Irreducible Complexity
Behe's argument that some systems cannot function without every part has been answered for each of his examples.
6 min readUpdated
Michael Behe introduced the term in Darwin's Black Box (1996). An irreducibly complex system is one "composed of several well-matched, interacting parts that contribute to the basic function, wherein the removal of any one of the parts causes the system to effectively cease functioning."
The argument is that such a system cannot arise by successive slight modifications, because the intermediate stages would be non-functional and therefore not preserved by selection. Behe's stated conclusion is that they were designed.
Behe's examples and their status
| System | Claim | Current position |
|---|---|---|
| Bacterial flagellum | A rotary motor whose ~40 proteins are all required | About 10 core proteins correspond to the type III secretion system, a functioning injection apparatus. Component subsets have other functions |
| Blood clotting cascade | Removing any factor breaks the cascade | Whales and dolphins lack Factor XII and clot normally. Puffer fish lack three factors. The cascade arose by gene duplication, and the duplicates are identifiable by sequence |
| Immune system | Antibody generation could not arise stepwise | The RAG1/RAG2 recombination machinery derives from a transposon; the Transib transposon family carries recognisably related sequences |
| Eukaryotic cilium | Intraflagellar transport requires all parts | Simpler functional cilia exist with fewer components; the protein families have identifiable homologues |
| Mousetrap (analogy) | Removing any part makes it useless | Not biological, but the analogy has been answered: reduced-part traps that still function have been built and demonstrated |
Why the logic does not hold
The argument assumes evolution builds systems by adding parts to a non-functional precursor. Several well-documented routes produce a system in which every part is now necessary.
Scaffolding removal. A structure can be built with supports that are later discarded. A stone arch cannot stand while being built and cannot lose a stone once complete, yet arches are constructed routinely — the centring is removed at the end. The finished arch is irreducibly complex and was assembled gradually.
Change of function. A part performing one job can be recruited to another. The type III secretion system injects proteins; a subset of flagellar components does the same thing. A structure that works as an injector can become part of a motor.
Duplication and divergence. A gene is copied, and the copy diverges. The clotting cascade's factors are recognisably related to one another in sequence, which is the signature of repeated duplication from an ancestral protease.
Increasing dependence. Two components that initially function independently can come to rely on each other over time, a process Müller described in 1918 as "interlocking complexity." Each step is advantageous, and the end state cannot be reduced.
The court case
Behe testified as the lead expert witness for the defence in Kitzmiller v. Dover Area School District (2005), which is the only trial to have examined the argument under cross-examination.
Judge John E. Jones III, a Republican appointee and a churchgoer, ruled that intelligent design is not science and that the Dover policy violated the Establishment Clause. On irreducible complexity specifically, the ruling found the argument to be a negative argument against evolution rather than positive evidence for design, and that Behe's examples had been rebutted in the peer-reviewed literature.
Under cross-examination Behe acknowledged that his definition of science would also admit astrology, and was presented with a stack of books and papers on the evolution of the immune system, which he stated he had not read and which he maintained were insufficient.
The full ruling is available at Kitzmiller v. Dover.
Common objections
"Co-option doesn't explain the whole flagellum, only part of it"
Behe's strongest reply, and it is correct as stated. The type III secretion system accounts for roughly ten of the flagellum's proteins. The remainder — the motor, the hook, the filament — are not explained by that single step.
No complete step-by-step history of the flagellum exists, and claiming otherwise would misstate the evidence.
What the co-option finding does is defeat the argument's central premise. The claim was that no subset of the system could have a function. A functioning secretion apparatus built from a subset of those proteins shows that the premise is false, which is enough to void the inference even without a full pathway.
The remaining question — how the other components arose — is ordinary unsolved biology, not a demonstration that they could not have.
"Behe has answered these criticisms"
He has responded at length, in The Edge of Evolution (2007) and Darwin Devolves (2019), and has shifted the argument toward the probability of multiple coordinated mutations rather than the impossibility of intermediates.
That shift is itself informative. The original claim was that intermediates could not function; the revised claim is that the required mutations are too improbable, which is a different and weaker argument.
The revised version has also been contested. Behe's estimate of the "edge" relies on chloroquine resistance in malaria requiring two specific mutations, and subsequent work found intermediate single-mutation states that confer partial resistance, removing the requirement for simultaneity.
"The blood clotting cascade still can't lose parts"
Within a given species, largely true — removing most factors causes haemophilia.
The comparison across species is what answers it. Whales and dolphins lack Factor XII entirely and clot normally. Puffer fish lack Factors XI, XII, and the contact pathway. These animals have functioning cascades with fewer parts, which is precisely the state the argument says cannot exist.
The sequences also record their own history. The factors are related to one another and to trypsin-like proteases, with the pattern of similarity expected from repeated duplication.
"This is an argument from ignorance — you can't prove a negative"
Often made against Behe, and it overstates the case in one respect. Behe does offer a positive criterion: he claims that systems of this kind are known from experience to be produced by intelligence.
The criterion does not do the work required. Human designers are observed producing machines, but the inference from "designers make machines" to "this biological system was designed" requires that no other process could produce it, which returns the argument to its negative form.
The structural problem remains. Irreducible complexity identifies systems whose history is not yet known and infers design from the absence of an explanation, which makes the argument weaker as biology advances rather than stronger.
"Kitzmiller was one judge's opinion, not a scientific finding"
Correct. A court ruling does not settle a scientific question, and courts have been wrong.
The proceeding is nonetheless useful evidence, because it is the only setting in which the argument's leading proponent defended it under sustained cross-examination by opposing experts, with both sides able to present evidence.
The ruling's specific findings — that no peer-reviewed publication supports the claim, that the examples had been answered in the literature, and that Of Pandas and People had been produced by replacing "creation" with "intelligent design" after the Edwards v. Aguillard decision — are factual findings from evidence rather than opinions about biology.
What the evidence shows
Irreducible complexity describes a real property: many biological systems fail when a component is removed.
The inference from that property to design does not hold, because several documented processes produce systems in which every part is now required — scaffolding removal, co-option, duplication and divergence, and increasing mutual dependence.
Each of Behe's original examples has been answered in whole or in part, and in two cases — the clotting cascade and the flagellum — the answer takes the specific form the argument said was impossible: functioning systems built from fewer parts.
The related design argument is covered in Intelligent Design.