A Critique of Behe, Dembski on “Irreducible Complexity”

[Note: Footnotes (superscripted) are at bottom–sorry, you gotta scroll up and down to read them; citations (parenthetical references at end of sentences) are not completed yet. Some say “DBB” for “Darwin’s Black Box” or something like that, but I have to track down the page numbers when I get time.]

Michael Behe, a biochemist and author of Darwin’s Black Box: The Biochemical Challenge to Evolution, has no quarrel with the “Darwinist” interpretation of the fossil record, rendering of geological ages (including belief in an “old” earth), or the claim that later species—including humans—evolve from ancestral species. Nor is there evidence that he is, professionally speaking, interested in the origins of the world, or of earth, as a whole. However, his work has gained the favorable attention of creation theorists who are anti-Darwinist on each of these counts. One way to highlight the continuity here is to note that Behe’s work takes its cue from the creationist adaptation of the classic “argument from design,” specifically the claim that certain natural structures display such complexity that their emergence was unlikely to have occurred minus (and therefore are best explained by) the deliberate intervention of some intelligent agency. Behe’s unique contribution to this line of thought is his application of (a form of) the argument to the cellular micro-mechanisms which are the subject matter of his field. These include the blood clotting factor in vertebrates, the cilia and flagella which motor bacteria, and production of antibodies by the immune system. His central contention is that the complexity of biochemical structures such as these is irreducible, that is, they display an interdependence of components such that the removal of (or their “reduction” by) any one of these components renders the system unworkable—or rather, that this feature proves the work of “intelligent design” (ID).

Though joined in Behe’s central concept, “complexity” and “irreducibility” can nonetheless be logically distinguished, and to some degree Behe separates them in his presentation. (1) He follows prominent Creationists in defining complexity as a property of natural systems, roughly, a function of the number of components of the system and the number of interconnections between these components. More importantly, he adopts a popular Creationist tack of giving complexity a probabilistic basis, prior to the introduction of irreducibility: In short, the greater the complexity of a biochemical structure, the lower the probability that the protein sequence describing the structure could have emerged by a spontaneous, random (that is, unguided (by “intelligence”)) combination of individual proteins. (“A system of low probability has a much better chance of emerging spontaneously as a result of a chain of random events” (DBB #).)

(2) Behe’s unique contribution is the “irreducible” modifier. This serves not only to increase the complexity of his structures in the presentation—that is, it enables Behe to calculate the probability of their random emergence as even smaller than were these structures merely “reducible”—but more crucially, to suggest that these structures are unaccountable for along the lines of any Darwinian narrative: Since the cellular mechanism in question is advantageous to the organism which uses it only as a whole, it could not have evolved by the slow, step-by-step (that is, “part” by “part”) accretive process described (on Behe’s reading) by the theory of natural selection.

Behe’s key thesis of ID is offered as the third prong of an exhaustive trichotomy of hypotheses explaining the emergence of biochemical structures: If (1) the structure’s random, immediate self-construction is too improbable to take seriously, and (2) its emergence by way of Darwinian natural selection is impossible, then (3) the theory of ID must be inferred. The present essay will focus on the concept of irreducible complexity (IC), paying less attention to the ID inference, that is, the contention that ID follows from IC.

On ‘Simple’ Complexity¹

Some creation theorists have openly cited the French mathematician Borel’s “Single Law of Chance,” which suggests that phenomena with a probability of occurrence lower than 1/1050 simply do not occur. Short of naming Borel (or his creationist paraphrasers) by name, DBB repeats this exact idea in laymen’s terms. While refusing to specify any lower limit on “happening,” he offers the composition of biochemical processes (when construed randomly as in (1)) as examples of “prohibitively improbable” events.

Behe likely figures the probability for clotting factor incorrectly, but we needn’t address his math to critique the assumptions guiding it. First, low probability is not the same as impossibility; it cannot absolutely determine against an event’s coming to be. Indeed, on reflection, it is clear that events whose probability is extremely low happen on a daily basis; in fact, ironically, it is simple enough to manufacture a situation in which one such event is virtually bound to happen. An example comes from a fellow ID theorist (though created to illustrate a point very different from ours) whom Behe is certain to have read: Constructing a die with the letters A though F painted across its six sides makes it possible to engineer an event consisting of 100 die-tosses. If we record the results of each successive toss, whichever combination of 100 letters is yielded will have had a probability of approximately 1/1077. The divisor of this number exceeds by 43 orders of magnitude a figure to which Behe himself appealed—calling it “horrendously large”—in calculating the improbability of a ‘randomized’ (unguided) blood clotting mechanism. (The fraction itself is by 28 orders of magnitude smaller than Borel’s lower limiting case, to which Behe may or may not be committed.)

It follows that very low probability is not a condition sufficient to eliminate the chance of occurrence, nor conversely to premise a proof of the guiding hand of ID. Behe himself seems to acknowledge this, by implication, in a preface to a colleague’s work which was written after DBB. In William Dembski’s Intelligent Design, he writes:

For example, if we turned a corner and saw a couple of Scrabble letters on a table that spelled AN, we would not, just on that basis, be able to decide if the were purposively arranged. Even though they spelled a word, the probability of getting a short word by chance is not prohibitive. On the other hand, the probability of seeing some particularly long sequence of Scrabble letters, such as NDEIRUABFDMOJHRINKE, is quite small (around one in a billion billion billion). Nonetheless, if we saw that sequence lined up on a table, we would think little of it because it is not specified [my emphasis]—it matches no recognizable pattern. But if we saw a sequence of letters that read, say, METHINKSITISAWEASEL, we would easily conclude that the letters were intentionally arranged that way. The sequence of letters is not only highly improbable, but also matches an intelligible English sentence. [Therefore,] It is the product of intelligent design.

Note that Behe’s “particularly long sequence of Scrabble letters” plays here precisely the role of the 100-letter string produced by the die in our example. This is very significant, for it makes clear that by the time of this writing Behe has come to share our contention that low probability by itself fails to warrant the inference of ID. Behe’s illustration, and its choice of the emphasized term, makes it also clear that he has adopted Dembski’s solution to the resultant theoretical “gap”: With respect to some event, Dembski regards low probability as a necessary but not sufficient condition to infer ID, becoming sufficient only when joined by a second criterion, which he calls “specification.” Loosely put, an event is specified when it displays a recognizable pattern. In turn, recognizability entails “detachability,” in the sense that to recognize a pattern as such requires that we have in mind some image, independent of—or “detached” from—the pattern observed. (Put crudely, this “image” is the thing we are ‘cognizant’ of, in the light of which the specification becomes a re-cognition.)

