Lost birds and natural selection

by Matthew Cobb

Over at The Atlantic, Rebecca Rosen has a nice piece about how nearly 900 homing pigeons got lost after being released from a site in Jersey Hill in upstate New York between 1968 and 1987. The disappearance of these pigeons were noted because they were part of a long-term experiment by Cornell professor William T. Keeton, into pigeon navigation. The explanation, as outlined by Rosen, has recently been put forward by Jonathan Hagstrum from the US Geological Survey, in a recent article in Journal of Experimental Biology. And the answer is ‘infrasound’. Hasgtrum’s abstract says:

Jersey Hill lies within an acoustic ‘shadow’ zone relative to infrasonic signals originating from the Cornell loft’s vicinity. Such signals could arise from ground-to-air coupling of near-continuous microseisms, or from scattering of direct microbaroms off terrain features, both of which are initially generated by wave–wave interactions in the deep ocean.

Hagstrum modelled what might have been happening to the sound produced by these various sources and concluded:

little or no infrasound from the loft area arrived at Jersey Hill on days when Cornell pigeons were disoriented there, and that homeward infrasonic signals could have arrived at all three sites from directions consistent with pigeon departure bearings, especially on days when these bearings were unusual. The general stability of release-site biases might be due to influences of terrain on transmission of the homeward signals under prevailing weather patterns, whereas short-term changes in biases might be caused by rapid shifts in atmospheric conditions.

The pigeon-lover Darwin could never have imagined it!

Interestingly, the catastrophic disappearance of a flock of birds played an important role in providing empirical evidence that natural selection could shape species. In 1898, American naturalist Hermon C Bumpus published an article in which he pointed out that, at the time, although many thinkers used natural selection as the framework to interpret the world,

‘we forget we are really using a hypothesis that still remains unproved and that specific examples of destruction of animals of known physical disability are very infrequent’

In February 1898 there was a terribly violent snowstorm and at the end of the storm, a group of 136 sparrows that had been hit by the storm were brought into Bumpus’ lab at Brown University. 72 of the birds survived, 64 perished. What separated the living from the dead? Bumpus measured each bird according to 10 criteria (size, weight, beak length etc). What he found was that the birds that died were generally those that departed most from the mean – the largest, the smallest, the lightest, the heaviest, etc. However, he also noticed that some birds with ‘extreme’ characters did survive, as long as they did not have several extreme variants.

Bumpus concluded with a view that natural selection tended to favour a ‘type’, rather than extreme variants:

‘Natural selection is most destructive of those birds which have departed from the ideal type, and its activity raises the general standard of excellence by favoring those birds which approach the structural ideal’

(This is the only sentence in the article that is in italics)

This raised a problem, which Bumpus did not address, and I invite readers to comment on below the fold. If natural selection favours a ‘type’ and is therefore clearly conservative, as it was in the case of these sparrows, how can selection explain ‘progressive’ evolution, and the appearance of new types? Remember, this was a real and perplexing argument a little over a century ago, decades after Darwin’s approach had become widely accepted.


If you want to read Bumpus’ paper and explore his data, Clark University and UWC have put the material on line as a way of getting their students to study exactly what happened on that snowy night in 1898, and what it means for our understanding of natural selection.

If you want to know more about the way that 19th and early 20th century thinkers tried to find evidence for natural selection, I strongly recommend Jean Gayon’s book Darwinism’s Struggle for Survival: Heredity and the Hypothesis of Natural Selection (Cambridge). Declaration of interest: I translated it.

h/t Ben Goren for the pigeons.


  1. gbjames
    Posted February 7, 2013 at 6:54 am | Permalink

    “how can selection explain ‘progressive’ evolution, and the appearance of new types”

    Environments change. And so selective pressures on populations change.

    • jesse
      Posted February 7, 2013 at 7:35 am | Permalink

      Excellently concise.

    • Gregory Kusnick
      Posted February 7, 2013 at 11:52 am | Permalink

      The “ideal type” is a moving target.

    • Diane G.
      Posted February 7, 2013 at 1:17 pm | Permalink

      Selective pressures may vary temporally, geographically, and across many other parameters we think of a defining a species’s niche space.

      • Another Matt
        Posted February 7, 2013 at 3:58 pm | Permalink

        Diane G., I was gonna reply with almost exactly this, but I was at the dentist. Thanks for putting it so succinctly — mine might have been two paragraphs. 🙂

        • Diane G.
          Posted February 8, 2013 at 2:56 am | Permalink

          Gee, thanks!

          And now that your comment caused me to reread mine, I hope it’s obvious to everyone that my first “a” is supposed to be an “as,” sigh.

