Language Log

Scottish dialect genetics

For some reason, the worldwide excitement over English cow dialects hasn't connected with the more localized excitement over Scottish crossbill dialects, which was also recently featured on the BBC News web site ("'Accent' confirms unique species", 8/15/2006):

Debate has raged for years among experts about whether the Scottish crossbill was unique, or a sub-species of the common crossbill.
[...]
DNA tests had shown the Scottish crossbill, common crossbill and parrot crossbill - which visits from Europe - to be genetically similar.
The results of long-running research has now found, according to the RSPB, that the Scottish variety is a distinct species of its own.
The society said it had a "Scottish accent", or call, which it uses to attract a mate from among other Scottish crossbills.

The logic here is puzzling. Cows, the BBC told us, learn their regionally distinctive moos from the farmers that tend them. (Now in fact, there's no evidence for -- or against -- regional variation in cow vocalizations. The whole thing was an empirically vacuous PR stunt. But we're talking about logic here, not evidence.) Part of the argument for the plausibility of the cow story was the well-known fact (it really is a fact) that many species of birds learn aspects of their songs, and sometimes thereby develop local song "dialects". But in this other story, separated by only a few days, the BBC tells us that Scottish crossbills, though "genetically similar" to their European cousins, are now to be treated as a separate species, because they have a "dialect":

RSPB Scotland's senior researcher Dr Ron Summers, who led the study, said: "The question of whether the Scottish crossbill is a distinct species, and therefore endemic to the UK, has vexed the ornithological world for many years and split the bird watching community.
"This research proves that the UK is lucky enough to have a unique bird species that occurs here and nowhere else - and this is our only one."

Well, maybe the crossbills are not among the bird species that exhibit vocal learning, but instead develop songs that are genetically programmed in every detail. That would rescue the logic of the story, but its author shows no sign of having considered the question one way or another, which is a little odd, since the word dialect suggests social construction rather than genetic determination. (In fact, crossbills -- a kind of finch -- do learn their songs. See below for some details.)

So I spent a few minutes finding and reading the press release at the site of RSPB Scotland. (That's the Royal Society for the Protection of Birds.) This revealed where the BBC's take on the story came from -- as in the case of the regional cows, they were basically just repackaging the PR agent's press release (with a good deal of straight copying).

Quaintly, the RSPB has a picture caption that reads:

'Celtic' crossbills differ in bill size from other crossbill species found in Britain, and just like native Scots, they have also been found to have a distinct Scottish accent or call.

And it's well known that Scots (or Homo edinburgensis as scientists call them) are a distinct species. The RSPB press release has more about the three types of crossbills:

Scotland's conifer woods are home to three types of crossbill -
* the common crossbill (with a small bill best suited to extracting seeds from the cones of spruces)
* the parrot crossbill (with a large bill suited to extracting seeds from pine cones)
* and the Scottish crossbill (with an intermediate bill size used to extract seeds from several different conifers).
All three are similar in both size and plumage, and DNA tests have showed that the birds are genetically similar, casting some doubt on the Scottish crossbill's status as a distinct species.

In fact, you can't really tell the different kinds apart just by examining them:

Although the three species differ in average bill size, the actual differences are small and cannot be used reliably in the field by ornithologists to identify crossbills.

But wait:

The calls, though, can be distinguished by sonograms, or sound pictures, made up from recordings. Crucially, this provides the basis for a method to survey crossbills and, for the first time, gain a clear picture of their numbers and distribution in Scotland.

So the RSPB did a "long term field study" in which they captured 46 mated pairs of crossbills of various types, to learn "if the birds mate with those with a similar bill size and call, and whether young Scottish crossbills inherit their bill sizes from their parents".

Results showed that, of 46 pairs of different types of crossbills caught, almost all matched closely for bill size and calls. In other words, the different types of crossbills were behaving as distinct species.
The small number of 'mismatched' pairs was too few to suggest that the different types are not species, but enough to account for their genetic similarity. The fact too that young crossbills had bill sizes similar to their parents showed that they inherited their bill sizes, and also supports the species status of Scottish crossbills.

Gee, I bet you could use a similar technique to demonstrate that various ethnic groups in the U.S. are separate species.

The RSPB press release tells us that "Scottish crossbills (as identified by bill size) also have quite distinct flight and excitement calls from other crossbills", but unfortunately, neither the press release nor its replication at the BBC tells us what crossbills' calls are like in general, and how the Scottish-dialect version differs. The Scottish crossbill page on the RSPB website says that it has "[a] 'chup chup' call with a fluty quality", whereas the common crossbill has "[a] loud 'chip chip' call; a warbling, twittering song", whereas the parrot crossbill has "[v]ery similar calls to crossbills, but thought to give a distinctive deep ‘kop-kop’ and ‘choop choop’".

Since the RSPB has no equivalent of the International Phonetic Alphabet at its disposal, I'm puzzled about the difference between the Scottish crossbill's "chup chup" and the parrot crossbill's "choop choop". We want spectrograms and scatter plots! (There are audio samples on the RSPB site, but only one recording per "species". I presume that the RSPB's field study has been published somewhere, but I haven't tracked it down yet.)

The RSPB's "long term field study" now must be supplemented with a larger and longer field study:

The next steps in the Scottish crossbill study are to find out its population size and habitat requirements.

With the current estimate of 1,500 birds for its global population, being little better than a guess, a detailed survey is crucially important to put together the right conservation and management measures to protect and conserve it.

