Reports of the National Center for Science Education

Of Mousetraps and Men: Behe on Biochemistry

In Darwin's Black Box: The Biochemical Challenge to Evolution, biochemist Michael Behe claims that biochemical systems exhibit a special kind of complexity - irreducible complexity - that cannot possibly have evolved and must have resulted from intelligent design. Like other intelligent-design creationists, Behe is vague about both the identity and methods of his intelligent designer, though he does distinguish between the hypothesis of natural design (by space aliens, perhaps) and that of supernatural design (1996, 248-9).

As Behe is aware, postulating intelligent design by space aliens only postpones a confrontation with the problem of the origins of complexity. After all, who designed the designers? Thus the unwary reader is pointed in the direction of a supernatural, undesigned designer. But if you were puzzled by biochemical complexity in the first place, this latter hypothesis, involving as it does an unknown supernatural being that employs unknown materials and methods, can hardly result in a net reduction of mystification.

Luckily we do not have to settle this matter. It turns out that Behe's intelligent design hypothesis is the result of his failure to consider relevant natural processes when trying to account for the origins of biochemical complexity. This problem arises in turn because Behe thinks about biochemical complexity with the aid of a misleading mechanical analogy - the well-designed mousetrap. The mechanical mousetrap is to Michael Behe what the mechanical watch was to William Paley. And it goes without saying that machines have designers.

So how should we think about design and designers? We will argue first that the historical process of the intelligent human design of technological artifacts, such as mousetraps, needs to be sharply differentiated from the hypothetical magical process of supernatural design and creation ex nihilo (literally from nothing). In fact, Behe's case derives its appeal from a failure to examine the details of the human design process. Naturally, he provides no details whatsoever of the hypothetical supernatural design process. Secondly, we will show why the mousetrap analogy fails to do justice to the richness of biochemical complexity. And thirdly, we will offer a conceptual framework that explains the origins of the irreducible complexity Behe finds so mysterious (see also Behe 2000). The key, as we shall argue, is that most real biochemical systems exhibit a type of complexity that we term redundant complexity: a form of complexity that results from natural evolutionary processes amenable to scientific study.

The Mousetrap Model of Biochemical Complexity

Behe's central thesis is that the biochemical systems we find in living organisms manifest irreducible complexity. He further contends that processes of the kind invoked in evolutionary biology cannot explain the origin of irreducibly complex biochemical systems. Behe explains:

By irreducibly complex I mean a single system composed of several well-matched, interacting parts that contribute to the basic function, wherein the removal of any one of the parts causes the system to effectively cease functioning. An irreducibly complex system cannot be produced directly (that is, by continuously improving the initial function, which continues to work by the same mechanism) by slight, successive modifications of a precursor system, because any precursor to an irreducibly complex system that is missing a part is by definition nonfunctional (1996: 39).

Behe contends that although intelligent design processes of the kind we find in engineering, for example, can give rise to irreducibly complex systems, evolutionary processes cannot.

Behe employs an analogy with well-designed mousetraps. A mousetrap has several components, all of which are necessary for catching mice. A precursor "trap" that lacked one of the components - the spring, the trigger, or the platform, perhaps - could not trap mice. Lacking even minimal function, it could not be improved through incremental adaptive evolution to become a functioning trap. We already know that mousetraps require intelligent human designers. Behe argues that functioning biochemical systems are like mousetraps. They could not have evolved through incremental adaptive evolution, and must be the products of super-human intelligent design. This argument, like all design arguments, has a surface plausibility. It is too bad that those who rely on design arguments have never taken the time to think clearly about what is actually involved in the intelligent human design of technological artifacts.

The Origin of Artifacts

It is essential to differentiate between the actual behavior of intelligent human designers on the one hand and the hypothetical action of a supernatural being on the other hand. Human engineers do not create or manufacture anything ex nihilo, nor are the processes of design secular miracles; to suggest otherwise is misleading. For ordinary objects like mousetraps, we can infer that they are designed without having any knowledge of their designer. Why? Because we all know that mousetraps are artifacts, and by definition artifacts have designers.

But, by studying the history of artifacts and engaging in reverse engineering, we can often uncover the identities of actual human designers, as well as the methods and materials they employed. That is why it is important to study the processes of human design, and not just the design itself. Can we analogously study the design process of the hypothetical supernatural designer? No proponent of supernatural design - not St. Thomas Aquinas, not Archdeacon Paley, not Michael Behe and his numerous sympathizers - has ever offered the slightest clue about how this could be done scientifically.

