table of contents
October 2004

Darwin’s Black Box

Darwin’s Black Box, The Biochemical Challenge to Evolution, Michael J. Behe, 1996, Simon & Schuster, New York, pp 307, pbk, ISBN 0-684-83493-6

This work offers major aid and comfort to creationists. It was written by a real biochemist, and purports to be a devastating critique of Darwinian evolution. It is quite flawed in its major arguments; however it does raise some important issues. I shall first consider some of the flaws and then some of the real open issues that it touches.

Some critical remarks


In the very beginning Behe makes a small but revealing error; he mixes up microevolution and macroevolution with small and large change. He then insensibly conflates this variation of existing features and the development of new features. These are three separate concepts and three separate issues. Microevolution is standardly used to denote evolution within a species or population and macroevolution is used to denote the evolution of species and higher taxa. In its own right this is not consequential; it is only a confusion of terminology. It is revealing, however, in that it illustrates that Behe is not well familiar with evolutionary theory. In my opinion, this relative ignorance (and it is relative for Behe appears to be quite well informed in his area of expertise.)


In Chapter 2 he does some quote mining to establish that there are doubts about the soundness of Darwinism. The usual suspects are rounded up, Gould, Eldredge, Margulis, Kaufmann, and company, along with some rather dubious sources, e.g. Yockey. (He seems to have a fondness for citing fringe scientists, e.g. Dembski.) Not only are these people and their issues (except for Margulis and Kaufmann) irrelevant to his thesis, he engages in the usual misrepresentations.

An inane suggestion

Towards the end of the book, chapter 10 (p227) he proposes an inane scenario which he surely must realize is inane. I quote:

The irreducibly complex biochemical systems that I have discussed in the books did not have to be produced recently. It is entirely possible, based simply on an examination of the systems themselves, that they were designed billions of years ago and that they have passed down to the present by the normal processes of cellular reproduction. Perhaps a speculative scenario will illustrate the point. Suppose that nearly four billion years 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.)
This is a jaw-dropper, fantasy biology at its worst. This horde of unused genes waiting in reserve would never survive, unaltered and unexcised, for billions of years. No one who is at all familiar with the speed and manner that prokaryote (and protist) populations evolve would seriously entertain such a thought.

Why, then, does Behe suggest such a thing? The most charitable explanation is that he had an off moment. A less charitable explanation is that he knows rather less about observed evolution of prokaryote and protist genotypes than one would expect.

There is a still less charitable explanation. There is a problem with the design hypothesis which I fancy Behe is well aware of. It will not do to have all of the implementation of the designed biochemistry being done at the origin of life. The designer(s) must intervene every so often over the course of billions of years. Thus the blood clotting sequence cannot be introduced until there a metazoan with a circulatory system. In turn the circulatory system cannot exist without a clotting mechanism. One is quickly led to the recognition that these irreducible systems cannot be introduced piecemeal; the entire genome of the organism has to be reconfigured. Moreover this cannot be done once or even a few times. There are many such “irreducibly complex systems” in many different lineages with greatly varying ages. In short the designer(s) has been actively fiddling with the biosphere for billions of years on a large scale.

In short, Behe’s thesis implies active guided evolution, something that he carefully avoids discussing and well he should because it is a hypothesis with problems. On one hand if God is the designer he has chosen a strangely circuituous and byzantine path to get to us. On the other hand beneficent intelligent aliens that drop in every so often to fiddle with the planetary genome is rather less than plausible.

The uncharitable view is that Behe understands quite well the implication, wants it to be accepted, is not prepared to defend it, and offers a palpable false front in lieu of the argument he is not prepared to make.

Faulty metaphors

Behe relies heavily on metaphors in his arguments. There are two I wish to call particular attention to.

The first is the motorcycle. Behe argues that the motorcycle could not have evolved from the bicycle in an incremental Darwinian fashion and offers it as a representative irreducible system. He is quite correct if we conceive of the evolution of the motorcycle from the bicycle in terms of incremental additions and modifications. But this is not how the first motorcycles were made. They were made by taking pre-existing components, a bicycle and an engine, and coupling them together. This required a relatively small incremental refinement. Once you have the motorcycle you can refine it incrementally, adapting the bicycle to the engine and vice versa, improving the coupling until you indeed have an “irreducibly complex” motorcycle.

