Natural selection in real time: birds of a feather don’t evolve together

June 10, 2011 • 6:06 am

Want to see evolution in action—in “real time”? A new paper in Ecology by André Desrochers on songbirds (access free) might fill the bill. (Excuse the pun.)  It shows—I’d prefer to say “suggests strongly” since there are a few problems—that the shape of feathers in North American songbirds has evolved over the last century in response to changes in patterns of forestation.

Here’s the idea: in the last hundred years, North American forests have changed drastically.   The boreal (i.e. high-latitude subarctic) forests of eastern North America have been cut back heavily, replacing old coniferous stands with younger deciduous ones.  Temperate non-forest habitats have also become more fragmented.  Conversely, the temperate forests of eastern North American, severely deforested in the 19th century, have reversed this trend, undergoing “afforestation” in the 20th century.  Afforestation also characterizes boreal early-successional forest.   I can’t vouch for these generalizaitons as I’m not an ecologist, but the authors support them with references.

It’s also known that birds with “pointier” wings have more energy-efficient sustained flight, and that pointiness (we’ll define it below), can evolve rapidly in birds.  If your habitat becomes more fragmented, it would be advantageous to evolve pointier wings to travel more efficiently between distant foraging and resting places.  Conversely, if your habitat becomes less fragmented, you should lose those pointy wings, which impose energy costs in takeoff; and ounder wings are also better for foraging in thick vegetation or close to the ground.

From these observations Desrochers made the following hypothesis (from the paper):

I tested the following predictions: over the last century, species mostly found in boreal, mature, coniferous forests and temperate non-forest habitats evolved more pointed wings in response to increased fragmentation, whereas species associated with temperate mature forests and boreal early-successional forests evolved less pointed wings because of relaxed selection for mobility. Additionally, I examined whether the above predictions were better supported in nonmigratory species than in neotropical migrant species.

Migrants should show less changes since they have the constant selection (unchanged over the century) for having wings appropriate for their yearly long-distance round trips.

Desrochers measured 21 species of birds (average 40 specimens per species) collected between 1900 and 2008; all were from collections at Cornell University at the Canadian Museum of Nature.  This enabled him to test for any long-term changes feather shape that could reflect evolution.

How did he measure pointiness? Here’s a female scarlet tanager showing how the measure was made (on the right wing only):

(a) is the distance between the carpal joint of the right wing and the distal end of the outermost secondary feather.  (b) is the distance between the same joint and the wing tip.  “Pointiness” is the index 100 X (b – a)/a, in other words a measure of how much, relatively, the wing tips extend beyond the secondary feathers.  This is called the “primary projection” in bird argot.  Desrochers also took an unrelated measure (bill length) just to see if other morphological traits might also have changed, indicating perhaps other selective pressures besides flight.

The results?  Pretty convincing:

  • Of the 21 species (there were actually 22 sets of measurements, for the red-breasted nuthatch, Sitta canadensis, was measured from both boreal mature forest, expected to select for pointier wings, and temperate mature forest, selecting for rounder wings), nearly every one changed in the direction predicted from its habitat. Of 12 species from temperate mature forest and boreal open habitats, eleven “evolved” rounder wings, as expected.  Of the ten species in boreal mature forest and temperate open habitats, all ten evolved pointier wings over the last century—also as expected given the habitat fragmentation.  These directions of change alone, regardless of their magnitude, are statistically significant.  Desrochers’ analysis also shows that eleven of the 22 trends were also statistically significant unto themselves, though I actually count 12 from his table.  At any rate, this is a very strong confirmation of his hypothesis.
  • The one species tested in both types of habitats, the red-breasted nuthatch, showed divergent evolution, as expected, evolving pointier wings in boreal mature forest and rounder wings in temperate mature forest.
  • Migratory status wasn’t consistently correlated with evolution (we expect migrants to show less evolution), but it was in boreal forest birds, with pointiness increasing more in residents than in migrants.
  • There wasn’t anywhere near this degree of changes in beak shape, which changed in only five mature boreal species (getting longer); and that change was of borderline statistical significance.