Intelligent design theorists, noting that various biochemical systems give the appearance of being designed, aim to argue that they are, like artifacts, actually the products of design. However, if they are not like artifacts - if the appearance of design is deceiving - then all bets are off. Although a biochemical system might give the appearance of being designed, the conclusion that the system actually results from design could be made evidentially respectable if, for example, we can find some compelling scientific evidence concerning the methods and materials employed, and also some compelling historical evidence concerning the identity of the designer.

Why should this be so? The reason lies in a fundamental difference between things such as mousetraps or watches, on the one hand, and things such as biochemical systems, on the other. Mousetraps and watches are antecedently known to be artifacts, and hence to have human designers (even if their identities and methods are obscure). For these objects, the question of design can often be safely separated from the questions of how designed and by whom.

But in the case of the alleged intelligent design of biochemical systems, these questions are all inextricably intertwined precisely because it is not known antecedently that biochemical systems result from deliberate design by a non-human agent (or agents) of supernatural origin. The very claim of design itself requires evidential justification. Providing evidentially grounded answers to the questions of how and by whom these systems were designed would simultaneously provide powerful evidence that the systems were indeed designed - a matter sorely in need of justification. This issue is made all the more acute because naturalistic evolutionary hypotheses exist to explain the same features that lead Behe to postulate a designer; one is presented below. For this reason, the features of biochemical systems that Behe points to cannot simply be viewed as the registered trademark of the creator or the hallmark of design. Behe has made an extraordinary claim, and its validation will require extraordinary evidence. Behe makes no attempt to meet this evidential requirement.

Moreover, in his argument Behe cavalierly ignores common facts about the human design process, which, like biological evolution itself, involves descent with modification. The intelligent human design of artifacts is frequently a historical process resulting from the generation of variation on existing technological themes along with selective retention of specific variants for further elaboration. Human engineers have long known that the problem-solving process is a historical, tinkering, trial-and-error process.

Indeed, it is the ability to produce multiple variants on themes, by varying parameters, that makes modeling and simulation such a powerful tool in the design of technological artifacts such as aircraft (Vincenti 1990). Even at the dawn of powered flight, the Wright brothers built, tested, and discarded numerous models in wind-tunnel tests. Many variants based on existing glider designs were tried, but only a few were chosen and selected for further elaboration. The ultimate fruit of this trial-and-error process was a powered machine adapted to an aerial niche! It was not magically created by some human intellectual whirlwind using pieces of junkyard scrap.

The concept of supernatural intelligent design derives at least some of its appeal from the fact that we humans have actual experience of the intelligent design of artifacts. But when what is involved in human design is properly understood, do we really want to understand the hypothetical supernatural design process by analogy with the bungling, tinkering, trial-and-error process of our own experience? And if the supernatural process is different from the human process, how is it different? And how could we settle disputes between rival hypotheses about the details of the supernatural design process? Until these issues are addressed, biochemistry's mysteries will not be solved through the invocation of supernatural design, because until they are dealt with, the appeal to supernatural design will be effectively no different from the claim that it all happened by magic. If ever there was an explanatory black box, this is it!

Perhaps our point is now clear. But if it is not, let us elaborate using some examples given by Gary Cziko:

In 1793, Eli Whitney's cotton gin that removed seeds from short-stapled cotton was based on the Indian charka, which had been in use for thousands of years to remove seeds from long-stapled cotton. Joseph Henry's electric motor of 1831 copied many of the mechanisms involved in the steam engine. The development of the first transistor at Bell Laboratories in 1947 ... owed much to the work of German physicist Ferdinand Braun who, in the 1870s, found that certain crystals conduct electricity in only one direction (1995: 163).

What at first glance might appear to be an ahistorical special human creation is really an artifact belonging to an historical lineage, where new artifacts result from the same variation-and-selection processes that are the staple of evolutionary explanations.

To see how this works, consider intelligently designed jet engines. These are clearly a different species of technological artifact from intelligently designed water-wheels. Yet over the last three hundred years we can trace a line of descent from water wheels to water turbines, and then from water turbines to steam turbines, gas turbines, and jet engines, with variation-and-selection processes playing important roles in all the major engineering transitions.

Needless to say, jet engines did not descend in a simple linear fashion from water wheels. Rather these artifacts emerged through processes involving horizontal transfers of modules from other evolving technological lineages (for example, fuel technologies, metallurgical technologies, and so on). The new modules were further modified as they were gradually incorporated through trial-and-error processes leading to the development of symbiotic relationships.

Interestingly, although there are some obvious and important differences between technological and biological evolution - the former is best explained in terms of intelligent (human) design, whereas the latter is not - they do not differ with respect to having benefited from horizontal module exchanges.

