Home| Letters| Links| RSS| About Us| Contact Us

On the Frontline

What's New

Table of Contents

Index of Authors

Index of Titles

Index of Letters

Mailing List

subscribe to our mailing list:


Critique of Intelligent Design

Evolution vs. Creationism

The Art of ID Stuntmen

Faith vs Reason

Anthropic Principle

Autopsy of the Bible code

Science and Religion

Historical Notes


Serious Notions with a Smile


Letter Serial Correlation

Mark Perakh's Web Site

The Designer-of-the-Gaps Revisited

A critique of William Dembski's article "Irreducible Complexity Revisited"

Richard Wein

Posted February 15, 2004

"Ignorance more frequently begets confidence than does knowledge: it is those who know little, and not those who know much, who so positively assert that this or that problem will never be solved by science."

Charles Darwin


  1. Introduction
  2. The Argument From Irreducible Complexity
  3. Co-option
  4. Causal Adequacy
  5. The Connection With Specified Complexity
  6. Conclusion
  7. Notes
  8. 1. Introduction

    Intelligent Design advocates claim that an intelligent designer was involved in the origin of certain biological structures, and that this conclusion is justified by a scientific inference from empirical evidence. The main arguments which they offer in support of these claims are those of Michael Behe and William Dembski. Those arguments have appeared in a number of popular books and articles, though never in a formal technical paper, and have been widely criticized by biologists and others. The criticisms have been many and varied, but the primary one has been that, once shorn of error and obfuscation, all that's left is the age-old god-of-the-gaps argument, also known as the argument from ignorance. This argument is based on the undeniable premise that scientists have not fully explained certain observed phenomena, and insists that those phenonema therefore cannot have a natural explanation and so must be the work of God. ID advocates replace the word "God" with "a designer" (who they nevertheless believe to be God), but the basic form of the argument is otherwise unchanged.

    Since the god-of-the-gaps argument has long been rejected by scientists and philosophers of science, and ID advocates realize how weak the argument appears, they insist they have a stronger argument. They claim they can demonstrate that the probability of biological structures evolving by natural selection is so small as to make it effectively impossible for such evolution to occur. Even if this claim were justified, they would still be relying on a god-of-the gaps argument, though a somewhat stronger one: these structures cannot have evolved by natural selection and biologists do not have another explanation; therefore they must have been designed. However, their position is actually far worse than this, because their probabilistic claim is unsupported by any substantive argument. When pressed to justify the claim, they merely fall back on another argument from ignorance: biologists have not given a fully detailed account of how the structures evolved. To conceal the emptiness of their arguments, ID advocates confuse their readers by the use of idiosyncratic and ill-defined terms, equivocal claims, misleading rhetoric, and--in Dembski's case--irrelevant mathematical notation.

    This article will not discuss the merits of the god-of-the gaps argument. My purpose is merely to demonstrate that the arguments of ID advocates take this form. If they wish to attempt a justification of the god-of-the gaps argument, ID advocates are welcome to try. What is far less welcome, however, is their attempt to represent the argument as something more impressive, by hiding it behind a smokescreen of pseudoscientific mumbo jumbo.

    Dembski's fullest exposition of his argument for intelligent design occurs in his book No Free Lunch [1]. Although he dresses it up in misleading discussions of probablity, complexity, information and specification, the core of Dembski's case is the argument from irreducible complexity. It is essentially the same argument made by Behe in his book Darwin's Black Box [2]. I have previously criticized No Free Lunch in detail [3], but, despite writing two lengthy responses to my critique [4, 5], Dembski has entirely evaded addressing the substance of my arguments.

    Since the publication of No Free Lunch , Dembski has written a number of articles in which his argument appears to have changed significantly. He has apparently abandoned a key part of his argument, though he never clearly acknowledges this. Here I will address his argument as he presents it in his most recent article, Irreducible Complexity Revisited [6]. [Just before the current article was finalized, Dembski published a new book, The Design Revolution. I have not yet seen that book, so cannot comment on any additional or altered arguments it might contain.]

    2. The Argument From Irreducible Complexity

    Dembski 's latest definition of irreducible complexity is as follows:

    A functional system is irreducibly complex if it contains a multipart subsystem (i.e., a set of two or more interrelated parts) that cannot be simplified without destroying the system's basic function. I refer to this multipart subsystem as the system's irreducible core.