  2. Rebecca Harbison
    Posted February 7, 2013 at 7:01 am | Permalink

    Well, my first thought is ‘something changes’. If a sparrow is well adapted for its environment, the mean sparrow is at a local maximum for relative fitness. The more it departs from that, the less fit it is, and the less likely it is to survive extreme events. (If it’s not getting as much food, it could more easily freeze, for instance.)

    If the environment changes — because some of the birds move to a new environment, or the climate changes, or humans do new things — then the maximum changes and suddenly some of the extreme sparrows might have an easier time of it, and the ‘mean sparrow’ shifts in that direction as the generations pass and the extreme sparrows raise more sparrow chicks.

    • Dominic
      Posted February 7, 2013 at 7:17 am | Permalink

      Yes – sparrows are finches & subject to the same trends/pressures as Galapagos finces but over a much larger range.

  3. Dominic
    Posted February 7, 2013 at 7:15 am | Permalink

    It is an interesting article! I tweeted it last week (our library covers sound & associated topics). Infrasound is fascinating -I was just wondering whether dinosaurs might have used it as elephants & whales do (reading Neil Shubin & his new book The Universe Within, a chapter talking about the importance of size for certain things to happen with animals, ie you have to be larger to hear lower frequencies). (Some people have heard an infamous ‘hum’ which has been suggested to be low frequency sound.)

    As to Matthew’s poser-
    If a population is large, won’t that tend to iron out differences if the population mixes easily?
    However as range extends does it takes longer for a trait to spread from a fringe area to the centre, while easier for a trait in the centre to expand because of more potential partners?
    Also size of animal & speed of reproduction – try measuring horn difference in 170 rhino – what differences are due to nutrition rather than genes? (thinking of David Raup, Bad Genes or Bad luck)…: genetic drift?

    By the way, as Matthew is too modest to mention it, take a look at this

  4. aboc zed
    Posted February 7, 2013 at 8:06 am | Permalink

    “how can selection explain ‘progressive’ evolution, and the appearance of new types”

    my opinion is this:

    LIFE as self-replicating chemical system went from most ‘simple’ lifeforms (single cell) to most ‘complex’ lifeforms (genus homo on top of food chain) simply because ALL lifeforms use and re-use the same configuration-space and the system of feedbacks within that configuration space would ‘destined’ LIFE to go the way of what we call ‘progressive’ evolution.

    That process of complexification (or evolutionary process) has lead to lifeform that is ‘deliberative capability’ making all of these observations possible.

    In a certain sense this process of complexication is complete with genus Homo being on top of the food chain ignorantly playing with fire of genetic modification without allowing itself time to actually understand all the relationships of the black box that is LIFEPLANETLIFE

    The evolutionary process therefore now is the evolution of DELIBERATIVE CAPABILITY.

    That is the genus Homo will not end with Homo Sapiens Sapiens simply because the momentum of evolution out of ignorance will eventually cap population growth and climate change, food crysis, pollution, damaged genetic laboratory will present the final environmental pressures on Homo Sapiens Sapiens producing a sub-species that I call Homo Cogitans.

    But we are at least several generations and a collapse of democracy and capitalism away from that.

    So no need to worry or do anything differently: we can continue solve our local optimization problem by becoming most successful peckers in our environment: learn how to do business, how to talk bs and how to become the biggest man/women in our little tribe of ignorance.

    Natural selection will do the work and each of us can enjoy our life while we can because at some point it will not be as pleasant and long as it is now! 🙂

    • Torbjörn Larsson, OM
      Posted February 7, 2013 at 8:42 am | Permalink

      I’ll take your teleology and respond with parasite simplification. IIRC over half of today’s species are parasites, so evolution is at least as much “simplification”.

      [The real response would be along Gould’s “it started out simple” argument, I believe, but perhaps it is better to leave that to the biologists.]

      And language capable intelligence is a rare trait, not an end run on a non-teleological process.

      Finally, a nitpick: the proposed definition for life fails nearly all species. As far as I know only a prokaryote from a South African mine has been shown to be entirely self-sufficient, just subsisting on energy from hydrogen liberated by radioactive minerals and the carbon it can scavenge.

      It could become a good definition of biosphere though, after removing “self-replication”: a robust chemical system.