Horizontal transfers play an important role in biological evolution. Modules evolving in one lineage can be transferred to other lineages, where they typically undergo further evolutionary modification. For instance, changing the example from jet engines to eukaryotic cells, evolutionary biologists now see the mitochondrial power plants as the fruits of a symbiotic union between at least two distinct prokaryotic lineages - the integrated endosymbiotic whole is greater than the sum of its prokaryotic parts.

Nor should the evolutionary theorist ignore horizontal plasmid exchanges by means of which genetic information in one bacterial lineage can find its way into another, distinct, lineage. On a larger scale, there is, of course, horizontal exchange through hybridization. In this process, of great importance in plant evolution, first-generation hybrids show a genuine mixture of characteristics from the distinct parental lineages.

And what of the origin of mousetraps? Variation-and-selection processes have played an important role here too. Since the US Patent Office opened in 1838, it has granted more than 4400 mousetrap patents. Currently, about 40 new mousetrap patents are issued each year. Ten times that many patents are turned away, mostly because they are not minimally functional. The Patent Office mousetrap taxonomy recognizes 39 subclasses, including "Impalers", "Smiters", "Swinging Strikers", "Choking or Squeezing", "Constricting Noose", and "Electrocuting and Explosive" (Hope 1990: 92)

Devices that kill mice by hitting them have a long and interesting technological evolutionary history - see Hornell (1940). The spring-loaded trap discussed by Behe appeared in the 1890s, and was patented in 1903 (nr. 744379) by John Mast, a Pennsylvania coleslaw manufacturer with a serious rodent problem. The spring-loaded trap did not result from design and creation from nothing - a secular miracle in Pennsylvania. Rather, Mast had studied existing mousetrap patents and had borrowed from 5 or 6 of them - thus showing the importance of horizontal information transfers - before filing his own patent application in October 1899 (Hope 1990: 94). Behe's mousetrap is in fact a technological hybrid, descended with modification from earlier traps in a complex historical evolutionary process. Although the mousetrap is intelligently designed, it did not appear by a magical, ahistorical process of special creation, the details of which are forever hidden from public view!

Biochemical Complexity and the Mousetrap Analogy

But what of the mousetrap analogy of biochemical complexity? Here we will present some examples from biochemistry that call into question the general biochemical relevance of the mousetrap analogy. We will argue first that Behe's mousetrap analogy leads him to ignore a crucial aspect of the biochemical complexity we observe in nature: the phenomenon of redundant biochemical complexity. Redundant biochemical complexity represents the biochemical and molecular footprints of evolutionary processes in action. Having explained and illustrated this concept, we will then argue that redundant complexity provides the key for a natural, evolutionary understanding of the origins of irreducible complexity. We do not pretend to have a complete account of evolutionary biochemistry. We suspect that the details will eventually emerge from continuing scientific research. But rather than speculate about these matters, we will focus instead on what we do know about biochemical systems.

While biochemical complexity has many sources, one of the key concepts underlying our current understanding of biochemical evolution is that of gene duplication, a process whereby a gene is doubled in a genotype. As a result of this process, one gene can continue the old function, while the duplicate is freed up to be co-opted to serve novel functional ends - the duplicate gene acquires mutations that change its activity. These mutations may be preserved or eliminated through the operation of natural selection. If preserved, these mutations can lead to new functions. More importantly for our purposes, gene duplication is also a central evolutionary source of some of the redundant complexity we actually observe in biochemical systems (Shanks and Joplin 1999). So what is redundant biochemical complexity?

We see redundant complexity when we notice that many actual biochemical processes do not involve simple linear sequences of reactions, with function destroyed by the absence of a given component in the sequence. Instead, they are the product of a large number of overlapping, slightly different - hence redundant - processes. Redundant complexity is also embodied in the existence of back-up systems, which can take over if a primary system fails. Finally, redundant complexity is observed in the phenomenon of convergent biochemical evolution, wherein systems with different evolutionary histories, perhaps using different mechanisms, nevertheless achieve similar biochemical functions.

Redundant complexity turns out to lie at the heart of the stability that biochemical processes manifest in the face of perturbations that ought to catastrophically disrupt systems like Behe's well-designed, minimalist mousetrap - the absence of any component of which should render the system unable to perform its function. To understand redundant complexity better, it will help to look at some examples.

