Darwin's Black Box at 30: irreducible complexity, co-option, and the bacterial flagellum

Darwin's Black Box created a major controversy and irreducible complexity remains influential.

Finding: partly supported, but not probative. The book clearly had public and institutional impact. Attention, sales, and continuing advocacy do not establish the correctness of its biological argument.

Article · TalkOrigins overview

Lehigh's disclaimer about intelligent design shows disciplinary embarrassment or suppression.

Finding: unsupported inference. The official page states the department's position and identifies Behe as a dissenting faculty member. It does not demonstrate coercion, suppression, or the truth of either side's scientific claims.

Lehigh statement · Article

Axe's 2004 experiment definitively shows functional protein classes are highly isolated in sequence space.

Finding: overstated. Axe measured functional prevalence under a defined assay and protein design. That is relevant evidence about constraints in that experimental landscape; it is not a proof that all functional transitions lack selectable intermediates or that evolution must leap between isolated endpoints.

Axe 2004 · Harms & Thornton 2013

The bacterial flagellum is irreducibly complex because removing any part destroys motility.

Finding: valid removal observation, invalid historical conclusion. Dependence in an extant machine does not show that every precursor had to perform motility by the same mechanism or that its components could not have had earlier roles.

TalkOrigins definition and context · Pallen & Matzke 2006

Ken Miller evades irreducible complexity by redefining it whenever a remnant can perform another function.

Finding: mischaracterized. The point of an alternate function is not to change the removal test. It is to challenge the assumption that the present system's basic function was the only selectable function available throughout its history.

TalkOrigins: Mullerian two-step · Gould & Vrba 1982 · Miller, Finding Darwin's God

A mousetrap base used as a doorstop does not refute irreducible complexity because it no longer catches mice.

Finding: formally correct but historically incomplete. A stripped-down flagellum need not already propel a bacterium for co-option to be relevant. The evolutionary question is whether parts or subsystems had selectable functions before later additions, modifications, or dependencies produced the present machine.

Article · TalkOrigins: Mullerian two-step · Gould & Vrba 1982

Muller predicted interlocking complexity as an expected consequence of evolution.

Finding: directly relevant and underweighted in the source article. In 1918 and 1939, H. J. Muller described how a feature that was initially merely advantageous or dispensable could become indispensable when later changes made other processes depend on it. This anticipates the distinction between an existing system's present dependence and the historical functions of its parts.

This does not by itself reconstruct the bacterial flagellum. It does refute the claim that an alternate earlier function is irrelevant in principle. It also shows that the core response predates both Behe's terminology and modern molecular-machine debates.

Muller 1918 · Muller 1939 · TalkOrigins discussion

The debate applies a “poof” standard asymmetrically.

Finding: interpretive asymmetry. Dembski treats evolutionary explanation as if finding functional precursors would require a sudden, functionless “poof” event to make the pathway believable. Biologists generally reject “poof” wherever it is asserted and instead test functional precursors, co-option, duplication, recruitment, and interlocking dependencies.

A precursor with a different function does not answer every historical question. But dismissing it because it does not perform the later system's function imposes an extra requirement on evolution. The article does not supply an equivalently specified mechanism for design insertion. This is a finding about the article's argumentative structure, not a claim that every creationist or intelligent-design advocate reasons identically.

Dembski article · Dembski 2004 · TalkOrigins

The type III secretion system cannot illuminate flagellum evolution because secretion is not motility.

Finding: present functions differ; conclusion does not follow. Homologous export machinery can be recruited into systems with different functions. Comparative work treats this functional difference as the evolutionary question, not as a reason to dismiss the homology.

Pallen & Matzke 2006 · Abby & Rocha 2012

Because animals appeared after bacteria, the flagellum must precede the non-flagellar T3SS.

Finding: partly supported but oversimplified. A flagellum-first direction is supported by important comparative work, but the inference depends on phylogenomics and homology, not merely broad organismal chronology. Earlier studies proposed a shared ancestor or competing histories.

Abby & Rocha 2012 · Gophna, Ron & Graur 2003

No plausible intermediate steps have been identified for the bacterial flagellum.

Finding: false as a blanket statement. Published models use ancestral cores, gene duplication, modification, deletion, recruitment, and subsystem co-option. They are historical reconstructions—not a complete replay in the laboratory—but they are more than an assertion that the parts appeared simultaneously.

Liu & Ochman 2007 · TalkDesign model · Abby & Rocha 2012

It is easier to remove parts than to add coordinated parts, so simplification is evolution's primary trajectory.

Finding: possible in some histories, not a general law. Deletion is one mechanism among duplication, recruitment, horizontal transfer, mutation, and regulatory change. The direction must be tested for each system.

Abby & Rocha 2012 · Puente-Lelievre et al. 2025 · Andrianova et al. 2025

Irreducible complexity remains decisive because no detailed Darwinian accounts exist for biochemical systems.

Finding: too broad and incomplete. Some histories remain incomplete, but incomplete reconstruction is not a failure of the evolutionary framework. The Krebs (citric-acid/TCA) cycle is a direct biochemical counterexample: published chemical-design analysis models its opportunistic assembly from reactions with earlier metabolic uses, while later experiments identify chemically feasible reverse-cycle and precursor networks. These studies do not settle every ancestral order or date, but they make pathway evolution a testable historical and mechanistic problem.

Meléndez-Hevia et al. 1996 · Muchowska et al. 2017 · Muchowska et al. 2019 · Liu & Ochman 2007 · Trotter et al. 2014 · TalkOrigins book review

Continuing criticism and unresolved details show that intelligent design is ascendant and is the best explanation.

Finding: not established. Attention, institutional conflict, and gaps in historical reconstruction do not supply a positive design mechanism or discriminating prediction. This conclusion is an inference from perceived difficulty, not a demonstrated causal explanation.

Dembski 2004 · TalkOrigins · TalkDesign