    As with previous attempts by Behe and Dembski to define irreducible complexity, this definition is vague and ambiguous in a number of respects. What constitutes a "part"? In his sole example (the flagellum of the bacterium E. coli), Dembski takes the parts to be proteins. But can a collection of proteins constitute a part? What about a piece of a protein? Also, what does it mean for a subsystem to be "simplified without destroying the system's basic function"? Is simplification limited to the removal of existing parts (as in Behe's original definition), or can it include the replacement of a part with a modified version? Can it include the replacement of the whole subsystem with a simpler subsystem operating on a quite different principle?

    Though Dembski gives only one example, he seems to assume that other irreducibly complex systems exist. It would be a mistake, however, to assume that the systems claimed as irreducibly complex by Behe are also irreducibly complex in Dembski's sense, since their definitions differ significantly. Without further clarification of Dembski's definition, it is impossible even to confirm that the bacterial flagellum is irreducibly complex in his sense.

    Fortunately, these questions need not concern us too deeply, since the concept of irreducible complexity turns out to be nothing but a red herring. Dembski argues that an irreducibly complex system cannot evolve by "a direct Darwinian pathway", which he defines as follows:

    A direct Darwinian pathway is one in which a system evolves by natural selection incrementally enhancing a given function. As the system evolves, the function does not evolve but stays put.

    Let us assume for the sake of argument that Dembski's conclusion is valid, and that the bacterial flagellum could not plausibly have evolved by a direct Darwinian pathway. Why should we care? Dembski's concept of direct Darwinian pathways is a straw man, an absurdly oversimplistic idea of evolution. In practice, biological systems are not restricted to maintaining the same function as they evolve. Their function can and does change. A system which has evolved for one use can be co-opted for another use. Co-option is discussed further in the following section.

    Clearly, having divided evolutionary processes into direct and indirect pathways, Dembski must make a convincing case against both possibilities if his overall case is to stand. So what is his argument against evolution by indirect pathways? Why, it's our old friend the argument from ignorance: biologists have not given fully detailed accounts of such pathways, so we must reject the possibility that they exist. To show that I am not misinterpreting him or omitting anything vital, I present his argument in full:

    How does the argument from irreducible complexity handle indirect Darwinian pathways? Here the point at issue is no longer logical but empirical. The fact is that for irreducibly complex biochemical systems, no indirect Darwinian pathways are known. At best biologists have been able to isolate subsystems of such systems that perform other functions. But any reasonably complicated machine always includes subsystems that perform functions distinct from the original machine. So the mere occurrence or identification of subsystems that could perform some function on their own is no evidence for an indirect Darwinian pathway leading to the system. What's needed is a seamless Darwinian account that's both detailed and testable of how subsystems undergoing coevolution could gradually transform into an irreducibly complex system. No such accounts are available or forthcoming. Indeed, if such accounts were available, critics of intelligent design would merely need to cite them, and intelligent design would be refuted.

    In making an argument from ignorance, it serves Dembski's purpose to exaggerate the ignorance of biologists. They may not have provided the exhaustively detailed ("seamless") account which he demands, but they have proposed outline accounts for several of the systems which Behe claims are irreducibly complex.

    As he has done in many previous articles, Dembski denies that he is making an argument from ignorance. But the denials are never substantiated. Instead, each one is followed by yet another argument from ignorance. All that changes is the wording. Here's another example:

    If after repeated attempts looking in all the most promising places you don't find what you expect to find and if you never had any evidence that the thing you were looking for existed in the first place, then you have reason to think that the thing you are looking for doesn't exist at all. That's the argument from irreducible complexity's point about indirect Darwinian pathways. It's not just that we don't know of such a pathway for, say, the bacterial flagellum (the irreducibly complex biochemical machine that has become the mascot of the intelligent design community). It's that we don't know of such pathways for any such systems. The absence here is pervasive and systemic. That's why critics of Darwinism like Franklin Harold and James Shapiro (neither of whom is an intelligent design proponent) argue that positing as-yet undiscovered indirect Darwinian pathways for such systems constitute 'wishful speculations.'

    Even if all its premises were justified, the above argument would remain an argument from ignorance. But the argument is even worse, in that at least one of the premises is false. It is not true that biologists have no evidence that indirect Darwinian pathways exist. Dembski is closing his eyes to all the palaeontological and biochemical evidence of co-option.