So, is this a good case of evolution in real time–in only three human generations?  I think so, but there are a few problems.  The most significant to me—and this is always the first thing that strikes me as a geneticist—is that there is no evidence that this change over time rests on changes in the frequencies of the birds’ genes.  Many ornithologists (and ecologists) often assume that if they see an animal change size or shape over a few generations, that change must automatically be genetic, and therefore the result of evolution via either natural selection or genetic drift.  But of course the change could be purely “developmental” or “phenotypic,” reflecting not genetic change but a purely developmental response to some unknown environmental change.

That’s not pure speculation, for there are plenty of examples.  The average height of Japanese, for example, has increased dramatically relative to Americans in the last generation.  Perhaps a Martian zoologist would, like some ecologists, attribute this change remarkably rapid evolution of increased height in the Japanese, probably due to natural selection.  But that’s wrong.  The height increase is not based on genes—it couldn’t be, for it’s happened way too fast. It resulted purely from an environmental change: the improved diet of the Japanese after the Second World War, which made them grow larger.

Many animals and plants have the ability to change their body shapes and appearances due to environmental circumstances (flamingos, remember, only become pink if they eat crustaceans and algae, incorporating the carotenoid pigments into their feathers).  I could point out other examples of ecologists making this fallacious “it’s all genetic” assumption, but I don’t want to embarrass my colleagues. Suffice it to say that without stronger evidence, seeing a trait change over generations leaves the question open if it really is genetic evolution.  The way to test this, I suppose, would be to release banded birds from single broods into diverse forest habitats, and see if living in those different environments could change the pointiness of their wings.

Desrochers tries to explain away this problem by invoking the concept of “heritability”: that is, the degree to which variation in a trait can be transmitted faithfully from parent to offspring within one population:

A second alternative explanation is that changes in primary projection may simply reflect phenotypic, as opposed to genetic, change (Gienapp et al. 2008). However, body measurements are highly heritable, with narrow-sense heritability (h2) generally between 0.6–0.7 in the case of wing length. . .

But this appeal to heritability is completely wrong, as has been pointed out for decades by the likes of Steve Gould, Richard Lewontin, and many other geneticists.  Just because a trait can be heritable within a population living in one environment (that is, a proportion of the variation in that population rests on variation in genes) says absolutely nothing about whether the difference in a trait among populations living in different environments (like Desrocher’s birds) has a genetic basis. The heritability of height is substantial in the population of North American humans, but one could not have used that to say that the difference in height between pre-war Japanese and Americans must have been largely genetic.  There was an important environmental difference there, too: diet.  All geneticists know that measurements of a trait’s heritability are confined to a single population in a single environment, and cannot be used to say anything about the genetic basis of differences in that trait between different populations in different environments. (This, of course, is the whole basis for the blow-up about differences in IQ between human “races,” who may inhabit different cultural and educational environments.)  Maybe, then, the differences in wing pointiness reflect some environmental modification of bird wings produced in different types of habitat.

Anyway, let me cease this rant and just let it serve as a lesson to ecologists to avoid assuming that changes over time are automatically genetic (or evolutionary)—and to not buttress this conclusion by specious appeals to “heritability.”

To my mind, that’s the biggest problem with this paper.  Desrochers mentions a few others—changes in food type, for example—but those seem unlikely based on the lack of changes of bill configuration.

I probably have been too hard on Desrochers.  To be fair, I think that he really has shown evolutionary changes in bird feathers in the predicted directions.  It is my gut feeling (nothing more) that there are probably not many environmental factors that could change feather pointiness, and so this could be genuine evolutionary change in a short period. In that case, it really would be a kind of landmark study—worthy of inclusion in textbooks along with the Grants’ work on Darwin’s finches.  But oh, how much stronger it would have been with some genetic data! (The Grants did, by the way, have that genetic data!).  I suppose I’m a bit peeved that elementary considerations of population genetics are being swept aside (or misused, in the case of heritability).  Nevertheless, I greatly admire Desrochers’ paper, and really hope he has some other evidence that wing pointiness cannot easily be changed by environmental factors alone.



