In celebration of the 20th anniversary of biochemist Michael Behe’s pathbreaking book Darwin’s Black Box and the release of the new documentary Revolutionary: Michael Behe and the Mystery of Molecular Machines, we have highlighted some of Dr. Behe’s “greatest hits.” The following is a talk delivered at the American Museum of Natural History on April 23, 2002. Behe spoke as a participant on a panel including ID proponent William A. Dembski and evolutionists Kenneth R. Miller and Robert T. Pennock. Eugenie C. Scott of the National Center for Science Education moderated. An introduction was given by Richard Milner, editor of Natural History. Get your copy of Revolutionary now! See the trailer here.
Thanks very much, Dr. Scott! It’s great to be back in New York City. I taught at Queens College and City University for three years in the early nineteen-eighties; my wife grew up on Cambreleng Avenue near 187th Street in the Bronx and our first child was born here, so New York holds many happy memories for our family.
My talk will be divided into four parts: first, a sketch of the argument for design; second, common misconceptions about the mode of design; third, misconceptions about biochemical design; and finally, discussion of the future prospects of design. Before I begin, however, I’d like to emphasize that the focus of my argument will not be descent with modification, with which I agree. Rather, the focus will be the mechanism of evolution — how did all this happen, by natural selection or intelligent design? My conclusion will not be that natural selection doesn’t explain anything; Rather, the conclusion will be that natural selection doesn’t explain everything.
So, let’s begin with a sketch of the design argument. In the Origin of Species, Darwin emphasized that his was a very gradual theory; natural selection had to work by “numerous, successive, slight modifications” to pre-existing structures. However, “irreducibly complex” systems seem quite difficult to explain in gradual terms. What is irreducible complexity? I’ve defined the term in various places, but it’s easier to illustrate what I mean with the following example: the common mousetrap. A common mechanical mousetrap has a number of interacting parts that all contribute to its function, and if any parts are taken away, the mousetrap doesn’t work half as well as it used to, or a quarter as well — the mousetrap is broken. Thus it is irreducibly complex.
Suppose we wanted to evolve a mousetrap by something like a Darwinian process. What would we start with? Would we start with a wooden platform and hope to catch mice inefficiently? Perhaps tripping them? And then add, say, the holding bar, hoping to improve efficiency? No, of course not, because irreducibly complex systems only acquire their function when the system is essentially completed. Thus irreducibly complex systems are real headaches for natural selection because it is very difficult to envision how they could be put together — that is, without the help of a directing intelligence — by the “numerous, successive, slight modifications” that Darwin insisted upon. Irreducibly complex biological systems would thus be real challenges to Darwinian evolution.
Yet modern science has discovered irreducibly complex systems in the cell. An excellent example is the bacterial flagellum which is literally an outboard motor that bacteria use to swim. The flagellum has a large number of parts that are necessary for its function — a propeller, hook, drive shaft, and more. Thorough studies shows it requires 30-40 protein parts. And in the absence of virtually any of those parts, the flagellum doesn’t work, or doesn’t even get built in the cell. Its gradual evolution by unguided natural selection therefore is a real headache for Darwinian theory. I like to show audiences this picture of the flagellum from a biochemistry textbook because, when they see it, they quickly grasp that this is a machine. It is not like a machine, it is a real molecular machine. Perhaps that will help us think about its origin.
I have written that not only is the flagellum a problem for Darwinism, but that it is better explained as the result of design — deliberate design by an intelligent agent. Some of my critics have said that design is a religious conclusion, but I disagree. I think it is wholly empirical, that is, the conclusion of design is based on the physical evidence along with an appreciation for how we come to a conclusion of design. To illustrate how we come to a conclusion of design, let’s look at the following. This is a Far Side cartoon by Gary Larson showing a troop of jungle explorers, and the lead explorer has been strung up and skewered. Now, everyone in this room looks at this cartoon and you immediately realize that the trap was designed. But how do you know that? How do you know the trap was designed? Is it a religious conclusion? Probably not. You know it’s designed because you see a number of very specific parts acting together to perform a function; you see something like irreducible complexity or specified complexity.
Now I will address common misconceptions about the mode of design, that is, how design may have happened.