    3. Co-option

    Co-option (also known as co-optation) is a fundamental process of evolutionary biology in which structures which have evolved for one use become employed for a different use. One of the best known examples is the evolution of the mammalian inner ear from reptilian jaw bones, an evolutionary sequence which is clearly demonstrated by the fossil record. Simon Conway Morris, a leading evolutionary biologist, writes:

    Cooption is, therefore, commonplace (e.g., Finkelstein and Boncinelli, 1994; Holland and Holland, 1999), perhaps ubiquitous, and just what we would expect in organic evolution. [7]

    Indeed, co-option is not just advantageous but absolutely essential to evolutionary theory. Complex new systems cannot appear suddenly out of nowhere, so whenever an organism begins to do something new, it must do so by co-opting an existing struture which evolved for another use. This could be as simple as a fish using its fins--which evolved for swimming--for scrambling over rocks in shallow water. Over time, if the conditions persist, natural selection may gradually lead to the fins becoming better adapted for scrambling and less useful for swimming. Eventually scrambling may become their primary function.

    Co-option has always been a thorn in the side of proponents of the argument from irreducible complexity, as it negates their simplistic arguments. In Darwin's Black Box, the first book to elaborate the argument, Behe employed a dichotomy, dividing evolutionary pathways into two groups, direct and indirect, the same division that we see in Dembski's recent articles. He presented a (rather muddled) argument against evolution by direct pathways, while failing to come up with any substantive argument at all against indirect pathways.

    When Dembski employed his own version of the argument in No Free Lunch, he abandoned the obfuscatory direct/indirect dichotomy, apparently believing that he instead had a knock-out argument against co-option. But the co-option scenario which he argued against was a straw man, another oversimplistic reading of evolutionary theory. It hypothesized that all the proteins making up the flagellum evolved independently for other purposes and were then simultaneously co-opted to produce a flagellum. He failed to consider the more realistic hypothesis that a substantial exisiting system, consisting of many proteins, was co-opted as a single unit. Yet this is exactly the type of hypothesis which biologists propose. Having overlooked this possibility, Dembski erroneously claimed that he had considered and rejected all possible options, and that he had therefore decisively eliminated Darwinian evolution as a reasonable explanation for the flagellum:

    These two conditions transform the definition of irreducible complexity into a vise that allows the Darwinian mechanism no room to maneuver. ... Let me stress that there is no false dilemma here--it is not as though there are other options that I have conveniently ignored but that the Darwinian mechanism has at its disposal. [8]

    Since the publication of No Free Lunch, Dembski's error has been pointed out to him and he appears to have recognised it. Though not acknowledging that he made a mistake or that he has changed his argument, Dembski has abandoned the claim above and fallen back on the argument from ignorance, while adopting Behe's direct/indirect dichotomy as a new smokescreen. He also now mentions the possibility which he previously overlooked:

    But what if instead co-option occurred more gradually and incrementally? In the evolution of the bacterial flagellum, imagine natural selection gradually co-opting existing protein parts into a single evolving structure whose function co-evolves with the structure.

    Here, Dembski is still limiting evolution to adding existing protein parts to a structure (why not larger parts?), but at least he is now considering the possibility that the function of the whole structure may change over time. Unfortunately for him, however, the only argument he can find to use against this possibility is another argument from ignorance:

    Minimally what's required are detailed, testable reconstructions or models that demonstrate how indirect Darwinian pathways might reasonably have produced actual irreducibly complex biochemical machines like the bacterial flagellum.

    4. Causal Adequacy

    In No Free Lunch Dembski appealed to a concept he called causal specificity. Never one to eschew the opportunity to invent another useless technical term, Dembski now replaces causal specificity with causal adequacy:

    This is where the argument from irreducible complexity needs to make a third key point, namely, an explanatory point. Scientific explanations come in many forms and guises, but the one thing they cannot afford to be without is causal adequacy. A scientific explanation needs to call upon causal powers sufficient to explain the effect in question. Otherwise, the effect is unexplained. The effect in question is the irreducible complexity of certain biochemical machines. How did such systems come about? Not by direct Darwinian pathways--irreducible complexity rules them out on logical and mathematical grounds. And not by indirect Darwinian pathways either--the absence of scientific evidence here is as complete as it is for leprechauns. Nor does appealing to unknown material mechanisms help matters, for in that case not only is the absence of evidence complete but also the very theory for which there's no evidence is absent as well.

    Thus, when it comes to irreducibly complex biochemical systems, there's no evidence that material mechanisms are causally adequate to bring them about. But what about intelligence? Intelligence is well known to produce irreducibly complex systems (e.g., humans regularly produce machines that exhibit irreducible complexity). Intelligence is thus known to be causally adequate to bring about irreducible complexity. The argument from irreducible complexity's explanatory point, therefore, is that on the basis of causal adequacy, intelligent design is a better scientific explanation than the Darwinian mechanism for the irreducible complexity of biochemical systems.