Now this sort of coupling of complex systems is just what Margulis suggests is one of the main features of cellular evolution. Behe dismisses Margulis’s ideas more or less out of hand with the observation that her schema requires complex systems to begin with. This dismissal is quite cavalier. Symbiosis of organisms is indeed a coupling of complex systems. However the mechanism is general. Given two systems A and C and a coupling b we can produce a more complex system AbC. In general the coupled system AbC can be simplified and refined to produce to produce a system A’b’C’ which is “irreducibly complex”.

More than that this composition of chemical cascades is one that is one that has been observed. Behe simply glosses over the issue.

The second metaphor is the one of the road. Behe introduces the tale of the hapless rodent who seeks his true love who his separated from him by a highway with a thousand lanes of traffic. Poor little fellow, he makes it across a few lanes, perhaps one or two or three or even four but he ends up being roadkill every time. The moral is that if you have to make too many changes to reach a goal and have to maintain viability in the interim you will never achieve the goal.

We may grant the unhappy fate of the unfortunate rodent. Consider, however, a similar metaphor. Again we have a highway with a thousand lanes of traffic. This time, however, the lanes are separated by grassy median strips. We have not one rodent but a family of them. Every so often some rodents dash across a lane to the adjacent median strip. They settle down, raise families, and entrench themselves. After a time they colonize the next strip and then the one after that and so on and so forth. The moral is that you can cover a great deal of hazardous territory if you have safety islands in between and can increase your numbers in the safety islands.

The question, then, is which metaphor applies. Behe suggests, without ever really making the case, that his chosen metaphor applies. For example, he observes that the clotting cascade is irreducibly complex. So it is in its present form. However that is not sufficient for his purposes. What he has to show and does not show is that is that it could not have evolved by successive extensions. That is, we start with a primitive clotting mechanism, a single stage producing an anti-coagulant with an enabler and a repressor. (Cells swarm with such mechanisms.) Couple that reaction subsystem with a similar one, and modify them to work together. The composite system is more sensitive and better regulated. Repeat as desirable.

The point is that for the Behe’s metaphor to apply one has to show that there are no safety islands. This he does not do. Moreover it does not suffice to argue that a particular cascade is irreducibly complex as it stands. One has to show that it cannot be produced by a composition which is then simplified. This he also does not do. This is particularly germane because there are well known examples of this process of composition and subsequent simplification.

Slighting kaufmann

Behe spends several pages discussing Kaufmann’s theories. Butchering would be a more apt description. I will confine myself here with the following quote (p190).

The above explanation may sound a bit fuzzy. Some of the fuzz is no doubt due to my modest powers of description. But a good deal is due to the fact that complexity theory began as a mathematical concept to describe the behavior of some computer programs, and its proponents have not yet succeeded in connecting it to real life.
There is indeed a lot of “hot air” in Kaufmann’s work, more mathematics and computer simulation than biology. However for a counter example, let us turn to page 273 where Behe quite blithely notes:
The transcribed mRNA is bound by a particle called a ribosome. Ribosomes are huge complexes consisting of fifty-two separate proteins (of which several are present in multiple copies) and three pieces of RNA with lengths of 120, 1,542, and 2,904 nucleotides. The ribosome can be readily broken down into two large pieces called the 30S subunit and the 50S subunit. Incredibly, the ribosome is self-assembling. Experiments have shown that when ribosomes are separated into their components and then remixed, under the right conditions the components will spontaneously reform ribosomes.
The ribosome is a spectacular example of Kaufmann’s “self organizing complexity”. Nor is it the only one. There are many such examples.

An overview of the argument

In summary, Darwin’s theory of natural selection is that within a species there are some heritable variations in the individuals. Those variations which are favorable will be preferentially represented in the succeeding generations. Over time the favorable variations will cumulate, effecting evolution of the species. Favorable variations are those variations which best fit the individual to the environment, i.e. evolution is adaptive.