These criteria appear to be met by the METHINKS… sequence. Indisputably, the probability is very low that the sequence could have come about by chance; and in some sense, we carry an understanding of this phrase which is prior to and independent of seeing it on the table—we would have recognized it anywhere.2 Finally, let us concede that these features together are what justify our inference that some (albeit human) intelligence has arranged the letters deliberately.3

However, it is less clear how his appeal to Dembski can help Behe when it comes to inferring ID from his featured biochemical structures. For whatever patterns these machines can be said to display are not of a “detachable” sort at all! For example, no kind of recognition marked the discovery of one of these systems. That is, its discoverer would not have had among his or her background knowledge some sample to which he might compare the newfound structure, yielding an understanding of it analogous to that of comprehending a phrase in one’s home language.4

On Irreducible Complexity

We shall consider whether “irreducibility” can be added to “complexity” to justify the design inference. In considering the question I shall use Behe’s “mousetrap analogy” as a kind of organizing device for critiquing the basic IC concept. First, however, some preliminary impressions concerning the overall structure of his argument are in order.

Some Preliminary Impressions on Argument from Ignorance, Arbitrary Classification

First, one is struck by what is essentially the “argument from ignorance” on Behe’s behalf. Recalling the ‘trichotomy’, Behe says plainly that the main “evidence” for the ID inference is that he “cannot imagine” either a possible pathway by which an (allegedly) IC structure could have evolved, nor, alternately, a workable version of the structure which is simpler (by one or more parts). I’m not prepared to say that this style of reasoning is never warranted. But “if evolution cannot explain biochemistry now, it can never explain; it cannot explain it in principle” seems a nonsequiter. Moreover, an argument from ignorance is disturbing in a natural-scientific context because it seems to be, at best, an untestable hypothesis, or at worst, a non-hypothesis altogether and therefore a research-terminus.

Relatedly, note how Behe casts the Darwinian “frontier” of ignorance cleanly along microscopic-macroscopic lines. Recall, he is willing to accept evolution by natural selection as holding at the organismic level, and his reasoning for this is the success it has had in explaining the origin of some organisms. Conversely, he contends that evolution fails to explain any of the biochemical machinery found in every living cell, so it is a dead (inapplicable) theory at the sub-cellular level. This is clearly a false rendering of the “frontier”; as we shall see, we do know something about the evolution of (even (apparently) irreducible) microscopia, after all. But what is interesting in its own right is the hasty significance Behe reads into this division, once, for good or for ill, it is made. In other words, even if it were the case that we know how several macroscopic organisms evolved, but not how any microscopic machines evolved, this fact would not justify Behe’s selective abandonment of the microscopic domain to ID (or, sooner, his abandonment of such as inexplicable with respect to Darwinism).

Consider: By analogy, a scientist (assume she shares at least Behe’s “portion” of Darwinism), when faced with some finite number of (plausible) Darwinist macro-narratives, could conclude that those other macro-organisms for whom narratives are not presently available are for that reason beyond the reach of Darwinism. In a kind of nominalist moment, she could easily manufacture some feature common to all the “storied” macro-organisms and absent from the “unstoried” macro-organisms and claim that Darwinism fails as a general theory because it only explains the first ‘type’ or ‘domain’—she could now use such a term—of entity. But of course, we would demand that she produce not some feature(s) in view of which, or purposes for which, the ‘domains’ can be considered different, but rather some feature which is relevant to the investigation of natural selection. Short of such provision, we accuse the scientist of a proffering an arbitrary distinction (or at best, an arbitrary employment or application of a distinction).5

Of course, real scientific practice is something like the reverse of hers: With apologies to Popper, showing how certain macroscopic structures must have evolved lends a kind of increasing credence to claims that the other, as yet “unshown” macroscopic structures also evolved. This is because whatever evidence we do have bolsters a hypothesis which makes claims over both nominal ‘groups’ of organisms. In this way hypotheses are best seen as research programmes; they are less true or false than promising or (relatively) unpromising. Their promise is enhanced precisely because of what they have already explained or predicted. But, as “promise” would connote, the claims they make always outpace the “verification” (again, apologies to Popper) of these claims. So, returning to Behe, for him to note that the Darwinist bundle of “hypotheses” has not yet realized its promise of accounting for microscopia is just to repeat that, indeed, it is hypothetical; and, therefore, more is needed of Behe to justify a preemptive truncating of its applicability than the simple observation that it has not been tested (“verified,” applied) in this domain.

In conclusion, it is incumbent on Behe to show what is in principle special about microscopia such that we should retain our (his own) promise that Darwinism can explain the remaining “unstoried” macro-organisms, while abandoning all of the former domain to the design inference. More positively, while Behe writes as though we must know all things to know any, an inductive argument that “Darwinism explains pn phenomena; therefore, it may very well explain biochemistry too” maintains until we see a reason to think otherwise; this is no less the case for evolutionary biology than for any other strong programme. How far time and research may extend the confirmed reach of Darwinism is (yet?) indeterminate. Complicating the question are the newness of the field of molecular evolution and relative lack of grant-funding (the “payoff” to biochemical research is often less dramatic than in other fields; part of this is due to the same newness of course). In addition are the resistance of biochemical structures to fossilization and the fact that these structures are very old, common to most eukaryotic cells, both features making their evolutionary pathways all the harder to reconstruct. All of this would lead us to expect a “lag” behind macro-research, but does not, I think, justify a despairing of the same.

Some Preliminary Impressions on the Definition of a System, IC

A related point concerns the failure of Behe to adequately define a “system.” That is, it isn’t always clear when one system ends and another begins. Naturally this affects whether or not we judge a candidate for system to be IC. For one, one must isolate a particular system for examination before it can be judged IC or not; for another, Behe defines IC partly in terms of its being a property of a system, so the latter has a kind of conceptual priority to the former. Behe frequently speaks of IC in terms of systems of systems; for instance: “Since each step necessarily requires several parts, not only is the entire blood-clotting system irreducibly complex, but so is each step in the pathway” (DBB 87). This in turn leads to difficulty in defining a “component” of a system and distinguishing it from the system-proper within which it operates.