    As usual, Dembski doesn't clearly define his term. A causally adequate (or "sufficient") explanation might be taken to be one which necessitates the observed outcome. But this is not how Dembski uses the term. Clearly, an intelligent designer does not necessitate the occurrence of irreducibly complex biological systems. The designer might have chosen not to produce any organisms at all. It appears that when Dembski says an explanation is causally adequate he means that the causes invoked by the explanation could have produced the result to be explained. His claim then seems to be that natural evolution could not have produced the irreducibly complex biological systems that we observe, but an intelligent designer could have done so. We have already seen that the first part of this claim (that natural evolution could not have produced irreducibly complex systems) is based on nothing more than an argument from ignorance.

    However, Dembski's argument here is not just about whether certain causes could have produced the observed result, but whether there is evidence that they could have done so. He claims there is no evidence that natural evolution could have produced irreducibly complex biochemical systems. This is untrue. Presenting the evidence for natural evolution is far beyond the scope of this article. Broadly speaking, it consists of the multitude of examples of ad hoc adaptation of organisms to their environmental conditions, adaptations which show no sign of direction towards any goal. This evidence and the absence of evidence for a designer make natural evolution the best explanation of the biological systems we observe (irreducibly complex or otherwise). In the absence of any good argument against the natural evolution of irreducibly complex systems, there is no reason to treat them as a special case. Furthermore, theoretical evidence of the power of natural selection to produce complex systems is provided by computerized evolutionary algorithms [9].

    Dembski goes on to argue that intelligent design is known to be causally adequate to explain irreducibly complex systems, based on the assertion that "humans regularly produce machines that exhibit irreducible complexity". Let's accept for the sake of argument that humans really do produce machines that exhibit irreducible complexity in Dembski's sense of the term (though he gives no examples). It does not follow that an intelligent designer could have produced the irreducibly complex biological systems that we observe. Dembski offers no evidence of the capabilities of any designer who was in a position to design biological systems. Nor does he offer any evidence that such a designer could have existed. Thus, Dembski is very far from demonstrating the causal adequacy of his explanation. He is just resorting to an old chestnut of religious apologetics in which divine explanations are invoked while ignoring all the explanatory problems associated with the nature and existence of gods. Replacing "divine" with "design" does not make this style of argument any more convincing.

    5. The Connection With Specified Complexity

    In his latest article, Dembski repeats the claim--made in many previous articles and books--that his notion of specified complexity provides a reliable criterion for detecting design. I dealt with this claim at great length in my critique of No Free Lunch [3], showing that Dembski's use of the term specified complexity was equivocal and deceptive, serving only as a smokescreen for his god-of-the-gaps argument. I will give an updated but much briefer treatment of the subject here.

    The first thing to note is that Dembski does not use the word complexity in its everyday sense, nor in any pre-existing technical sense. Instead, he uses it in a misleading sense of his own: for Dembski, complexity is defined to mean improbability. He also uses the word information in the same idiosyncratic sense. As a result, Dembski's specified complexity and information have quite different meanings from these same terms as used by other writers. Yet Dembski shamelessly conflates his own idiosyncratic meaning with those others, causing endless confusion. Of course, when Dembski means improbability, he should write "improbability", and not mislead his readers by calling it complexity or information.

    Further confusion is caused by Dembski's persistent refusal to clarify what improbability he is referring to. Sometimes he seems to mean the improbability of an event with respect to a specific natural hypothesis, and usually this is a hypothesis involving a uniform probability distribution, such as a hypothesis of purely random combination of components. At other times, he seems to mean that an event is complex if it is improbable with respect to all the natural hypotheses we can think of. At still other times, he seems to mean that an event is complex if it is improbable with respect to all possible natural causes.

    In his latest article, Dembski continues his policy of equivocation. First he writes that:

    Darwinists object to this approach to establishing the specified complexity of irreducibly complex biochemical systems. They contend that design theorists, in taking this approach, have merely devised a 'tornado-in-a-junkyard' strawman.