Desrochers, A. 2011.  Morphological response of songbirds to 100 years of landscape change in North America. Ecology 91:1577-1582.

h/t: Birds and Science, via Matthew Cobb

29 thoughts on “Natural selection in real time: birds of a feather don’t evolve together

  1. Regarding the Japanese, should one assume a diet is “improved” just because it makes people taller?

    1. I was just about to say something about that…

      It’s true that younger people are a good deal taller than their parents and certainly their grandparents–largely due to increased protein in the post-war diet. I’m taller than my mother-in-law, and most women her age, but shorter than nearly every teenager I meet these days. On the other hand, it remains to be seen whether the diet that has allowed increases in height will be able to compete with the long-lived generations who grew up on a lower-protein/very-low-sugar pre-war diet. My husband’s grandparents, for example–all born around the first world war– all lived to be about 94-95.

      Jury’s still out on the ‘health’ of the diet–but definitely correct that increases in overall height in younger generations are due to diet and not genetics.

      *Interesting* post! Red-breasted nuthatch:-)) (Did you mean to write “temperate mature forest…and boreal mature forest…”? 3rd and 4th lines of the first bullet point).

      1. Yes fascinating. I was forming the question about whether it was evolution or just variation in the population while reading the first part, that you tackled in the second part.
        My father was about 5’8″ & when he moved to Norfolk in the early 1960s he told me he was taller than most of the local people. The tallest local person was Sir Edmund Bacon(d.1982). The nobs were able to eat well when Norfolk labourers were only able to have dumplings made of dough as they were so poor in the 19th c. Perhaps there is a trade off – the bigger that you are the more calories you need.

  2. I didn’t like the title of this post because ‘Natural Selection’ always happens in ‘real time’

    As a mere layman I appreciate the points that you’ve raised here – it’s a privilege to be let in on the fine points of a discipline ~ It must be a great temptation for scientists to (unconsciously) reduce their standards when new evidence leans towards a favoured result.

    Also in the link I couldn’t see a mention of what precautions Desrochers (& Bélisle, Morand-Ferron & Bourque) took to ensure a random sample of birds was chosen for measurement. What if ‘pointier’ winged birds within a species in a given area have a behaviour that makes them more likely to be measured ?

  3. Would sampling from the approximately 850 museum specimens used in the study allow any genetic conclusions to be drawn reliably and at a reasonble cost?

    Would you have to have already identified the loci contributing to the state in each species separately?

    Sorry if the questions are ill informed – I suppose preservation could have degraded the DNA enough to make extraction impossible or really hard, and the sample sizes could be too small, especially if these are Quantitative Trait Loci…

  4. Interesting. Another question: could the changes in the habits of the birds lead to body changes? For example, humans who are put to work, say, loading ships for a living, would develop bigger shoulder muscles.

    If I remember correctly, Darwin said something about ducks in OOS; he noted that the ducks that walked more had bigger legs and those that flew more had stronger wings.

    1. Tim – The Grants are two people – Peter and Rosemary. Regarding your question, if Dr. Coyne is referring to the study I think he’s referring to, the Grant measured certain traits found in Darwin’s Finches (beak size, I believe) and found strong correlations between the measurements found in parents and the measurements found in offspring.

    2. The Grants had almost all the birds in their Galapagos study population banded, so they could measure the morphological traits on parents, offspring, etc., and use the correlations among these measurements to estimate the heritability. The most obvious is the regression of the offspring measurement on the parents’ measurement. The slope of this regression is equal to the proportion of the total variation in the measurements that is due to genetic variation. This estimate of heritability is an estimate within the measured population, and does not necessarily reflect the heritability of variations among populations.

  5. The most obvious environmental effect on feathers is the wear that occurs after molting, usually on the tips and edges. This wear, however, is well known to ornithologists, and Desrochers mentions that worn specimens were not included in the analysis. Wear is not a yes/no state, so some degree of wear may have affected the included specimens. To produce the clear results in the paper, there would have to be some systematic effect of wear. Flying in closed habitats means the birds’ wings probably strike the vegetation more often. But does this “sharpen” them, or blunt them?

    1. What about wear in the old museum specimens? Would the date of collection be known in all cases ie post moult or pre-moult?