Redundant Pathway Complexity

If we examine the central catabolic pathway of glycolysis (the interconnected series of reactions by which glucose is broken down to release usable energy), it looks superficially as though the product of one reaction in the series is required as the substrate for the next reaction in the sequence. Thinking of glycolysis on the mousetrap model, one would expect that removing one component - enzyme, substrate, or product - would shut down the pathway and prevent the continual production of energy. In fact, almost every step in this pathway is redundantly complex. As an example, let us look at a key step, the production of glucose-6-phosphate from glucose, catalyzed by the enzyme hexokinase.

Not only does hexokinase activate the relatively stable glucose (Bennett and Steitz 1978), but it is a multipurpose enzyme that in part controls the rate of the first part of the glycolytic pathway by directing the chemistry of glucose either to build up more complex molecules (anabolism) or to harvest the energy stored in glucose (catabolism). The direction of chemical activity is dependent only on the concentration of the substrates, products, and various components of the pathway (Voet and Voet 1995).

One might assume, therefore, that here we have a good example of Behe's irreducible complexity. Remove the enzyme and the reaction should stop. But this intuition rests only on a superficial characterization of this step in the pathway. Looking at the fine details -- where the devil proverbially lurks - reveals an unexpected complexity in what initially appeared to be a simple, straightforward chemical situation.

In typical vertebrate tissue, redundant complexity is manifested in the existence of several different variants (isoforms) of hexokinase. All of these are present, as a result of gene duplication and differential expression, in varying proportions, in different tissues. The proportions of the variants differ for the specialized functions of the different tissues in which they are present, depending on whether the tissue requires rapid utilization of energy (as in muscles) or is involved in converting glucose into the storage form glycogen (as in the liver). Removal of a given variant of hexokinase does not disrupt glycolysis, although it may have an effect on the efficiency with which a function is achieved. So there is redundant complexity here, in the first, seemingly simple and straightforward, step of the glycolytic pathway.

Each of the other components of the rest of the glycolysis pathway manifests similar redundancies. Remove glucose, and the pathway can utilize numerous other hexose (6 carbon atom) sugars to supply the next product. Knock out one enzyme variant, and the other variants in the tissue can take over its function - maybe not quite as efficiently, but as Behe concedes, efficiency can be improved by natural selection over evolutionary time. There are back-up systems too. For example, if all the variants of hexokinase were removed, there are alternative pathways, such as the pentose phosphate pathway, that can supply the needed products (Martini and Ursini 1996).

It is a hallmark of many evolved biochemical systems that there are typically multiple causal routes to a given functional end, and where one route fails, another can take over. The existence of variants of a given enzyme are evolutionary legacies - legacies by means of which one and the same enzyme can be adapted to serve different specialized functions in different specialized tissues.

Genetic Knockout

Another way in which we can see the general inadequacy of the mousetrap analogy of biochemistry is simply to remove specific sections of an organism's genome. This procedure has recently been applied to mammals. Researchers can now target a specific gene in mice and "knock it out" (Travis 1992). Such knockout mice are valuable models for human diseases in gene function experiments. However, such mice do not always give the expected result - they do not exhibit the predicted functional deficits - due to the type of redundant complexity we have been discussing.

One example concerns the gene p53, which was originally identified as a tumor suppression gene, but has subsequently been found to be involved in a number of fundamental cell processes. For example, it plays roles in gene transcription, the cell cycle, programmed cell death (apoptosis), DNA replication, and DNA repair processes (Elledge and Lee 1995).

If you thought of this case as a genetic mousetrap, you might be tempted to think that the removal of this gene, involved as it is in all of these vital processes, would lead to catastrophic collapse of the developmental process - a bit like removing the spring, trigger, or platform from Behe's mousetrap. But this is not the case, since p53 knockouts in mice yield offspring that are viable and fertile, although susceptible to the early appearance of spontaneous tumors (Dowehower and others 1992). This suggests the following dilemma: either p53 is not required for embryonic development or there are redundant ways in which the function of the missing component is compensated for (Elledge and Lee 1995). The evidence at hand supports redundant complexity, since there are at least 400 proteins associated with the proper control of the cell cycle alone (Murray and Hunt 1993), and it would appear that some of these other proteins pick up the slack created by the missing p53. Such mice can still be caught in mousetraps!


We are discovering more and more about the nature and role of redundant complexity. Consider the new field of genomics. The study of genome sequences has revealed some startling findings about the complexity and organization of biological organisms. The genome of the yeast Saccharomyces cerevisiae contains many redundant sequences. Fifty-three duplicated gene clusters, making 30% of the yeast genome, have been identified (Clayton and others 1997). Such findings concerning gene duplication lead to an interesting question concerning Behe's use of the mousetrap analogy. That is, how few genes does it take to maintain a free living organism? Experiments at Celera Genomics are currently underway to knock out all nonessential genes from Mycoplasma pneumoniae.