Most likely, a “system” is something one isolates for certain purposes, while for other purposes the same entity may be better seen as a component of a broader system. This “relativity” suggests another criticism of Behe’s macro-micro dichotomy. Vertebrate biology on a macro-scale offers examples of the “crucial interdependencies” which are a hallmark (Behe says this much) of biochemical IC structures, tempting one to label them IC as well: A heart is useless without a neural system to tell it to pump, a brain cannot function without a respitory system to provide it with oxygen or a digestive system to provide it nutrients, and so on. Darwinism, and indeed the portion that Behe himself admits, accounts for the gradual, simultaneous evolution of the component “sub-structures” of the vertebrate body. Behe must explain, then, why he dismisses the possibility that the sub-systems of a broader biochemical system could have evolved in like manner, their interdependence even becoming irreducible at a (later) point.. (This possibility will be explored further later.)

The Mousetrap Analogy

To illustrate his thesis of IC in everyday terms, Behe offers the example of the standard 5-piece (not including staples) spring-snap mousetrap. The device consists of (1) a wooden base, (2) a hammer, (3) a spring which provides force to the hammer, (4) a holding bar which restrains the hammer when the trap is set, and (5) a catch which secures the holding bar when the trap is set and provides a trigger for the trap. (The cheese, a sixth part, would sit on the catch. But this part is inessential, as a mouse unseeking food could still accidentally walk across the catch.) Since, says Behe, the mousetrap depends on all five parts working in concert to catch a mouse, it is IC. This feature makes it impossible to take apart the trap and then “reverse” this process to build up a sequence of five mousetraps, each with one part more than its predecessor model and therefore slightly better at catching mice. “[Y]ou cannot catch a few mice with just a platform, add a spring and catch a few more mice, add a holding bar and catch a few more…All the components have to be in place before any mice are caught” (MC 178).

Behe sees this IC quality as analogous to is biochemical systems. (a) Reduction by a single “part”—in this case, an individual protein—would render the entire system inoperable; the system is IC; (b) Like “reverse engineering” the mousetrap, (Darwinian) evolution by natural selection describes a stepwise, or part-by-part, process of slight modifications; (c) Of course, only beneficial modifications (mutations) are selected for (ones that increase the fitness of the organism, i.e., which modify a system in a way which enhances fitness for the organism possessing it). Therefore, (d) such a system “cannot have evolved in a Darwinian fashion” (DBB 25): A Darwinian “pathway” to the final system would require a series of “reduced” intermediate versions of the final system—which would be either of zero benefit, or harmful, to the organism—and therefore nonselectable.

A Critique of the Mousetrap Analogy

It is actually remarkably easy to undercut Behe’s analogy. For instance, anyone can purchase a mousetrap, remove the catch and instead wedge the holding bar under the edge of the spring, or even push it into the soft wood where the catch used to lie. This version of the trap is less effective at catching mice than the original—it is less stable, for one, prone to springing without a mouse’s “aid.” But it is better than no trap at all. John McDonald of the University of Delaware has created a series of drawings whereby Behe’s mousetrap is reduced further from the 5-part original to a 4-part, 3-part and 2-part model, down to a 1-part model amounting to a simple bent spring. The greater the reduction, the less effective each trap becomes at catching mice; but at no step has the trap become dysfunctional.

McDonald’s series is most readily useful on two counts: First, it speaks directly to the problems with Behe’s “argument from ignorance” by showing how a structure which appears to be IC at first may be in time shown to be reducible. It illustrates how the appearance of IC can reflect a lack of creativity on an observer’s part, or a failure to conduct the right experiments. (By analogy, “experimenting” with the mousetrap was essential to McDonald’s envisioning alternate pathways.) Therefore, time may yield a proper Darwinian narrative for molecular structures, and terminating lines of research in view of an ID inference could work against acquiring the real story. This begs the more fundamental problem of how to recognize IC, and whether it can ever be done.

(I) Showing that What Appears to be IC is Actually Reducible

Second, McDonald’s series exemplifies one way of responding to Behe’s charge that biochemical structures are irreducible and therefore designed: As with the trap, we might show that what appears to be biochemical irreducibility simply is not.

On this count, consider the Krebs Cycle (also known as the tricarboxylic acid or citric acid cycle). This is a key metabolic operation of the cell, performing many crucial functions. It is the second stage in glucose metabolism, produces units of carbon, etc. Though Behe does not mention Krebs by name, to all appearances the cycle is an IC biochemical system. Using Behe’s criterion, one cannot imagine which enzyme could be removed, leaving a functioning cycle intact. However, in some systems there is manifest a shorter pathway cutting across the cycle, called the “glyoxylate shunt.” Note that this variant route was discovered after discovery of the (apparently irreducible) original. Further research revealed that some species lack a Krebs cycle altogether, along with the gene sequence that determines it. These discoveries demonstrated that Krebs is not reducible, but one of various pathways to the same function(s).

Another example speaks directly to Behe’s most-cited example of IC. In a unique anti-ID polemic by Shanks and Joplin, there is described an experiment led by Thomas H. Bugge in the mid-1990’s which drew upon the then-fresh technique of biochemically deleting selected genes from an animal’s genome. From one group of mice, researchers removed a gene responsible for the production of fibrinogen, an enzyme crucial for blood clotting. From a second group was removed a gene responsible for the production of plasminogen, which helps to curb blood clotting from advancing to the point of trombosis. As predicted, the first group of mice suffered hemorrhages and the second were plagued with trombotic problems. However, when researchers cross-bred the two groups, the forthcoming generation of mice, while lacking both of the isolated genes, were free of the problems associated with the first two parent groups. So while removal of one regulatory protein was fatal, removal of two prompted a surrogate regulatory mechanism to take over the same basic function. Again, what at once appeared to be IC proved with time and experimentation to be in fact reducible.

“Reducing” the Krebs and the clotting factor does not free evolutionists from having to show how these mechanisms plausibly evolved. But it shows that, if IC ever rules out such a Darwinist explanation ipso facto, it does not apply in these cases. It confirms our suspicions that biochemical IC may be impossible to diagnose; which should shake Behe’s confidence in reading ID off of other structures similarly tagged.

