Glenn Branch's picture

Falsifia-behe-lity Revisited

Physical model of a bacterial flagellum

It’s never certain what the response to a blog post will look like, of course; I understand that. When I wrote “Falsifia-behe-lity,” I didn’t anticipate that commenters, especially those on NCSE’s Facebook page, would be particularly interested in making NSFW conjectures about the illustration that accompanied the post. (For the record, it shows “a physical model of a bacterial flagellum. It was imaged and modeled at Brandeis University in the DeRosier lab and printed at the University of Wisconsin, Madison. It was fabricated on a ZCorp Z406 printer from a VRML generated at Brandeis.”) But I did anticipate that commenters might think that I was trying to discuss the issue of falsifiability in general, and although I took pains to try to make it clear that I was addressing just Behe’s arguments in “The modern intelligent design hypothesis” (2001), I thought that it was possible that I would be pressed to offer a general disquisition on falsifiability.

I wasn’t, which is just as well; I estimate that it would take at least 25,000 words to do it right, and who wants to read a 25,000-word blog post? But I was asked a question that is, perhaps, equally interesting and is, certainly, easier to answer: how does my criticism of Behe’s attempt to turn the tables—that is, to argue, with regard to the bacterial flagellum, that “intelligent design” is and “Darwinism” is not falsifiable—comport with Behe’s claim, also to be found in “The modern intelligent design hypothesis,” that his critics are trying to have their cake and eat it too? Behe writes, “Now, one can’t have it both ways. One can’t say both that ID is unfalsifiable (or untestable) and that there is evidence against it. Either it is unfalsifiable and floats serenely beyond experimental reproach, or it can be criticized on the basis of our observations and is therefore testable.” As it happens, the previous discussion helps to show why Behe is wrong here too.

First, let me summarize the main points of “Falsifia-behe-lity.” Behe assumes (and I conceded for the sake of argument) that a theory is scientific only if it is (or entails hypotheses that are) falsifiable. He defines “Darwinism” (with regard to the bacterial flagellum: I’ll usually omit the qualification hereafter) so as to be unfalsifiable and “intelligent design” so as to be falsifiable. But the unfalsifiability of Behe’s “Darwinism” is unimportant, since there are instances of it that are falsifiable. And the falsifiability of his “intelligent design” is not genuine, because in order for it to contradict “Darwinism” it needs to be redefined in such a way that renders it unfalsifiable after all: as (I suggested) “some intelligent process did in fact produce the bacterial flagellum.” So redefined, like “Darwinism” its unfalsifiability is unimportant, but unlike “Darwinism” it’s unclear that there are any falsifiable instances of it that have any plausibility.

Now, why is Behe inclined to accept “intelligent design” with respect to the bacterial flagellum? Ignoring the details, the basic argument, to be found in Darwin’s Black Box (1996) and elsewhere, is:

The bacterial flagellum is irreducibly complex.

It is impossible for irreducibly complex structures to be produced by unintelligent processes.

Therefore the bacterial flagellum was produced by intelligent processes.

(Behe sometimes hedges, saying not that it’s impossible, but that it’s extremely unlikely for irreducibly complex structures to be produced by unintelligent processes. This doesn’t affect any point that I’m going to make here.)

There’s a lot that’s wrong with the argument, of course, and to his credit Behe discusses a few of his critics. Russell Doolittle argues that the vertebrate blood-clotting cascade (rather than the bacterial flagellum) is not irreducibly complex, citing knock-out experiments in mice, while Kenneth R. Miller argues that it is, contrary to Behe, possible for unintelligent processes to produce irreducibly complex structures, citing work by Barry Hall on the experimental evolution of a lactose-utilizing system in E. coli. Of course, Behe argues that Doolittle and Miller are mistaken about the science, and there have been replies by or on behalf of Miller and Doolittle, and so on; you can decide for yourself who wins the scientific battle. (Hint: not Behe.) But the point of interest here is not the scientific details, but Behe’s failure to distinguish between criticizing the premises of his argument and criticizing its conclusion.

At least to a first approximation, it’s plausible to regard the premises of Behe’s argument as falsifiable: they make risky claims about the world, and thus they can be shown to be false if (for the first premise) the bacterial flagellum can be shown to perform its function even after the loss of one of its parts or (for the second premise) unintelligent processes can be shown to produce irreducibly complex structures. (I say “to a first approximation” because there are unresolved questions about the exact definition of “irreducible complexity” and about the meaning of “unintelligent processes” that Behe fails to address; but I won’t pursue any line of criticism based on those questions here.) But—as argued in “Falsifia-behe-lity”—“intelligent design” with regard to the bacterial flagellum, as Behe ought to define it, is unfalsifiable by his lights. So it is not at all inconsistent to criticize his premises as empirically false and his conclusion as unfalsifiable.

It is perhaps helpful to stress that there’s nothing logically anomalous in the argument’s having premises that are falsifiable but a conclusion that is unfalsifiable. Here’s a homespun example that I used in “Falsifia-behe-lity”:

Janet Parker died from smallpox in Catherine-de-Barnes, England, in 1978.

Therefore someone died from smallpox somewhere sometime.

The argument is clearly valid (unlike Behe’s), and the premise is falsifiable, but the conclusion, by Behe’s lights, is unfalsifiable: for there are a potentially infinite number of deaths (including deaths in the future and elsewhere in the universe) to investigate.

Just as Behe’s attempt to turn the tables fails, so his attempt to charge his opponents with having their cake and eating it too fails. It will perhaps not be surprising that his vaunting rhetoric about falsifiability fails as well. Behe writes, “The danger of accepting an effectively unfalsifiable hypothesis is that science has no way to determine if the belief corresponds to reality.” That’s not quite right, of course. If you accept, e.g., Pallen and Matzke’s falsifiable hypothesis about the evolution of the bacterial flagellum, then (given that it appeals only to unintelligent processes) you accept “Darwinism” in Behe’s sense, but accepting the latter unfalsifiable claim poses no danger. On the other hand, when the proponents of “intelligent design” insist on a definition of “intelligent design” on which it is unfalsifiable, and never present any plausible instances of it that are falsifiable, then there is a serious danger—at least to their scientific credibility.