If these experiments succeed, the resulting minimal organism will be noteworthy as a genuine example of a genetic version of Behe's mousetrap. If the organism really is genetically minimal, the absence of any component will be fatal. And such a minimal organism will be peculiar precisely because it will be a laboratory artifact - a drastic artificial modification of a redundantly complex natural system.

Redundant Origins of Irreducible Complexity

The existence of redundant complexity is evidence of the operation of evolutionary processes at the biochemical level. But it does not show that Behe is wrong to point to the existence of irreducible complexity. Let us suppose that some of his candidate examples of irreducible complexity are correct. Redundant complexity gives us the tools to explain the origins of what Behe found so mysterious. To see how redundant complexity might explain the origins of irreducible complexity, let us borrow an architectural image from A G Cairns-Smith, a biochemist interested in the origins of biochemical complexity (1986: 59-60).

Consider a free-standing arch of stones. It manifests irreducible complexity in that the keystone at the top of the arch is supported by all the other stones in the arch, yet these stones themselves cannot stand without the keystone. In other words, all the component stones depend on each other. Take away any stone, and the arch collapses.

Notwithstanding this fact about arches, it is nevertheless possible to construct them in gradual stages. You cannot gradually build a self-supporting, free-standing arch using only the component stones, piling them up, one at a time. But if you have scaffolding - and a pile of rocks will suffice to support the growing structure - you can build the arch one stone at a time until the keystone is in place and the structure becomes self-supporting. When this occurs, the now redundant scaffolding can be removed to leave the free-standing structure.

The study of developmental processes suggests that an important biological role is played by removable scaffolding in the formation of all manner of elaborate structures, including body parts and neural pathways. For example, developmental scaffolding, in the form of an initial superabundance of cells, can be removed by programmed cell death (apoptosis). This process plays a crucial role in the developmental sculpting of such structures as fingers and toes (Campbell 1996: 980; Lewis 1995: 15).

Natural evolutionary processes give rise to the redundant complexity we observe in biochemical systems. But these redundancies may also provide, in concert with extant functional systems and structures, the biochemical scaffolding to support the gradual evolution of systems that can ultimately manifest irreducible complexity when the scaffolding is reduced or removed. By the operations of natural selection, some of these biochemical arches will be retained for further evolutionary elaboration, while others will be eliminated. In effect, irreducible complexity results from the evolutionary reduction of redundancy in redundantly complex systems - systems that are themselves the fruits of evolutionary processes.

In an earlier paper (Shanks and Joplin 1999), we argued that self-organizing chemical reactions - many of which are suitable for demonstration in the classroom or laboratory - can give rise to irreducibly complex chemical systems. Our claim here is that redundant complexity provides another natural evolutionary route to the same end. We are a long way from having to abandon natural science in favor of supernatural hypotheses concerning the origins of biochemical complexity.


Although there is much that we do not know about the biochemistry of living systems, and Behe points to some good examples, we do know that they are not like designed artifacts such as mousetraps. Behe's case against evolution is a good example of the perils of being trapped by a metaphor - a metaphor that Behe has not properly understood. A closer look at human intelligent design processes reveals not secular versions of theological design and creation ex nihilo, but instead complex manifestations of analogs of evolutionary processes - this time in the domain of the cultural evolution of technological artifacts. Descent with modification is as important in the origin of artifacts as it is in the origin of species.

We have also argued that many evolved biochemical and molecular systems exhibit redundant complexity. This kind of complexity simultaneously accounts for the stability of evolved biochemical systems and processes in the face of even quite radical perturbations, for biochemical and metabolic plasticity, and, mainly as a result of gene duplication, for the co-optation of extant structures and processes in the course of evolutionary time to serve novel functional ends.

More importantly, redundant biochemical complexity points to the general biochemical inadequacy of the mousetrap analogy while providing a natural evolutionary basis for the appearance of biochemical systems manifesting irreducible complexity.

Of course, for some types of engineering problem, intelligent human designers build in redundancy and back-up systems. Perhaps Behe might want to argue that the resulting artifacts, with their engineered redundancy, suggest a more sophisticated design analogy. The trouble here is that naturalistic, evolutionary processes give rise to similar biochemical redundancies. And evolutionary processes do so without appeals to supernatural biochemical designers of unknown identity, using unknown materials and methods.

By Niall Shanks and Karl Joplin
This version might differ slightly from the print publication.