(II) Showing How What is/Appears IC Could Nonetheless Have Evolved From an Earlier Form

Another possible way of responding to Behe’s irreducibility argument is suggested by considering McDonald’s series “in reverse”. That is, instead of descending from the standard mousetrap to the simple spring, we can imagine McDonald’s series as progressing from the simpler to more complex devices, gaining fitness with each step. McDonald himself has presented one such reverse series (). Where Behe’s original trap represented (if misleadingly) a single organism, the reverse series intends to represent the developmental pathway which the form of this final organism “took,” or would (could) take if it “evolved” as biological creatures do. McDonald’s basic point is to show how a ‘Darwinesque’ accretive developmental story can be conceived of a structure which appears to be IC.

Of course, despite its imaginative element, this development is supposed to say something about actual biological structures, and the parameters of this comparison will be dealt with in the next section. For now, McDonald’s reverse series suggests that Behe’s argument could be deflated by telling a similar story for actual biochemical structures. (a) Note that this approach does not require proffering The True Story of how a given structure actually came to be. Since Behe’s charge is that his chosen structures could not have possibly evolved, we merely need to show how it might have evolved to rebut this claim. (Of course, whatever story we tell must be consonant with, constrained by, our knowledge of evolution; showing how a structure could have evolved just means “showing how it could, given what (else) we know, have evolved.”) (b) Clearly, this kind of response does not even require one to deny that the final organism is IC. Indeed, a developmental story can be told of biological structures that are widely thought to be IC and which have never been functionally “reduced,” conceptually or actually.

The development of the mammalian ear from reptilian jaw bones is one of many examples. Though macroscopic, the ear is helpful for our purposes because its evolution is particularly well-documented in the fossil record, and therefore Behe almost certainly accepts the standard narrative. The three smallest bones in the human body—malleus, incus and stapes—carry sound vibrations across the middle ear from the eardrum to the “oval window.” Removal or modification of any [part of] of them would disable this function and result in loss of hearing; in this sense, the system is IC. The fossil record documents their development precisely, and very nearly as completely as one could ask: During the evolution of mammals, the bones originally forming the rear part of the reptilian jaw were pushed further back and reduced in size until they migrated into the middle ear of mammals. (The process got going in part for a fact the reader may know, that snakes are not completely deaf but can hear to a rudimentary degree by sensing vibrations in their jaw bones.) This case is also appropriate because for decades creationists pointed to the ear as a system that could not have evolved by this Darwinian process and therefore must be the product of ID. These critics mockingly noted that the postulated “intermediate” reptiles could neither hear nor eat while waiting for two of their jaw-bones to migrate into their middle ears. However, in a dramatic moment in the history of evolutionary biology, Harvard’s A.W. Crompton produced a fossil with a clear coupling of the jaw joint, permitting it to both eat and hear during the shift and “freeing” natural selection to select for the intermediate forms.

Recalling Krebs, this way of responding to Behe can be made in biochemical terms as well. The work of Russell Doolittle shows plainly how the blood clotting cascade could—and, his peers tend to concede, probably did—evolve by a process of gene duplications of serine proteases which functioned originally as digestive enzymes (Creation/Evolution vol. 16: pp 55-70). Short of replicating his work here in all its detail, let us note that—unlike Behe’s “explanation” for the same structure—Doolittle’s theory makes testable predictions. Evolutionary theory in general claims that the DNA sequences of organisms reflect and correspond to their specific evolutionary histories, such that that organisms with shared ancestry will display corresponding sequence portions. The relevance of this claim is not diminished when it is an organism’s microscopic features which are under consideration. Therefore, the developmental path which Doolittle theorizes has produced blood clotting animals should have left its unique “footprint” in the DNA sequences of these creatures. This framework permits a host of predictions to be generated, inviting several of Doolittle’s peers to join the investigation and vindicating his work across the board.

These examples prove that IC can in at least some cases come about as an “emergent property.” A “part,” such as a second articulation of the jaw bone or ‘extra’ proteases, which is optional at one point, may, with the addition or modification of other parts, become indispensable at a later stage of development—a phenomenon Behe officially regards as impossible. Because of this phenomenon, the property of IC can never by itself rule out the possibility of evolutionary “emergence.”

ID Defense of the Mousetrap Analogy:
Behe’s, Dembski’s Critique of Disanalogy

Behe and Dembski have sought to defend the mousetrap analogy by problematizing McDonald’s simplified version, considered as a reverse progression. Most of this counter-criticism amounts to charges that the reverse “progression of mousetraps has little connection to biological reality” (Dembski) specifically, it does not present a good model of “a Darwinian process” (Behe). For one, even the simplest mousetrap requires “the involvement of human intelligence…[and the] involvement of intelligence at any point in a [Darwinian] scenario is fatal” for Darwinism (Behe).

These criticisms are misplaced. If McDonald’s progression represents a biological phenomenon, it represents only certain aspects of this phenomenon while abstracting away from certain other aspects—just as, being “merely” an analogy (and not an instance) of this phenomenon, it must. Of course, those respects in which it is disanalogous to the evolutionary process are precisely those it carries over from Behe’s original model—making the latter’s critique a puzzling one: If Behe argues that McDonald contradicts his (own) thesis of non-design in representing biochemical process with an intelligent contrivance, Behe must, to be consistent, admit that he himself begged the question of ID by representing the same reality with an intelligent contrivance in the first place! Recall also that Behe introduced the Scrabble-letter series NDEIRUABFDMOJHRINKE as an illustration of a chance or random construction. This he contrasted with the METHINKS… series of letters, illustrating a product of intelligent design. However, it is clear that both illustrations were created by Behe himself and therefore are both actually products of design. The fact that the gibberish was actually the product of design did not prohibit Behe from using it to illustrate something about, or, some aspect of, a chance (or, non-designed) construction. Nor should the mousetrap’s being a product of design prohibit it from representing something about a product of “chance” mutation-and-natural selection.