My book Darwin’s Black Box, in which I flesh out the design argument, has been widely discussed in many publications. What have other scientists said about it? Well, they’ve said many things — not all flattering — but the general reaction is well summarized in a recent book The Way of the Cell, published last year by Oxford University Press, and authored by Colorado State University biochemist Franklin Harold, who writes, “We should reject, as a matter of principle, the substitution of intelligent design for the dialogue of chance and necessity (Behe 1996); but we must concede that there are presently no detailed Darwinian accounts of the evolution of any biochemical system, only a variety of wishful speculations.” Let me take a moment to emphasize Harold’s two points. First, he acknowledges that Darwinists have no real explanations for the enormous complexity of the cell, only hand-waving speculations, more colloquially known as “Just-So stories” — how the rhinoceros got its horn; how the bacterium got its flagellum. I find this an astonishing admission for a theory that has dominated biology for so long. Second, apparently he thinks that there is some principle that forbids us from investigating the idea of intelligent design, even though design is an obvious idea that quickly pops into your mind when you see a drawing of the flagellum or other complex biochemical systems. But what principle is that?
I think the principle boils down to this: Design appears to point strongly beyond nature. It has philosophical and theological implications, and that makes many people uncomfortable. But any theory that purports to explain how life occurred will have philosophical and theological implications. For example, the Oxford biologist Richard Dawkins has famously said that “Darwin made it possible to be an intellectually-fulfilled atheist.” Ken Miller has written that “[God] used evolution as the tool to set us free.” Stuart Kauffman, a leading complexity theorist, thinks Darwinism cannot explain all of biology, and thinks that his theory will somehow show that we are “at home in the universe.” So all theories of origins carry philosophical and theological implications.
But how could biochemical systems have been designed? Did they have to be created from scratch in a puff of smoke? No. The design process may have been much more subtle. It may have involved no contravening of natural laws. Let’s consider just one possibility. Suppose the designer is God, as most people would suspect. Well, then, as Ken Miller points out in his book, Finding Darwin’s God, a subtle God could cause mutations by influencing quantum events such as radioactive decay, something that I would call guided evolution. That seems perfectly possible to me. I would only add, however, that that process would amount to intelligent design, not Darwinian evolution.
Now let’s look at common misconceptions about biochemical design.
Some Darwinists have proposed that a way around the problem of irreducible complexity could be found if the individual components of a system first had other functions in the cell. For example, consider a hypothetical example such as pictured here, where all of the parts are supposed to be necessary for the function of the system. Might the system have been put together from individual components that originally worked on their own? Unfortunately this picture greatly oversimplifies the difficulty, as I discussed in Darwin’s Black Box. Here analogies to mousetraps break down somewhat, because the parts of the system have to automatically find each other in the cell. They can’t be arranged by an intelligent agent, as a mousetrap is. To find each other in the cell, interacting parts have to have their surfaces shaped so that they are very closely matched to each other. Originally, however, the individually acting components would not have had complementary surfaces. So all of the interacting surfaces of all of the components would first have to be adjusted before they could function together. And only then would the new function of the composite system appear. Thus the problem of irreducibility remains, even if individual components separately have their own functions.
Another area where one has to be careful is in noticing that some systems with extra or redundant components may have an irreducibly complex core. For example, a car with four spark plugs might get by with three or two, but it certainly can’t get by with none. Rat traps often have two springs, to give them extra strength. They can still work if one spring is removed, but they can’t work if both springs are removed. Thus in trying to imagine the origin of a rat trap by Darwinian means, we still have all the problems we had with a mousetrap. A cellular example of redundancy is the hugely complex eukaryotic cilium, shown here in cross-section, which has multiple copies of a number of components, yet needs at least one copy of each to work, as I pictured in my book.
Many other criticisms have been made against intelligent design. I have responded to a number of them at the following locations.
I will now discuss how I view the future prospects of a theory of intelligent design. I see them as very bright indeed. Why? Because the idea of intelligent design has advanced, not primarily because of anything I or any individual has done. Rather, it’s been the very progress of science itself that has made intelligent design plausible. Fifty years ago much less was known about the cell, and it was much easier then to think that Darwinian evolution was true. But with the discovery of more and more complexity at the foundation of life, the idea of intelligent design has gained strength. That trend is continuing. As science pushes on, the complexity of the cell is not getting any less; on the contrary, it is getting much greater. For example, a recent issue of the journal Nature carried the most detailed analysis yet of the total protein complement of yeast — the so-called yeast “proteome.” The authors point out that most proteins they investigated in the cell function as multiprotein complexes — not as solitary proteins as scientists had long thought. In fact they showed that almost 50 percent of the proteins in the cell function as complexes of a half dozen or more, such as the polyadenylation machinery shown in this figure from the paper. To me, this implies that irreducible molecular machinery is very likely going to be the rule in the cell, not the exception. We will probably not have to wait too long to see.
