14/n)
In fact, a quick skim of the literature gives the overwhelming impression that there is barely a SKIM of evidence for parapatric speciation across tropical mountains in birds, and not much more in reptiles, insects, or plants (15/n)
So why on earth, then, did we think that the case of Syma kingfishers in New Guinea might be different? (16/n)
Besides A.L. Rand's endorsement (see above), we had a handful of reasons (17/n):
First, unlike most previous studies of avian elevational replacements w/ fixed morhological differences, taxonomic limits in Syma are (were) v. fuzzy—they really do seem more like ecotypes than "real" species or subspecies...(18/n)
...highland populations are a bit larger (Bergmann's rule? 
), they have a slightly lower pitched call, a single mountain pop has a little bit of a darker bill, and there's a weird island population that we'll come back to. That's it. (19/n)
), they have a slightly lower pitched call, a single mountain pop has a little bit of a darker bill, and there's a weird island population that we'll come back to. That's it. (19/n)
Second, New Guinea has a different geography & geologic history than the Andes. In particular, the sky island mountains of the Huon Peninsula are extremely young and have never been connected by montane forest to the adjacent Central Ranges...(20/n)
...functioning, in our eyes, much like an isolated crater lake would in the cichlid speciation analogy (21/n)
Third, kingfishers generally seem pretty bad at colonizing the highlands (they are an overwhelmingly lowland group, and Syma megarhyncha is the only montane specialist kingfisher on New Guinea), so we thought selection pressures might be particularly strong (22/n)
With this in mind, we collected a handful of fresh tissues, many more toepads from museum specimens, sequenced DNA a few different ways, measured specimens, and aggregated all known recordings of the species (23/n)
We then:
1) inferred a gene tree from mtDNA
2) described nuDNA clusters
3) described the distribution of phentoypic traits and
4) tested alternate demographic models (24/n)
1) inferred a gene tree from mtDNA
2) described nuDNA clusters
3) described the distribution of phentoypic traits and
4) tested alternate demographic models (24/n)
What did we find?
1) highland and lowland mtDNA linages are reciprocally monophyletic (at least if you ignore a weird species-level insular population of the lowland taxon)
2) all three form distinct nuDNA clusters
3) phenotypic traits are multimodal but overlapping...(25/n)
1) highland and lowland mtDNA linages are reciprocally monophyletic (at least if you ignore a weird species-level insular population of the lowland taxon)
2) all three form distinct nuDNA clusters
3) phenotypic traits are multimodal but overlapping...(25/n)
...& 4) the best fit model is one of divergence in allopatry ~650 kya followed by ~250 ky of secondary contact at a little over 1 migrant per gen