McDonald’s (reverse) series, at best, intends to illustrate: (Some features of) How some biological system could have gradually developed from a simpler and less fit to more complex and fitter form, via intermediate steps. Whether intelligent design or the “blind” forces of mutation and natural-selection caused this development is an altogether separate matter from the one McDonald is—qua the illustration—concerned with. (Though, of course, he very much concerns himself with that question at other points). That is, the cause of the natural process the mousetraps represent is among those “respects” of reality of which the mousetrap series is not intended to be representative; it is among those features from which the analogy abstracts.6 Such criticisms targeting the cause of McDonald’s analogy, then, consider his mousetraps only in their bare artifactuality, and therefore are beside the question of that which they are intended to represent.

Dembski’s Critique of Modification

Recall Behe’s challenge to produce a functional mousetrap which is simpler by at least one part. (To avoid confusion, note that while McDonald’s response took the form of a series of traps, this series was prior to his “reverse progression” of the same; these are separate illustrations in the service of separate (if related) points.) Dembski critiques McDonald’s original, reduced series of traps by arguing that, each time he reduces the standard trap by a part, he must modify some of the remaining parts to retain functionality of the machine.

One is justifiably tempted to dismiss this critique by simply pointing out that Behe’s original challenge to reduce the trap never included the stipulation that other parts could not be modified. However, it proves more fruitful to explore the assumptions behind Dembski’s charge, and to critique those specifically. First, this enables us to draw Dembski’s case better, or at least more explicitly, than he has, so that we sidestep blowing our energies on a strawman. More importantly, Dembski’s objections to McDonald’s more elaborate “reversal” are located in his objections to McDonald’s reduced series, in a way which he does not make entirely clear but which can be traced on the same assumptive level.

Ultimately, I wish to argue that, in presenting his “reverse series,” McDonald has once more followed the terms and spirit of Behe’s original analogy to undermine the same; and that Dembski, who appreciates the same original analogy, perversely uses its built-in disanalogies—and every analogy has them—to reality to criticize, not the original analogy, but the implications McDonald fairly and consistently draws out of it.

(The following is wordy; please bear with me.) Again, Behe introduced the trap as an illustration or analogy of the concept of IC as manifest in a biological system. And again, it is this fact which makes it a perfectly reasonable response for McDonald to reverse the process and [attempt to] build up a mousetrap, part-by-part, as an illustration or analogy to the evolution of an (or, of the same) IC biological system. That is, McDonald is hardly transgressing the ‘boundaries’ or ‘spirit’ of the original analogy in doing this. In fact, Behe’s challenge is cast in much the same terms as those of McDonald’s “reversal,” even suggesting it: When Behe argues (recall) that you can’t start with a platform and catch a few mice, add a spring and catch a few more, add another one of the 5 parts and catch even more, etc., his point can only be that: This part-by-part accumulation is in fact (or, were it possible and actual, would be) analogous to a Darwinian evolutionary process—at least, if it could increase fitness along the way. (Call acceptance (or use) of this analogy—simply, seeing that such an increasingly fit mousetrap is like an increasingly fit organism—“Assumption P.”) The difference in his and McDonald’s take on the mousetrap analogy is only that, while (a) Behe accepts assumption P and then proceeds to show that, the way in which the mousetrap could not in fact ‘benefit’ (increase fitness) from such a part-by-part accumulation is analogous to, or illustrates, the way in which a biochemical process could neither ‘benefit’ from an analogous part-by-part accumulation; (b) McDonald also adopts P, but proceeds to show that the way in which a mousetrap can indeed ‘benefit’ from part-by-part accumulation is analogous to the way in which a biochemical process could ‘benefit’ from an analogous process.

Of course, McDonald has to modify the remaining parts in order to show this, and this feature of his approach bothers Dembski. We are now in a position to see how the latter’s critique of ‘modification’ connects to another: He charges that the “reverse engineering” of the mousetrap, proceeding as it does by the addition of parts, employs developmental “leaps” which are markedly un-Darwinian. That is, increasing the fitness of the traps by adding whole new parts (considering modification to others, or not) is disanalogous to the slow, gradual accumulation of parts described by Darwinism (and therefore that any “argument by analogy” to Darwinism fails along with the failed analogy). Therefore, Dembski appears to have locked McDonald in a no-win situation: On the one hand, McDonald cannot traverse the developmental “gaps” in a way that “looks” Darwinian, short of modifying parts. But if the 5 parts are indeed taken as the analogues of real Darwinian steps in the illustration, then introducing any modifications “in between” the stipulated steps seems like going beyond the permitted confines of the analogy—a kind of “cheating.” This (unstated) feeling may be driving Dembski’s objection.

Of course, it is precisely the original analogy that constrains McDonald to accept the 5 mousetrap parts as corresponding to the “steps” in the “evolution” of the trap. I contend that the disanalogies Dembski wishes to exploit are once more simply carried over from the original into McDonald’s appropriation of it. In fact, looking, with Behe, at the mousetrap as a model of a biological system, each step of McDonald’s scheme is better viewed as the sum of multiple smaller steps. Though McDonald’s progression shows only 4 steps of the path from 1- to 5-part construction, an ideal rendering of the process could show any number of intermediaries. These intermediate steps can themselves be seen as “modifications” of the 5 “standard” parts. Moreover, if the mousetrap were to develop like a biological organism or system, one would expect that the 5 “milestone” parts (or the first 4, if we stop the development at the 5th) might change their shapes. As shown by the mammalian ear, a biological system can modify its “parts” at any step along the evolutionary process.

These considerations have prompted McDonald to update his progression in the form of an “gradually developing mousetrap” (). This new (online) version is animated, permitting the author to render the intermediate steps, from a bent piece of wire to the 5-part standard, as a kind of gradual growth. No longer limited by 5 simple steps, McDonald illustrates the development of an (apparently) IC structure in a way which is more like (more analogous to) the real evolutionary process. It illustrates how modification of “parts” is acceptable and indeed essential to such a process, and therefore appropriate to its representation. (Don Stoner, a creationist critic of Behe’s mousetrap analogy, has also developed a non-illustrated version which gradualizes the process by describing intermediary steps ().)

All of this shows the trouble with posing Behe’s challenge, as Dembski does, in these terms: “How does a mousetrap get from mousetrap-version-1 to mousetrap-version-5, by route of versions 2, 3 and 4, in a way which resembles the gradualist progression described by Darwinism?” The main problem is that the 5 steps are ‘already’ disanalogous to those of a Darwinian process, and limiting the analogy to these leaves no room for a developmental story which “looks” Darwinian. But amending the process to one more resembling the gradualist one of Darwinism shows that such a development entails modification. Indeed, it amounts to a series of modifications; in a gradualist model, isolating specific “parts” which “undergo” modification becomes somewhat arbitrary (hence our scarequotes around “parts” throughout).