Dr Jeremy Wilson, head of research for RSPB Scotland said, 'Clarifying the status of the Scottish crossbill as a distinct species, and devising a survey method based on the bird's calls are exciting steps forward.

'We hope to carry out the first full survey of the numbers and distribution of Scottish crossbills in 2008, after which we will be better placed to understand how best to manage conifer woodlands in Scotland to secure the future of a bird found nowhere else in the world.'

Another, more cynical, argument for the RSPB's conclusion begins to suggest itself. Echoing Max Weinreich's observation that "a language is a dialect with an army and a navy", we might suggest that "a species is a phenotypic variant with a protected habitat". (Or take a look at the Wikipedia article on species for a sketch of the reasons why the concept "species", up close, is just about as contested as the concept "language".)

But if you ask me, the crossbills are getting their "accents" from those Scottish birders. At least, that's what the cow-dialect theory tells us.

[Hat tip: Edward Wilford.]

[Update: a few minutes with Google Scholar establishes that crossbills are indeed among the bird species that exhibit vocal learning. Thus P.C. Nundinger, "Call Learning in the Carduelinae", Systematic Zoology, 28:3 270-283 (1979). From the abstract:

Experiments and field observations document vocal imitation in six cardueline species representing four genera. Flight call learning was found in all birds studied; learning of many other call types was observed in two genera. Additional evidence extends call learning to other carduelines bringing to eight the total number of genera in which call learning has been observed. Call learning is perhaps characteristic of subfamily Carduelinae, and the taxonomist should consider the possibility that learning may affect the patterns of all adult cardueline vocalizations. The taxonomic value of cardueline calls in particular and passerine calls in general is re-examined in light of this extensive call learning.
[emphasis added]

From the body of the article, commenting on an experiment with call sharing in a pair of captured white-winged crossbills (Loxia leucoptera), who were mated by the experimenter's choice rather than their own:

I conclude that the call sharing exhibited by this pair of Loxia leucoptera is not due to chance but is the result of vocal imitation, and that vocal imitation affects most, or even all, of the call types in this species.

J.G. Groth ("Call matching and positive assortative mating in Red Crossbills", Auk, 1993) studied appalachian crossbills (Loxia curvirostra), and offers a plot of calls from members of 24 mated pairs:

Fig. 1 Audiospectrograms for call notes of 24 pairs (labelled A-X) of crossbills in Virginia. For each pair, call note of male illustrated on left and female on right. Short horizontal marks along vercial axes are at 2, 4, and 6 kilohertz, and width of each box represents 140 msec.

The overall conclusion of this paper is that

These findings are consistent with the hypothesis that distinctive forms of crossbill represent reproductively isolated groups (i.e. species).

However, his main evidence is there is a very wide range of body and bill sizes and shapes, and the shapes and sizes of mated pairs are highly correlated. The call types also match, but as he observes:

The Appalachian crossbills also showed a pattern of assortative pairing based on acoustic characters, but this observation is trivial because call matching was a prerequisite for identification of birds as mates. No information is available on the structure of the calls of these birds before they became associated with their mates.

Indeed, the very wide range of exactly-shared call types makes it seem unlikely that the call sharing is entirely genetic, though Groth's experience with captive crossbills was different from Nundinger's:

In two captive pairs with mates having intitially different call structures that produced nests and successfully fledged young, the mates never matched each other's flight calls.

(though this may have been because of the age at which the birds were captured). The role of genetics vs. vocal learning is left unclear:

The process by which crossbills choose their mates is not known. Bill size correlates with conifer preference in crossbills, and calls could function as signals giving information on morphology and, therefore, habitat preference, of individuals. A question that remains is whether vocalizations, visual assessment of morphology..., habitat preferences, or combinations of these and/or other cues provide the basis for mate choice in crossbills.

Some other background: Craig W Benkman "Divergent selection drives adaptive radiation of crossbills", Evolution, 57:5, (2003)

[C]rossbills are a recent adaptive radiation where the processes involved in population divergence may still be active. Red crossbills in North America are categorized into nine call types that are recognized by distinct vocalizations. At least seven of these call types are specialized for foraging on different species of conifers that hold seeds in partially closed cones through winter, and reproductive isolation is evolving between populations that have diverged in the last 10,000 years.

Also see Craig W Benkman, "Reciprocal Selection Causes a Coevolutionary Arms Race between Crossbills and Lodgepole Pine", The American Naturalist, 162:182-194 (2003).

And Sophie Questiau et al., "Phylogeographical evidence of gene flow among Common Crossbill (Loxia curvirostra, Aves, Fringillidae) populations at the continental level", Nature, 83:2 (1999):

Common Crossbill subspecies have been described according to morphological traits, vocalizations and geographical distribution. In this study, we have tried to determine whether the subspecies correspond to clear-cut mitochondrial DNA lineages ... We find a mixing of the mitochondrial haplotypes at the continental level among the different types or subspecies previously described. Morphological differentiation (in bill size and shape essentially) shows the possibility of rapid local adaptation to fluctuating resources (coniferous seeds), without necessarily promoting the development of reproductive barriers between morphs.

No doubt this was somehow taken into account in the RSPB study of Scottish crossbills, but it would be nice to know how. And the whole business seems like a good opportunity to raise issues of evolution, the meaning of the concept "species", and so forth. The position exemplified by the RSPB work seems to be that more-or-less reproductively isolated populations, with behavior that is somewhat differentiated from nearby groups, and average differences in some morphological characteristics (even if the distributions overlap a great deal), are separate species. By this definition, wouldn't Indian castes be different species?]

Posted by Mark Liberman at August 27, 2006 08:15 AM