    What evolutionists object to is the habit that antievolutionists have of calculating a probability based on a hypothesis of purely random combination and then presenting this as the probability of the system evolving. Dembski performed just such a calculation in No Free Lunch [10], but has been extremely vague about just what relevance it has to his argument. In previous articles, he has implied that this probability calculation takes natural selection into account, even though it does not do so:

    Convinced that the Darwinian mechanism must be capable of doing such evolutionary design work, evolutionary biologists rarely ask whether such a sequence of successful baby-steps even exists; much less do they attempt to quantify the probabilities involved. I attempt that in chapter 5 of my most recent book No Free Lunch . There I lay out techniques for assessing the probabilistic hurdles that the Darwinian mechanism faces in trying to account for complex biological structures like the bacterial flagellum. The probabilities I calculate--and I try to be conservative--are horrendous and render natural selection entirely implausible as a mechanism for generating the flagellum and structures like it. [11]

    He includes a similarly deceptive implication in his latest article:

    Yet even with the most generous allowance of legitimate advantages, the probabilities computed for the Darwinian mechanism to evolve irreducibly complex biochemical systems like the bacterial flagellum always end up being exceedingly small.29

    Footnote 29 here is Dembski's, and refers to section 5.10 of No Free Lunch, the whole of which is concerned with a probability calculation based on a hypothesis of purely random combination. In fact, Dembski misleadingly fails to state in No Free Lunch that his calculation is based on such a hypothesis (even though his general method of inferring design requires him to identify "all the relevant chance hypotheses" [12]), but it follows from an analysis of the calculation, and he has since acknowledged that this is the case:

    Wein therefore does not dispute my calculation of appearance by random combination, but the relevance of that calculation to systems like the flagellum. [4]

    This is the only probability calculation which Dembski has ever provided for the origin of a biological system. In his latest article, he gives a lengthy description of how one might go about performing such a probability calculation, and his aim here does seem to be to take natural selection into account. But his discussion remains purely theoretical, since he does not provide any figures.

    He also lists a number of alleged hurdles that evolution must overcome. These are all rather vague and based on oversimplistic views of evolutionary biology. Some of them also involve more appeals to ignorance. I won't deal with them individually. As long as the argument from irreducible complexity remains the mainstay of Intelligent Design advocates, there is no reason why critics should spend their time dealing with additional arguments which are presented only vaguely. If Dembski believes he has found a better argument than the argument from irreducible complexity, let him say so and present it in a professional manner.

    6. Conclusion

    The arguments for Intelligent Design are purely negative. Rather than offer evidence of the existence of any designer, they attempt to rule out natural evolution of certain biological structures. One such argument, the current favourite of the Intelligent Design movement, is the argument from irreducible complexity. However, instead of addressing evolutionary theory as actually propounded by evolutionary biologists, the substantive part of the argument is aimed at a simplistic parody of the theory which ignores one of its most fundamental elements, namely co-option of existing structures. The remainder of the argument is just a disguised version of the old argument from ignorance, or god-of-the-gaps argument. Intelligent Design advocates are forced to rely on the god-of-the-gaps argument because they are committed to detecting the action of an intelligent designer in biology despite the inescapable fact that no evidence of such a designer exists.

    7. Notes

    1. William Dembski, No Free Lunch: Why Specified Complexity Cannot be Purchased without Intelligence, Rowman & Littlefield, 2002.

    2. Michael Behe, Darwin's Black Box (Simon & Schuster, 1998), pp. 39-40.

    3. Richard Wein, "Not a Free Lunch But a Box of Chocolates: A critique of William Dembski's book No Free Lunch", The Talk.Origins Archive, May 2002, http://www.talkorigins.org/design/faqs/nfl. Also posted at TalkReason.

    4. William Dembski, "Obsessively Criticized But Scarcely Refuted: A Response To Richard Wein", May 2002, http://www.designinference.com/documents/05.02.resp_to_wein.htm.

    See also my response to this article:
    Richard Wein, "Response? What Response? How Dembski has avoided addressing my arguments", The Talk.Origins Archive, May 2002, http://www.talkorigins.org/design/faqs/nfl/replynfl.html. Also posted at TalkReason.

    5. William Dembski, "The Fantasy Life of Richard Wein: A Response to a Response", June 2002, http://www.designinference.com/documents/2002.06.WeinsFantasy.htm.

    6. William Dembski, "Irreducible Complexity Revisited", January 2004, http://www.designinference.com/documents/2004.01.Irred_Compl_Revisited.pdf.

    7. Simon Conway Morris, "Evolution: Bringing Molecules into the Fold", Cell, Vol. 100, 1-11.

    8. No Free Lunch, p. 287.

    9. Richard Lenski, Charles Ofria, Robert Pennock & Christoph Adami, "The evolutionary origin of complex features", Nature, May 2003.

    10. No Free Lunch, pp. 289-302.

    11. William Dembski, "Does Evolution Even Have a Mechanism?", April 2002, http://www.designinference.com/documents/04.02.AMNH_debate.htm.

    12. No Free Lunch, pp. 72 & 123.