(III) Showing How the Components of What is/Appears IC Could Nonetheless Have Evolved, Separately, With Separate Functions

Posing Behe’s challenge in terms of this question also assumes that biochemical systems only evolve in a linear, “branchless” fashion, proceeding from a rudimentary version of the system having some function, through various changes to a more developed one having precisely the same function. (Part of this idea is captured in the model of children’s tinkertoys, where some functional device is envisioned, and more and more parts are stuck together toward this single end.) Written out of this scenario is the possibility that any of the 5 parts might have evolved to have other, separate functions, prior to their enlistment in a broader mousetrapping function. Nor are we overextending Behe’s analogy to wring out positions its author fails to actually hold. In his words, “[I]n order for biochemical systems to evolve, multiple systems”—which for Behe, can only correspond to a multiplicity of functions—“would have to arise simultaneously—which can’t be done” (26).

Consider an example from a biology professor at Brown University:

I removed two parts from a mousetrap (leaving just the base, spring, and hammer), and used that 3-part device as a functional tie-clip. I then detached the spring from the hammer, and used the device as a keychain. If I had cared to, I might have used the base and spring (2 parts) as a paper clip, my tie clip (glued to a door) as a door knocker, the catch as a toothpick, or the base as a paperweight.

We can agree with Behe that natural selection cannot favor a non-functional system. But as these examples show, to say as he does that some function is abolished by removing one part of an IC system does not mean that function per se is abolished. Clearly, portions (“parts”) of an IC system can be perfectly, but alternately, functional in other contexts. Their parts can work—differently—apart from one another. This suggests yet another way of responding to Behe’s irreducibility argument: One could show how an IC system might have had precursors which were selected to fill other functions in the past. This would debunk Behe’s claim that an IC system had to be fully assembled at once because natural selection would have no “incentive” to favor development of its individual parts. Nor is this kind of scenario without its real biological analogue. Arguably the best example is from biochemistry itself, and is among Behe’s favored IC’s. Behe would contradict most of his disciplinary peers to deny that the bacterial cilium could have evolved though an assemblage of proteins originally used for other purposes in the cell. Crucial portions of the eukaryotic cilium, such as the linking proteins tubulin and dynein, and more complex microtubule and cross-bridging components, all retain other functions within the same cell. (Also, the known relatives of ciliated organisms share “protocilia” organelles.) These functions offer all the necessary “incentive” for evolutionary forces to select for these portions. The mechanism of gene duplication, which Behe accepts as holding in other contexts, offers a plausible means for the development of the cilium from the same parts. Thanks to the same device, the cilium itself has been demonstrably adapted in turn for another function, as a sub-structure of the vertebrate photoreceptor cell. (Ironically, the photoreceptor cell is yet another structure Behe cites as evidence of design!)

On Irreducibility and the Character of the Intelligent Designer

Investigating irreducibility suggests another issue closely associated with the concept of ID. Clearly, Behe intends the word “intelligent” in a purely qualitative sense. That is, he notes that ID is neutral with respect to the specific character of the designing agency for which it argues. To this end “Intelligent Design” is actually redundant; even a relatively dumb designer would count as “intelligent” on this definition. However: To say that the arguments in DBB do not entail any conception of designer (beyond possessing a bare capacity to design) is not to say that the question has no place in a proper treatment of the work.

Let me try to make the case, as the point is subtle. First, it is quite possible to distinguish between (a) a theorist’s views expressed at one time, or in one forum, and (b) those expressed at another; and these views can be perfectly compatible without any relation of entailment holding between them in either direction. Second, it is possible to distinguish between (a) a theorist’s intentions regarding the interpretation of his own views, and (b) the uses to which this work will likely be put. (A naive loyalty to ‘author’s intentions’ has no more role to play in philosophical than in literary criticism.) In both scenarios (b) can be interesting regardless of (a), while it remains sensible from a stylistic viewpoint that both might be targeted together.

I find these considerations relevant to my treatment of Behe’s work in view of the following:

(1) There is little doubt among Behe’s serious critics that the One, supernatural, God of Judeo-Christianity is Behe’s personal candidate for cosmic intelligence. This is due to more than speculation. For instance, Behe takes the scientific community to task for dismissing his ID theory merely because “the other side [of the “elephant” labeled intelligent design] might be labeled God” (DBB 233). This charge has been repeated by Behe in numerous public fora. The criticism is consistently that his peers would allow themselves to be intellectually compromised by such a fearful conclusion—never that such a conclusion, fearsome or not, is in any way unwarranted.

(2) The theory articulated in DBB is not a thing apart from the broader ID “school.” The most prolific among the members is W. Dembski, whom Behe cites repeatedly and for whose polemic Intelligent Design he penned a preface. In this work Dembski claims that “Christ is never an addendum to a scientific theory but always a completion” (ID 207). Dembski’s anthology Mere Creation remains the first and most representative collective expression of the ID school. The editor’s comment that, “As Christians we know naturalism is false,” appears in the very introduction to the work, proposing that the “we” intends the writers to follow (14). Many of these authors are openly committed to Christian Theism. Behe is not only among these contributors but invoked (glowingly) by virtually all of them. All of this suggests that Behe shares their Theism and that, regardless, his work is being used in its service.

(3) Similarly, the collective work of the ID school is neither a thing apart from the broader Creationist school. A 1996 document of the Creationist Center for the Renewal of Science and Culture credits ID theory as desirable for the Center’s “wedge strategy” of striking Darwinism at its allegedly weakest points. The avowed idea is that establishing ID, or rather, securing its credibility in the public mind, is one step toward promoting the credibility of Theism, which is a step toward popularizing Christianity specifically ().