Another example comes from a paper published in the Journal of Molecular Biology two years ago, which showed that some enzymes have only a limited ability to undergo multiple changes in their amino acid sequence, even when the enzymes function alone, as single proteins, and even when the changes are very conservative ones. This led the author to caution that “homologues sharing less than about two-thirds sequence identity should probably be viewed as distinctive designs with their own optimizing features.” The author pictured such proteins as near-islands of function, virtually isolated from neighboring protein sequences. This may mean that even individual proteins from separate species that are similar but not identical in their amino acid sequence might not have been produced by a Darwinian processes, as most scientists thought, and as even I was willing to concede. Perhaps even I give too much unearned credit to Darwinian theory.
Finally, to show what research questions might be asked by a theory of intelligent design, I’d like to briefly describe some of my own recent work. This is the title slide of a seminar I gave six weeks ago to the biotechnology group at Sandia National Laboratory. The title, “Modeling the evolution of protein binding sites: probing the dividing line between natural selection and intelligent design,” points to a question I’m very interested in exploring. If you are someone like myself who thinks that some things in biology are indeed purposely designed, but that not all things are designed, then a question that quickly arises is, where is the broad dividing line between design and unintelligent processes? I think that question has to be answered at the molecular level, particularly in terms of protein structure.
Drawings of the bacterial flagellum picture proteins as bland spheres or ovals, but each protein in the cell is actually very complex. This ribbon drawing of bovine pancreatic trypsin inhibitor gives a little taste of that complexity. Now, proteins are polymers of amino acid residues, and some structural features of proteins require the participation of multiple residues. For example, this yellow link is called a disulfide bond. A disulfide bond requires two cysteine residues — just one cysteine residue can’t form such a bond. Thus, in order for a protein that did not have a disulfide bond to evolve one, several changes in the same gene first have to occur. Thus in a real sense the disulfide bond is irreducibly complex, although not nearly to the same degree of complexity as systems made of multiple proteins.
The problem of irreducibility in protein features is a general one. Whenever a protein interacts with another molecule, as all proteins do, it does so through a binding site, whose shape and chemical properties closely match the other molecule. Binding sites, however, are composed of perhaps a dozen amino acid residues, and binding is generally lost if any of the positions are changed. One can then ask the question, how long would it take for two proteins, that originally did not interact, to evolve the ability to bind each other by random mutation and natural selection, if binding only occurs when all positions have the correct residue in place?
Although it would be difficult to experimentally investigate this question, the process can be simulated on a computer. Here is a sample of the data I have generated over the past year or so. The filled circles are data points from a number of simulations which were all fit by the following equation, the details of which I will not bother you with here. These results were presented at the meeting of the Protein Society last summer in Philadelphia.
In the next slide the log of the expected time to generate what I call “irreducibly complex” protein features is shown as a function of the log of the population size and the log of the probability of the feature. The yellow dot is the time expected to generate a new disulfide bond in a protein that did not have one if the population size is a hundred million organisms. The expected time is roughly a million generations. The red dot shows that the expected time needed to generate a new protein-binding site would be a hundred million generations. Using data from these simulations as well as Bill Dembski’s concept of probabilistic resources, we can come to several broad, tentative conclusions: 1) that undirected irreducibly complex mutations cannot have been regularly involved in the evolution of large animals — the time frame would be too long; and 2) that undirected IC systems of the complexity of two or more protein binding sites cannot have been regularly involved in the evolution of vertebrates. This work assumed that all mutations were neutral. Future work could investigate such questions as, what if intermediate mutations are selected against? And what happens if there is competition between IC mutations and single-site mutations?
The broad motivation behind this work is to start getting some good numbers to plug into Bill Dembski’s explanatory filter, to try to come to a reasoned conclusion about where in nature design leaves off.
In summary, I want to leave you with four take-home points: 1) that the question is open: no other scientific theory has yet explained the data; 2) that intelligent design is an empirical hypothesis that flows easily from the data, as you can tell by looking at a drawing of the flagellum; 3) that there is no “principle” that forbids our considering design; and best of all, 4) that there are exciting research questions that can be asked within a design framework.