      • aboc zed
        Posted February 7, 2013 at 8:56 am | Permalink

        it is all in the definitions and the ‘rubber of language’

        we can orgue about what ‘simplification’ and ‘complexification’ mean in this context but it only would make sense if it were of practical importance for either you or me

        as it stands each of us always goes back to our own ‘local optimization problem’ : some go to slave the companies they work for, some go to be slavedrivers or high priests of the nations

        as a species we are incredibly ignorant and primitive and the way we do not respect science and relegate to be a mere ‘enabler of ignorant bumbling and stumbling into the future’ for some is sad for some is natural for some does not exist

        thanks for your clarifications – i will look into readings that fall into them

    • Michael Fisher
      Posted February 7, 2013 at 9:48 am | Permalink

      Are you Alex Todorov fromHERE ? Someone going by “AIT” & also “Alex” has posted identical stuff here at WEIT around half a dozen times over the past two years

      I remember because of your phrase Homo cogitans

      • aboc zed
        Posted February 7, 2013 at 10:52 am | Permalink

        the site you linked to is not mine

        my views are entirely my own although I admit I borrowed the phrase “Homo cogitans” from the material on the site

  5. Marcoli
    Posted February 7, 2013 at 9:40 am | Permalink

    Dang it, I accidentally posted in the wrong spot. Here it is again where it belongs.
    This all seems pretty straightforward, at least in that some standard textbook answers will do the job here.
    1) Natural selection can only really select ‘against’ a trait, it does not directly select ‘for’ particular traits.
    2) Stabilizing selection was applied to the birds with extreme traits. Different traits were being selected against (wings too long / wings too short, etc.) Stabilizing selection more generally applies when an environment is stable, and the population has been optimized to that environment.
    3) If conditions change, or if a new environment opens up that becomes populated, NS will select against varieties that are less fit in that new environment — even if they were once the more fit. I am reminded of selection against finches with smaller beaks during droughts on the Galapagos — those finches are reasonably fit during times without drought. The result is directional selection. The average of the population shifts toward the more fit b/c of selection against the less fit.
    4) How does any of this lead to anything NEW? The standard answer, which seems adequate to me, is that sometimes directional selection is applied toward REPURPOSING structures for a new use. This is called EXAPTATION.

  6. W.Benson
    Posted February 7, 2013 at 10:37 am | Permalink

    Selection against the extremes and directional selection are not mutually exclusive. Showing that elephants smaller than 10 pounds and larger than 50 tons are at a disadvantage relative to the 2 ton ‘type’ is irrelevant to the question of directional selection, for example, to diminish mean size from, say, 2.2 to 2.1 tons.

    Bumpus believed he had shown selection against extreme variants. His data also suggest that individuals with longer bone measurements may have tended to survive better after being disabled by the ice storm. The differences are small, on the order of 1%-2%, and are not significant. However, they are consistent with Bergmann’s rule. It would be hard to detect a survival difference of only a few percent with the small sample size available to Bumpus.

    If I am not mistaken, after house sparrows were introduced into North America from Europe and have expanded over much of the continent where they have evolved according to Bergman’s rule, i.e., sparrows today are bigger where temperatures are cooler and smaller where it is hotter, putatively because of the demands of thermoregulation.

    “How can selection explain ‘progressive’ evolution”? Answer: by considering appropriate situations, situations in which ecology demands ‘progressive’ adaptations for which selectable genetic variation exists, e.g., ‘progressive’ insecticide resistance in crop pests.

    For data on Bergmann’s rule in house sparrows, see Johnson & Selander (1973), Baker (1980), Murphy (1985), McGillivray & Johnston (1987) and Yom-Tov (2001), cited in
    S. Meiri and T. Dayan. 2003. On the validity of Bergmann’s rule. J. Biogeography, 30:331-351 (pdf available for downloading).

    • W.Benson
      Posted February 7, 2013 at 4:27 pm | Permalink

      Heck, I read ‘progressive’ to mean ‘directional’ evolution. The Bumpus study has nothing to do with the ‘type’ unless he means, as I think he does, average phenotype of the species. The study certainly gives no insight into the evolution of increased complexity (= progress?).

      As said above, showing that there is selection against extreme phenotypes is not the same thing as refuting directional selection. The criterion for stabilizing selection (for me, at least) is the maximization of fitness of the phenotype corresponding to the real population mean. Bumpus did not demonstrate this.

      Matthew says: “If natural selection favours a ‘type’ and is therefore clearly conservative, as it was in the case of these sparrows . . .” This last assertion is questioned above. Moreover, for natural selection to be “clearly conservative”, it would have to be stabilizing in (almost) all cases investigated, not just (unconvincingly) for a small sample of house sparrows disabled on a cold winter’s night in 1898.

      That said, phenotypes are no doubt in dynamic equilibrium with selection, and selection may fluctuate in direction as ecological conditions ‘cycle’. Of course useful genetic innovations and qualitatively new environmental challenges will be game changers.