      1. The date of collection is an important data element for museum specimens, and though a few specimens may lack such data, most should have it (especially ones collected in the 20th century).

  6. “I leave you with my admonition to ecologists: ….GIVE SOME EVIDENCE THAT THOSE DIFFERENCES ARE BASED ON DIFFERENCES IN GENES.”

    So a new era of intense collaboration between evolutionary geneticists and ecologists has emerged at UC??? 😉

    Seriously, though, this ecologist agrees, but points out that obstacles do exist: 1) Not many ecologists have a great deal of experience with molecular genetic techniques, 2) Cooperating with a genetics lab can be logistically challenging (not impossible, of course), 3) The “start-up” cost of equipping one’s own lab to do this, on top of all of the ecological work, can be prohibitive, etc. etc.

    Nevertheless, I’ve just started down this road with a new project. I would very much LOVE to back up any ecologically important trait changes with genetic evidence.

  7. In compensation, geneticists can sometimes measure natural selection without having the foggiest idea what produces it.

    Three cheers for collaboration.

  8. I’d like to know the magnitudes (not just the presence of statistical significance) for the differences. It might be that the differences are so small that subconscious measurement error could explain the result. The measurers should have been blind to the provenance of the birds and the hypothesis of the experimenter. (Perhaps they were–I have not seen the paper yet.) This might seem unnecessary, but I think experimenter bias (not fraudulent but completely unconscious) can be stronger than people generally recognize.

    1. As for the magnitudes, the paper is free.

      As for blinding, sure, but this is one “easily” repeated experiment.

    2. ” All measurements were made with digital calipers under a
      dissecting microscope. Specimens with apparent molt of
      flight feathers or extensive feather wear were infrequent
      (n , 30 specimens) and discarded from analyses. A
      random sample of 128 specimens was measured twice
      (nonsequentially, i.e., ‘‘blind’’), to assess measurement
      error. Median differences between measurements of the
      same specimens were 0.25 mm, 0.30 mm, and 0.16 mm
      for secondary feathers, wing chord, and total culmen
      respectively, yielding intra-class correlation coefficients
      (measurement repeatabilities) .98.9%. I estimated
      temporal changes in primary projection for single
      species with linear models using the following covariates:
      year of collection and sex. Mean temporal changes
      in primary projection for each of the four species groups
      were obtained from mixed-effects linear models, with
      year of collection as covariate, and species and sex
      within species as random effects (SAS Institute 2009). In
      the case of mature forest species, migratory status
      (migrant or not) was also included as a fixed-effect
      covariate (species from open habitats were all migrant).
      Mean changes in culmen length for each of the four
      species groups were assessed with year of collection as a
      covariate and species and sex within species as random
      effects. Model residuals were examined visually through
      diagnostic plots and no strong departures from normality
      or homoscedasticity were noted.”

      1. Thanks Torbjörn. While the variability of the measurements (when repeated on the same birds) was small, it was not negligible. These random measurement errors only establish a lower bound for unconscious experimenter error. I do not want to say that this particular result is due to unconscious measurement bias, but measuring soft, flexible things like this to sub-millimeter precision certainly is difficult, and is the sort of situation in which biologists should show more sophistication regarding the measurement process. Humans are not machines and unconscious experimenter bias should be routinely controlled in delicate experiments like this..

  9. ***who may inhabit different cultural and educational environments.) ***

    That is one of the reasons to consider there might be selection for a different distribution of traits? Greg Clark makes that argument in ‘A Farewell to Alm’s’.

    1. I took that to mean that cultural and environmental factors might be affecting the expression of a given trait, rather than genetic factors.

  10. red-breasted nuthatch

    Kewt! I had to run down the etymology of the swedish word for our ‘sitta’ (Sitta europaea); “nötväcka”.

    Obviously “nöt” [en. “nut”] correspond, but “väcka” is from old swedish “vækia” := “hugga (vak)”, where “vak” := ice hole. Apparently its ability to hack out previously stored nuts (presumably) reminds of ice fishing.

    Which is another kind of cute I guess.

Leave a Reply