I trust these considerations justify, at minimum, a kind of “procedural equation” of Behe’s designer with the God of Western Theism. That is, they demonstrate that there is a utility in equating these agencies, for certain purposes. One such purpose is critiquing those who equate them for their purposes. For example, one could critique the avowed creationists in (3) vis-a-vis their equation of Behe’s designer and God the Father. This might require a shift of focus from ours up to now, to (something like) “Behe’s theory as creationists interpret or apply it,” but it is no less still a critique of Behe. Similarly, given (1) and (2), I wish to critique (something like) “Behe’s theory as he, in other moods (for other purposes?7), interprets or applies it”—or, “as he tacitly, and ‘ultimately’, intends it to be interpreted or applied.” And similarly, I wish to do this in such a way that the critique of Behe and of ‘his designer considered apart from the question of his/her character’, made hitherto, is not diminished in relevance.

Of course, in such a case, a theological critique is warranted. (I contend that, if Behe equates the two, then a critique of design is already a critique of God—“designer” would refer to God.) Specifically, it now makes sense to investigate whether design looks Godlike; and if it does not, this becomes a blow to ID theory—to Behe on his own express terms. (In any case, if you don’t buy my ‘procedural equation’, the stronger point will be shown that, whatever the author’s intention, DBB is not altogether neutral as to the designer’s character; certain traits just fall out of the concept of ID itself.)

Minding all of this, investigating Behe’s appeal to irreducibility suggests two basic lines of critique:

Irreducibility as a Lack of Troubleshooting Mechanisms

Irreducible complexity can be a mark of relatively poor design. In the world of human artifice, there is often enough occasion to criticize a particular design as fragile, lacking reinforcement of some kind. Such a charge does not amount to saying that the design is dysfunctional per se, but rather that it is (excessively) vulnerable to becoming so in the event that something goes wrong. Granted that no (human, imperfect) designer can be expected to anticipate, or to mechanize against, all conceivable mishaps; and too much reinforcement against one contingency may create vulnerabilities of another sort or indeed bring problems of its own (a ships-hull could be so glacier-proof thick that it admits no cargo, or sinks in the harbor). But the difficulty of devising a general rule for adjudicating these cost-benefit tradeoffs does not render it impossible to conceive of specific systems that are deficient—or less “intelligent” in design—insofar as they lack troubleshooting or “backup” mechanisms. Consider an automobile that lacks a spare tire, rendering the entire vehicle dysfunctional whenever one blows. It is in just this sense that truly “irreducibly complex” designs “live on the edge” of functionality; when something goes wrong, a lack of compensatory features hobbles them. The point is that this irreducibility may sooner bespeak the work of a shortsighted or otherwise inefficient designer than the perfect one of Western Theism.

Biochemical Complexity as Redundant, not Irreducible

It may be countered that biochemical mechanisms do contain such compensatory mechanisms, which are activated when certain components fail but remain “dormant” otherwise. This fact is incontestable. But surely it amounts to saying that these machines could have been “designed” without such mechanisms and still, under normal circumstances, function—and therefore wars against the thesis that they are IC.

Along these lines, Behe frequently portrays irreducibility as a system’s inability to function with one of its parts missing. At least, this is the impression one gets from his illustrations of this feature. For example, Behe presents the case for the irreducibility of his mousetrap by cycling through the various problems that (would) result from removing each of its parts in turn, while leaving the others intact. (Recall: “[I]f there were no platform, the other pieces would fall apart…[I]f there were no spring”—that is, if no spring were attached to the platform, which has ‘returned’ since the first example—“the mouse would not be pinned to the base” (MC, 178).)

But this approach may indicate a lack of imagination on Behe’s part. Recall Bugge’s gene deletion experiment, which we used to demonstrate that the blood clotting factor in mice was not IC, but instead functionally reducible. In their account of this study, Shanks and Joplin go a step further, recognizing that a better description of such a system is “redundant complexity” (Niall Shanks and Karl H. Joplin, Redundant Complexity: A Critical Analysis of Intelligent Design in Biochemistry. Philosophy of Science, 66 (2) (June 1999): 268). This redundancy renders the “design” of the clotting factor vulnerable for its excesses rather than, recalling the previous section, for its deficiencies. It is as though a designer took a perfectly functional system and added two extra parts, such that removing either one of them causes serious problems to the system. In this case alike, then, any designer we postulate is open to charges of poor planning or execution, and consequently appears far from the perfect God of Judeo-Christianity.

Means of Design

These considerations show the possibility and fruitfulness of questioning the concrete character of design beyond its mere ‘inferability.’ Indeed, if scientific hypotheses are to be judged in part by their suggestiveness, or ability to generate interesting lines of research, Behe’s hope to direct serious scientific attention to the design inference, it would seem, depends on the coherence of just such kinds of investigation. Of course, it also shows that the answers to these questions may trouble the theory.

Along these lines, we might ask just how, and at what point in the course of natural history, Behe’s biochemical designer went to work. A “deistic”-type theory whereby some agent set the initial, physical “ground rules” for the universe—like gravity, the behavior of fundamental particles, etc.—may be coherent but fails to address any of the questions Behe is concerned with. The same can be said of a designer who came along after the world’s beginning, jump started life in its most rudimentary form (Dembski’s ‘LIFE’) and then stepped back. Clearly, Behe is concerned with biological processes which emerged at some later point—much later, and indeed, at multiple separate points—along the process of development of life on earth. Remember that, assuming design took place, its direct action would occur at the genetic level; the biochemical structures themselves are crafted only indirectly, by way of the genes that code for them.8 Therefore, either of two scenarios must hold: (1) The designer reaches in and adds newly-crafted genes to the genome at various points, as environmental, or other phenotypic, developments dictate—not unlike a software provider provides updates as consumer needs expand; or (2) all of the design work for these yet unmanifested processes were encrypted in the initial code of the first primitive cell; here, the periodic intervention of the designer merely engages or activatesgenes crafted long ago. This latter is Behe’s own view:

Suppose that nearly four billion years ago the designer made the first cell, already containing all of the irreducibly complex biochemical systems discussed here and many others. (One can postulate that the designs for systems that were to be used later, such as blood clotting, were present but not ‘turned on.’ In present-day organisms plenty of genes are turned off for a while, sometimes for generations, to be turned on at a later time). (11)

There are two serious problems with this scenario; it is incredible that Behe never considers them. First, the length of DNA code for a bacterium is a mere fraction of that of a modern multi-cellular organism. There are, in fact, no reams of “turned off” genes in the contemporary bacterium corresponding to the great variety of structures Behe wishes to account for. (Their presence, of course, being what we would expect if Behe’s theory were adopted, given the part of evolutionary theory he himself accepts.)