  7. Jim Thomerson
    Posted February 7, 2013 at 11:20 am | Permalink

    There was a study by the late Clark Hubbs, maybe as far back as the 1950s. He studied a population of darter fish in a small spring fed stream. During the breeding season, the temperatures at the spring head were uniform throughout the day. A ways down stream, the temperature varied over 24 hours.

    Hubbs stripped eggs and sperm from the darters to produce fertilized eggs. He raised fry under conditions of constant temperature and changing temperature. Fry from the spring head population survived significantly better at the constant temperature, while fry from the downstream population survived significantly better under varied temperature.

    Hybrid fry survived better at mildly changing temperature than at natural downstream change, or at constant temperature. Hubbs took this as evidence that differences in these different subpopulations were genetically based. So here is directional selection within a population along an environmental gradient

    • John Scanlon, FCD
      Posted February 9, 2013 at 1:24 am | Permalink

      That’s very cool, and the elegance of the demonstration requires the population to be nearly as one-dimensional as the gradient (imagine individual fish lined up -M-F-M-F- so they can only mate with immediate neighbours). Most environmental variables can’t be aligned with a single spatial dimension, let alone a one-dimensional population, but… what Diane G said.

  8. cherrybombsim
    Posted February 7, 2013 at 11:53 am | Permalink

    “If natural selection favours a ‘type’ and is therefore clearly conservative, as it was in the case of these sparrows, how can selection explain ‘progressive’ evolution, and the appearance of new types?”

    This is the $64 question, right?

    You could get really, really deep into this, but I’ll just put in 2 cents worth. Selection against birds with wings too long or wings too short is selection against the phenotype, which is not the same thing as selection against the genotype. If there were “a” gene for wing length, or some group of genes that had an additive epistatic effect, you could make a strong link. But we know there is no such single gene, and that epistasis is not necessarily linear so the effect of pruning the phenotypic deviants on the genome itself is problematic.

    You can have selection against extreme phenotypes at the very same time that rare genotypes are actually increasing in the population.

    Just a thought.

  9. Posted February 7, 2013 at 8:51 pm | Permalink

    Forgive my novice comment, and maybe someone would be kind enough to clarify things for me, but the conclusion that natural selection favors a type from evidence drawn on a single event seems erroneous.

    First, the storm was a unique meteorological event, which in this case favored birds of some narrow range of characteristics. Clearly, the birds that died were all part of a successful species. Unless these birds were to live in a storm ravaged environment, the storm(s) would pose no significant selective pressure, which was already proven by the presence of the birds killed by the storm.

    Second, because evolution and environmental change generally moves at a snails pace (geologic & climatic) natural selection is by its nature a long term process. “Types” are the result of selective processes. There are types because natural selection dictated the type, and over time types change.
    To go from a single event to the conclusion Bumpus suggested seems like a huge leap.

    I think it would be most interesting to understand why “the birds that died were generally those that departed most from the mean.”

  10. Jerry Schwarz
    Posted February 8, 2013 at 1:05 am | Permalink

    I think most of these comments are missing the point. Selection can only operate on existing variation. A change in the environment will not cause a change in a phenotypic or genetic trait if there is no variation in that trait to operate on. So the question is how is that variation maintained in the face of the normalizing selection.

    • Gregory Kusnick
      Posted February 8, 2013 at 1:37 am | Permalink

      Normalizing selection needn’t have the effect of eliminating genetic variation. There could be many different genotypes that produce the “ideal” phenotype and would therefore be preserved by normalizing selection. Those same genes, in different combinations, might produce phenotypes that would be selected against, but there are still enough copies of those genes in successful phenotypes to maintain diversity.

      Consider a pair of dice as a model, with 7 as the “ideal” phenotype. Roll the dice 100 times; for each throw that comes up 7, pass those two digits into the next generation. But there are six different ways to make 7, using all the digits in equal proportion. So no particular digit is selected for or against; all digits pass through, and all 36 combinations remain possible in each subsequent roll, despite strong selection for the 7 phenotype.

    • Diane G.
      Posted February 8, 2013 at 3:05 am | Permalink

      Because environmental variation happens over a shorter interval than the time it takes for genetic variation to be lost via normalizing selection.

      That’s how it’s maintained. If you meant to ask how it arises in the first place–oh, periods of allopatry, say? Non-random mating?

