Arguments from personal incredulity versus inductive arguments from sampling
April 16, 2013
All I want to do in this post is make people aware that there is a difference between these two things, and occasionally that affects those of us who work in natural history.
In one of his books or essays, Stephen Jay Gould made the point that in natural history we are usually not dealing with whether phenomena are possible or not, but rather trying to determine their frequency. If we find that in a particular population of quail most of the birds eat ants but some avoid them, then we know some things: that quail can tolerate eating ants, that quail are not required to eat ants, and that both strategies can persist in a single population.
This idea has obvious repercussions for paleoart, especially when it comes to “long-tail” behaviors. I dealt with that in this post, and also in the comment thread to this one. But that’s not what I want to talk about today.
Sometimes it is useful to talk about things that never happen, or that have at least never occurred in the sample of things we know of. Obviously how certain you can be in these cases depends on the intensity of sampling and the inherent likelihood of a surprising result, which can be hard to judge. If you argued right now that T. rex lacked feathers because no T. rex specimens have been found with feathers, you’d most likely be wrong; it is almost certainly just a matter of time before someone finds direct evidence of feathers in T. rex, given the number of T. rex specimens waiting to be found and the strength of the indirect evidence (e.g., phylogenetic inference, analogy: ornithomimids are known to be feathered even though most specimens are found without feather impressions). If you argue that sauropods are unique among terrestrial animals in having necks more than five meters long, you’re most likely right; being wrong would imply the existence of some as-yet undiscovered land animal of sauropod size, or with seriously wacky proportions (or both), and our sampling of terrestrial vertebrates is good enough to make that extremely unlikely.
The reason for this post is that sometimes people confuse that last argument, which is about sampling and induction, with the argument from personal incredulity.
For example, in our no-necks-for-sex paper (Taylor et al. 2011), we included this passage:
Sauropoda also had a long evolutionary history, originating about 210 million years ago in the Carnian or Norian Age of the Late Triassic, and persisting until the end-Cretaceous extinction of all non-avian dinosaurs about 65 millions years ago. Thus the ‘necks-for-sex’ hypothesis requires that this clade continued to sexually select for exaggeration of the same organ for nearly 150 million years, a scenario without precedent in tetrapod evolutionary history.
One of the reviewers argued that we couldn’t include that section, because it was just the argument from personal incredulity writ large, like so:
There are no other known cases of X in tetrapod evolutionary history, and therefore we don’t believe that the case in question is the sole exception.
…with the second part of that unstated (by us) but implied. But we disagreed, and argued (successfully) that it was an argument based on sampling, like so:
There are no other known cases of X in tetrapod evolutionary history, and therefore it is unlikely that the case in question is the sole exception.
Now, it is perfectly fair to criticize arguments like that based on the thoroughness of the sampling and the likelihood of exceptions, as discussed above for T. rex feathers. Just don’t mistake arguments like that for arguments from personal incredulity.* On the flip side, if someone makes an argument from personal incredulity, see if the same thing can be restated as an argument about sampling. Maybe they’re correct but just expressing themselves poorly (“I refuse to believe that the moon is made out of cheese”), and maybe they’re wrong and restating things in terms of sampling will help you understand why.
* If you want to get super pedantic about it, they’re both arguments from ignorance. But one of them is at least potentially justifiable by reference to sampling. Absence of evidence is not necessarily evidence of absence, but it may get to be that way as the sampling improves (e.g., there is no evidence of planets closer to the sun than Mercury, and at this point, that is pretty persuasive evidence that no such planets exist).
Parting shot: one thing that has always stuck in my head from Simberloff (1983) is the bit about imagining a large enough universe of possible outcomes. And I’ve always had a perverse fascination with Larry Niven’s “Down in Flames”, in which he pretty much demolished his Known Space universe by assuming that every basic postulate of that universe was false. Neither of these follow directly on from the main point of the post, but they’re not completely unrelated, either. Because I think that they yield a pretty good heuristic for how to do science: imagine what it would take for you to be wrong–imagine a universe in which you are wrong–and then go see if the thing that makes you wrong, whatever it is, can be shown to exist or to work. If not, it doesn’t mean you’re right, but it means you’re maybe less wrong, which, if we get right down to it, is the best that we can hope for.
The photos have nothing to do with the post, they’re just pretty pictures from the LACM to liven things up a little.
- Simberloff, D. (1983). Competition theory, hypothesis-testing, and other community ecological buzzwords. The American Naturalist, 122(5), 626-635.
- Taylor, M. P., Hone, D. W., Wedel, M. J., & Naish, D. (2011). The long necks of sauropods did not evolve primarily through sexual selection. Journal of Zoology, 285(2), 150-161.