Saturday, September 27, 2008

What the Hell is Niche Conservatism?

Niche conservatism is all the rage; virtually unheard of at the turn of the Millenium, 2008 has already seen nearly 30 papers on the subject. The only problem? Nobody seems to agree on what it is!

In a recent exchange in Ecological Letters, Jonathan Losos and John Wiens try to cut through the confusion. Let's begin with Losos' definition of niche conservatism as a pattern that "results when closely related species are more ecologically similar that [sic] would be expected based on their phylogenetic relationships" (emphasis mine). In advocating this definition, Losos argues strongly against the view that niche conservatism is a synonym for phylogenetic signal (i.e., the tendency for related species to be more similar than species drawn at random from a phylogeny). To make this equation, he argues, would defy the original meaning of the term niche conservatism and expands its scope so broadly as to render it meaningless. As Losos notes, some degree of ecological similarity is expected among closely related species even under the simplest models of evolution (e.g., Brownian motion resulting from shifting selective pressures or drift). Niche conservatism, meanwhile, results when species are more similar than expected under this simple model due to constraint by natural (stabilizing) selection, gene flow, pleiotropy or the absence of variation. Although Wiens agrees on this point, many recent studies claiming to support niche conservation involve nothing more than simple tests of phylogenetic signal. In Losos' view, these studies provide evidence that is necessary, but not sufficient, for the identification of niche conservation.

Although I've previously confounded niche conservatism and phylogenetic signal in my own work, I've come around to the view shared by Wiens and Losos. Another influential voice in the field -- Michael Donoghue (2008) -- seems to straddle the fence when he says "strictly speaking, it is not necessary to link PNC with the view that there are constraints on niche evolution, [but] I believe that it is the relative difficulty of making major ecological shifts that explains [patterns of niche conservation]." What do you all think?

If people are into this topic, perhaps we can try to tackle Losos and Wiens arguments about how and when niche conservatism should be studied, and whether it represents a pattern, a process, or both...

Literature Cited:
Losos, J. B. 2008. Phylogenetic niche conservatism, phylogenetic signal and the relationship between phylogenetic relatedness and ecological similarity among species. Ecology Letters 11:995-1003.

Wiens, J. J. 2008. Commentary on Losos (2008): Niche Conservatism Déjá Vu. Ecology Letters 11:1004-1005.

Losos, J. B. 2008. Rejoinder to Wiens (2008): Phylogenetic niche conservatism, its occurrence and importance. Ecology Letters 11:1005-1007.

Donoghue, M. J. 2008. A phylogenetic perspective on the distribution of plant diversity. PNAS 105:11549-11555.

8 comments:

Dan Warren said...

I've read the first Losos paper, but have yet to read either side of his exchange with Wiens. Is there a concrete suggestion for how we can tell the difference between a slow Brownian process and one that is potentially faster than what we observe but is constrained by some external force? In other words, how can we tell whether the niche is "conserved", or whether it just evolves more slowly in some groups than others? I agree with Losos' idea in principle, but it's just not clear to me yet how it can be applied in empirical studies. Does it rely on the hypothesis that niche conservatism, if it exists, results in a pattern of niche evolution that is measurably non-Brownian?

Glor said...

Yes, he does suggest that the hypothesis of niche conservatisms requires a pattern that is measurably non-Brownian. I'm with you when you suggest that this is problematic. Not to dodge the question, but I'm actually coming around to Wiens' point of view that we shouldn't worry so much about identifying niche conservatism per se. Losos suggests a bunch of studies that require an a priori assay of niche conservatism, but it seems like you could get by with simpler tests in these cases. For example, predictions of ancestral niches and other types of comparative studies seem to require only tests of phylogenetic signal and its nature, not comprehensive tests of niche conservatism.

Dan Warren said...

I agree completely - just identifying niche conservatism isn't a particularly productive way to study the evolution of the niche, and we should instead be focusing our efforts on ways of studying patterns and rates of niche evolution. Jeez, we should write a paper about that or something.

Glor said...

Haha. I'm not sure how many such papers will be written before this starts to sink in...

liam.revell said...

Dan, regarding your first comment: "how we can tell the difference between a slow Brownian process and one that is potentially faster than what we observe but is constrained by some external force?" In our recent paper, Luke, Dave Collar, and I show that there is no relationship between the evolutionary rate & phylogenetic signal under pure drift (i.e., BM). However, "constraint," depending on its form, can tend to depress phylogenetic signal most when the evolutionary rate is high. I realize that signal and PNC are not synonyms (in fact, we recommend explicitly against synonymizing the two terms, as has been suggested by some other authors). I think our paper contains several other points that are also somewhat relevant to Losos (2008).

Dan Warren said...

Thanks, Liam! It sounds like an interesting paper, and something that I definitely need to read.

As for getting away from this "niche conservatism yes/no" approach that's unfortunately dominating the literature right now, my feeling is that the questions will get better when the methods get better. I think most people would acknowledge that there are much more interesting questions to ask about the evolution of the niche, but the methods needed to address those questions quantitatively just haven't been worked out yet.

Luke J. Harmon said...

I agree with Liam (unsurprisingly!) that our paper is very closely related to this. There is a bit of momentum towards fitting more specific models of character evolution to phylogenetic data, and it seems like this is the future for these questions.

liam.revell said...

From a practical perspective, the paper can also provide some guidance as to whether phylogenetic signal is even expected to be high when the niche is conserved. A simple example: if the niche is perfectly preserved on a given phylogeny (i.e., all lineages are under strong stabilizing selection to the same optimum), then the expected phylogenetic signal is very low. This is pretty unsurprising - however some of our findings regarding the relationship between evolutionary process and phylogenetic signal are more interesting. We also recommend that rather than attributing phylogenetic signal to this process or that - we should consider fitting an explicit model for the process we hypothesize, and then evaluating the fit of that model in comparison to others. This might be a more fruitful approach than a potentially endless discussion of what PNC really means.