  11. madscientist
    Posted February 8, 2013 at 1:42 am | Permalink

    I seriously doubt that the answer has to do with infrasound. Pigeons being the size that they are will be extremely poor receptors of infrasound – that’s one hell of a loud (infrasound) bang for a pigeon to hear even only a few km from home. In fact it should be possible to calculate the energy that must go into a single pressure wave to be detectable by a pigeon at Distance X – we only need to know what the pigeon’s distance from the coop is and the pigeon’s threshold of hearing for the given frequency. A (moderately large) volcanic eruption can be detected by instruments far more sensitive than any bird at a distance of a few hundred km, but those are damned huge pulses.

    • Gregory Kusnick
      Posted February 8, 2013 at 11:55 am | Permalink

      I wondered about that too. But on the other hand, if there’s a strong correlation between the pigeons’ flight paths and the acoustic contours of the landscape (as the abstract seems to indicate), then that correlation needs explaining somehow.

      Note also that the abstract doesn’t specify the mechanism by which pigeons detect these sounds. It may have nothing to do with hearing. Just guessing here, but it doesn’t seem wildly implausible that they could detect low-frequency pressure waves through sensory neurons in their feather follicles, allowing them to use their spread wings as a kind of acoustic interferometer for navigation purposes.

  12. Posted February 8, 2013 at 3:20 am | Permalink

    Peter and Rosemary Grant (as documented in the excellent “The Beak Of The Finch” by Jonathan Weiner) found that the differences between finches, such as say a large beak for crushing hard nuts were only actually useful during extreme conditions on the little island (Daphne) where they did much of their research. In normal circumstances when food was plentiful, the different finches behaved much the same. So the individual species of finch are not adapted to just a particular environment, but rather that environment under the most challenging conditions possible, which may happen extremely infrequently, such as a greatly extended period of drought. When food becomes scarce the birds need to use their particular adaptations to tackle the more difficult types of food (such as really hard nuts).

    Taking the above into account one can see that evolution isn’t going to favour reversion of a population to one particular type, but rather to a set of different types, each one of which is a way of exploiting the environment in it’s most extreme challenging condition. You can also see how speciation might occur in these circumstances when those finches without useful adaptations for adversity get winnowed out.

  13. Posted February 8, 2013 at 9:18 am | Permalink

    Seems to me that lots of people here are missing some key points.

    First, I find Bumpus’s appeal to Platonic idealism entirely unconvincing on purely aesthetic grounds. That ship sailed a long time ago and got eaten by sea monsters as it fell over the edge.

    All the sparrows showed was that there was a local evolutionary maximum for survival in that one particular environment. And, considering that nowhere near as many humans as sparrows died in that particular environment, we can also conclude, in an extremely narrow sense, that the sparrows made some decidedly unintelligent evolutionary choices in the 300 million years since our last common ancestor with them, if you’ll spare the anthropomorphization.

    It is entirely reasonable to suggest that there are all sorts of other local maxima for sparrows. For one, that environment in which they died en masse was decidedly atypical; it could well be that the outliers would have out-thriven the others in the wild.

    It’s even possible that the variations were themselves largely neutral outside of the stormy environment but that they would have enabled more beneficial or deleterious mutations in the future.

    But, mostly, evolution is a statistical thing. On average, for the most part, the average of the population will be fairly optimal. That average drifts and ideal optimization will be somewhere other than the median…but, in general, the closer an individual is to the median, the odds are said individual is to the optimum.



    • John Scanlon, FCD
      Posted February 9, 2013 at 1:29 am | Permalink

      That ship sailed a long time ago and got eaten by sea monsters as it fell over the edge.


  14. garardi
    Posted February 9, 2013 at 12:17 am | Permalink

    It seems to me that the type have been selected for but they are not perfect therefore the majority of deviations being deleterous and only a small number of deviations will show qualities that are advantageous therefore the mean type will be most selected and the majority of deviations will fail.

  15. Allautin@gmail.com
    Posted March 30, 2013 at 10:48 am | Permalink

    Common sense go at query!? Not provided by a reader who could readily follow Bumpus’s experimental details.

    “If natural selection favours a ‘type’ and is therefore clearly conservative, as it was in the case of these sparrows, how can selection explain ‘progressive’ evolution, and the appearance of new types”

    This does not in anyway preclude variation or spread (albeit constrained around a mean or morphotype). Indeed the data (and Bumbus’ experiment) “requires” the spread (genetic/morphologic). Requires it as a background supposition – transformed into data by the experiment. That is to say

    This the experiment provides the evidence – that a bedrock variability is required in Darwinism.

    The other requirement is persisting change in environment/selection pressure. It seems this is provided secondarily in the experiment, qualified by the facf that Bumpus’ experiment was from a thin time dimension – a snap shot.

%d bloggers like this: