Monday, December 4, 2017

Check your (taxonomic) biases at the door

Many of us like to believe that we are conceptually-oriented researchers; our particular study organism(s) are just means to an end, the end goal being to answer broad questions about how evolution or ecology works. But our citation patterns suggest otherwise. We are more likely to read, and hence more likely to cite, papers about organisms of direct relevance to our own work. That results in some unfortunate conceptual balkanization. Case in point: I’m writing an Annual Reviews paper on (non)parallel evolution, with several co-authors. Collectively, we have all studied sticklebacks (though we have some other study organisms in our history). So we tried very hard to diversify our citations (taxonomically; I must admit I haven’t gone through and checked the gender of first authors, for instance).
            In one section of the manuscript we discuss Langerhans and Dewitt’s 2004 AmericanNaturalist paper on Shared and Unique Features of EvolutionaryDiversification.  In that paper’s appendix they lay out a multivariate statistical approach to measure the extent to which evolution has occurred in parallel across multiple replicates, or in a unique direction in each replicate. That’s a very useful approach, that is very conceptually general and widely applicable (hence its great fit to The American Naturalist). So, I wanted to find a plant paper that cited this article, to use as an example illustrating their analytical approach. The problem is, I found only three plant papers out of the 200 that cited Langerhans and Dewitt (2004) according to Web of Science (as of December 1 2017). That’s 1.5%.  On the other hand, I noticed a suspicious number of fish citations. So I went back and noted down the taxonomic focus of each of the 200 papers citing their article (a few were reviews that had no particular taxonomic focus, so I didn’t tally those). Here’s the breakdown (sorted by # of citations):

Taxonomic group
# citations to L&D 2004


















The really striking thing here is how taxonomically biased this is. The single genus Gasterosteus has five times as many studies citing L&D than all studies of insects, which collectively are of course at least as diverse as stickleback, and possibly even as important from a practical standpoint.
            Why is this widely-applied method being effectively ignored by the vast majority of researchers? I think the answer is pretty simple: because the original paper was applying their method to study fish evolution, fish evolutionary biologists were more likely to read it. Worryingly, this suggests that when we publish a new method, we might reach the largest audience if we omit application of the method to any one taxon (or, if we apply it to many diverse taxa). Equally worringly, it suggests to me that because of our inherent taxonomic biases we are missing the boat on many important ideas and methods.
            The other hypothesis of course is that parallel evolution may be an exceptionally hot topic among fish biologists, especially stickleback- and Poeciliid-researchers. Bandwagons happen. We tried really hard in our Annual Review manuscript to branch out and cite studies involving something other than stickleback. That’s hard both because this is the system we are collectively most familiar with, and because this is such a great parallel-evolution system that many of the leading studies on the topic use this organism.

            There’s not an easy solution to this, other than to check our biases at the door: when writing, and when reading, ask yourself how diverse your citations (or paper choices) are. Try to step out of your comfort zone, read something about an organism you know nothing about, at least once a week if not more often. You’ll learn a lot of biology in the process, and maybe get some new ideas that help you step out of your taxonomic box.


  1. If you're looking for plant studies involving parallel evolution, my impression is that edaphic evolution to serpentine soils has produced some good stuff; my friend Nishanta Rajakaruna could probably help you out.

    "which collectively are of course at least as diverse as stickleback, and possibly even as important from a practical standpoint" :->

  2. That is an excellent example. I was especially interested, however, in cases where people applied Langerhans & Dewitt's analysis.

  3. I've come from the plant world to the stickleback world very recently. The one major thing that I've noticed separating stickleback (and perhaps fish, generally) folks from other taxonomic folks - in the way we think - comes from the types of habitats our organisms live in. Namely, (more-or-less) discrete lakes/stream populations, vs. (more-or-less) continuous terrestrial populations. It's hard to identify independent cases of parallel evolution in plants, whereas each lake is an independent instance of parallel evolution in fish.

    There are few analogs to the 'lake/stream' in other organisms. Fish nestle into studies of parallel evolution perfectly, because they are (1) confined to a discrete habitat, (2) do not typically disperse eggs or adults between such habitats (like Daphnia ephippia, for example), and (3) unintentional human-mediated dispersal is relatively difficult.

    To this end, it might even be the case that 'allopatric parallel evolution' is actually more frequent in fish than any other type of organism. So, I might add another possible answer to your question of why it's only being cited by fish researchers: maybe fish papers are the only papers that can truly test these hypotheses in a robust way. Or, at least, they are the most amenable to tests.

    Ben has a good example of plants adapting to discrete habitat features. This happens but I would guess it's less 'independent' than in fish. Timema stick insects - and host plants generally - are another example. And such cases of parallelism seem less common than in fish.

    I should add that I am looking forward to reading your Annual Review. Especially curious to read about speciation between populations that have evolved in parallel.

  4. It could be that the phenomenon is more easily assessed, and more frequent, in fish compared to many other taxa, but parallel evolution and this method does not require discrete categories of habitats. Although it is easier when habitats are discrete.

    But a couple problems with this notion is that 1) there’s a lot of work on parallel evolution in organisms other than fish, and that 2) another paper that used this method a couple of years after the original Am Nat paper has been cited quite a bit by studies on herps (reptiles and amphibians). That is, Langerhans et al. 2006 in Evolution (Shared and unique features of diversification in Greater Antillean Anolis ecomorphs) used the L&D 2004 method, but on lizards and not fish, and subsequently has a different taxonomic distribution of citations:

    Herps: 37%
    Fish: 31%
    Isopods: 11%
    Birds: 3%
    Spiders: 3%
    Plants: 3%
    Mammals: 1.5%
    Crickets: 1.5%
    Virus: 1.5%
    No taxa: 8%

    So, maybe the phenomenon or its investigation is more common in herps and fish? Or maybe citations have a serious taxonomic bias...


Sticklestock center

"There are two kinds of readers.  Those who have read the Lord of the Rings.  And those who are going to." There are two kinds of ...