Second, “for generations” is a grand understatement of how long the postulated genes would have to lie dormant before being expressed. Those coding for the eukaryotic cilium would have to pass for hundreds of millions of years before being engaged; those for blood clotting proteins would pass for over a billion years. During this long period, the dormant genetic strings can be expected to bear mutations at the standard rate. Normally, the mechanism of natural selection would weed out these mistakes as they are manifest in unfit phenotypes from regular generation to generation. However, in “cold storage” natural selection cannot touch them, so errors will accumulate one on top of another. This means that by the time the designer deems conditions proper for “turning them on,” these genes will be altered far beyond the design as it was originally conceived.

It is unimaginable for all or most of these mutations to be benign with respect to expression. More importantly, it is unlikely that environmental conditions should change so dramatically and in just the right ways from the parent generation to the birth of the “expressing” organism that the latter could even survive to reproduce within it. This would be correct from a statistical viewpoint even if the genes under consideration coded for macroscopic features. But in fact, if Behe’s thesis of IC were correct, any mutation which expresses a change in a biochemical structure would render it inoperable; and we are talking about a mindboggling number of mutations—an “unnaturally” high number, if you will.


1. By “simple” complexity I mean only complexity considered apart from the question of irreducibility.

2. Granted, to argue, as Behe correctly does, that we recognize the phrase in whole or as a whole becomes tricky, as we may not have seen this particular phrase before. (Neither may we recognize it as a quotation from Shakespeare.) Nonetheless, we do recognize it, and we wish to say that this business of recognizing amounts to something more than just seeing it as sensible—as some phrase—or even as a phrase in English. Behe probably would locate our recognition of the phrase as such in the fact that we clearly recognize its word-parts, combined with (our recognition of) the grammatical framework which unites them in the particular way it does.

3. Actually, Dembski’s criteria are probably wrong here. For one, they are too permissive, failing to restrict, for example, “humanoid” natural sculptures (like New Hampshire’s Old Man of the Mountain—R.I.P.—or the Sulukhat woman of the Caucasus), which arguably combine improbability with a recognizable pattern, from generating a design inference. Second: The nonsense Scrabble sequence and its English counterpart are arguably both “specified,” though to differing degrees. Granted, this is because they are both a product of design, though again to differing degrees: In both cases some intentional agent had to set up the table, to spill and spread out the letters, flip over any face-down ones and arrange them along a straight line, rotating letters as needed to read the correct way; clearly, for this agent to have gone a step further and spelled out a recognizable phrase on the line would not have crossed a line from nondesign to design, but only have increased the degree to which the scenario is, was already, a case of design. So Dembski must acknowledge that these concepts are matters of degree—and therefore, that something more must be added to his criterion that designed entities are “specified” plain and simple. Needless to say, both the sensible and nonsensical Scrabble scenarios are highly improbable, so the second criterion of improbability won’t do it. What is it that demarcates improbable/specified/less-designed entities from improbable/specified/more-designed entities such that the ID inference is fair in one case and not the other? Anyhow, these deficiencies, while provocative, do not affect our having accepted Dembski’s criterion for the sake of the present argument.

4. Recalling footnote , it may be the case that a biochemist’s cognizing such a structure—discovering a brand-new one, we’ll say—amounts to some kind of ‘recognition’. But rather than being a counterexample, this is at best analogous to a linguist’s coming across a phrase in an unknown language. The scientist may recognize his discovery as a biochemical structure, just as the linguist might recognize the phrase as ‘language’, or a language pattern, or even a type of language pattern (a phrase, statement)—but neither in terms of its specified character. The linguist can recognize that some linguistic structure is specified while failing to recognize it in terms of this specificity, to recognize it in the sense of the METHINKS… pattern; and the fact that a biochemist can recognize some structure as a distinct biochemical structure does not depend on the structure being specified.

5. Of course, Behe’s macroscopic and microscopic ‘domains’ are categorically different for a great many common human purposes. The fact that we can see one, and not the other, that one is composed of (parts of) the other, etc., ensure that this will always be the case. I suspect these ready differences are behind Behe’s quickness to argue that, because we only know how the first evolved, the second must not have evolved. But to cite even very great differences between things does not mean that they are different for every purpose, or for present purposes.

6. In like manner, to a Christian’s saying that God is analogous to her human father, it is no criticism to say that, while her father has blue eyes, God does not. Eye color (or, having fleshy eyes in the first place) is not one of the respects in which the analogy was intended to hold. Actually: It is more like Behe himself has said God is like his own father, to which McDonald has countered, it is more like his mother, to which Behe responds, “That analogy is bogus because my Mother has blue eyes, and how can God have blue eyes, or any eyes at all?.” (I do feel that Behe’s “disanalogy critique” is particularly sloppy and reaching.)

7. That is, purposes other than those explicit in DBB.

8. Weirdly, Dembski tends to forget this, writing as though the designer must craft the micromechanisms directly. Possibly he means to say this. But clearly, it is not the theoretic “default option” for ID: Genetic-level crafting is the obvious analogue to the evolutionary ‘crafting’ of macroscopic structures; that is, one would expect that, all things being equal, the designer would simply alter the genes him-/herself, just as mutations alter them on an evolutionary story. Dembski does not argue for his assumption that the designer crafted the structures directly, as opposed to genetically, nor deals with the problem (remember Lamarck?) of how these phenotypic changes could be translated back into the genotype to be passed on. His failure to act as though that any argument is needed, and to anticipate the objections that those versed in genetic theory could be expected to produce, indicates a lack of thinking on his part rather than a deliberate, thought-out point.

2 responses to “A Critique of Behe, Dembski on “Irreducible Complexity”

  1. Hola! The Darwinism v. Intelligent Design issue is far more complex than either Dembski et al’s angle, or yours. For the most concise, balance, and nuanced perspective on the age old question of naturalism v. creationism, see: ‘Debating Design: From Darwin to DNA’, eds Dembski & Ruse.

    • It is more complex. I was responding to someone’s angle, so my response is equally “angular” (partial, narrow). Thanks for the reference; when my immediate interests swing back around to this issue, I’ll remember it.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s