In Darwin’s time the nature of heredity and the source of variations was not known. In succeeding years the role of genes, mutation, and Mendelian inheritance was worked out. This lead to the neo-Darwinian synthesis in which mutations (spontaneous changes in the genome – there quite a variety of ways in which this can happen) are the source of variation. The neo-Darwinian synthesis can be summarized by the aphorism, “Evolution is the change in allele frequencies in a population over time.”

Darwin’s approach to evolution is a black-box approach. That is, we consider the gross morphological features (the traits) of organisms and concentrate on how the traits have adaptively altered over time. A good example of the Darwinian approach is the study of Darwin’s finches in The Beak of the Finch. The beaks of Galapagos finches vary a good deal, both within species and between species. Natural selection acts very promptly to select those varieties that have beaks that are optimally shaped.

Another example is Dawkins’ discussion of the evolution of the eye. He points out that there is a sequence of gradations of functional variations of the visual system that occur in nature, ranging from the photo sensitive eyespot through primitive lensless cups all the way up to the full vertebrate eye. Each version has adaptive value; each step in the sequence offers additional adaptive value.

The main feature of the black box approach is that we consider the functional utility of adaptations; we do not consider how the transition from one set of traits to another is actually effected. Darwin’s “black box” approach was necessary in view of the limited knowledge of the biochemistry of life extant in his time.

The neo-Darwinian synthesis was also a “black box” approach. At the time when the synthesis was being hammered out (1938-1950) relatively little was actually known about the biochemistry of genetics and how genes determined traits. The synthesis was a resolution of population genetics (which was principally a mathematical study) and the observations of the naturalists and biologists.

(Behe does not mention the argument about “soft” inheritance versus “hard” inheritance. The biologists held out for soft inheritance, i.e., traits were mostly inherited in a continuous spectrum whereas the geneticists held out for hard inheritance, i.e., traits differed discretely. The argument was resolved by making the assumption that soft inheritance was governed by many genes.)

In the past 50 years the black boxes have been opened up. We now know a great deal more about the actual biochemical processes that occur in life. These processes are complex and have a high degree of specificity.

Behe introduces the notion of “irreducible complexity”. There are some problems with his formulation but the concept is simple enough. A system composed of discrete components is irreducibly complex if there is a system function for which all components are necessary. He uses the standard mousetrap as an example; people can and have quibbled about whether it is truly irreducibly complex which is unfortunate because it does illustrate the concept nicely. He then devotes several chapters to giving examples of biochemical systems which are irreducibly complex. The examples include:

  • The synthesis of AMP, a basic chemical used in all cells.
  • The cilium in Eukaryotes.
  • The clotting sequence in mammals.
  • The antibody formation and defense sequence.
Most of these examples are cascades (the cilium is not). That is, there is a sequence of reactions, each one dependent on the result of the previous one, accompanied in some cases by a repression mechanism to turn the cascade sequencing off. Some, e.g. the AMP synthesis, are a one in, out sequence. In some the cascade has a multiplier effect, i.e., one in, many out.

Behe argues that these systems are irreducibly complex. He goes on to argue that they cannot be the result of Darwinian evolution and must therefore have been designed. The argument against Darwinian evolution (as he portrays it) is a strong one; there are real problems with simple conceptions of Darwinian evolution. His conclusion, that these systems must have been designed, rests on very weak arguments.

What is darwinian evolution?

Evolution, as Darwin formulated it, is very much a matter of small incremental refinements which depends on soft inheritance. That is, variations in the expressed phenotype (apparently) vary continuously. Existing traits can be emphasized or de-emphasized, depending on the exigencies of selection. New features can develop by exaptation. An essential feature of the Darwinian schema is that features must be functionally adaptive during all stages of their evolution.

In the neo-Darwinian synthesis it is assumed that variation arises in the genome via mutation (increasing the variety of alleles) and is reduced by selection and genetic drift. In general the coupling between the observed phenotype (which is subject to selection) and the genotype is very indirect in multi-cellular organisms. Most of the evolution of the genotype appears to be genetic drift; most alleles are neutral, i.e., they have a zero selection coefficient.

The nub of Behe’s argument is that Darwinian evolution cannot produce irreducibly complex systems. The argument is that such systems have no functional value until unless all of the components are present; incremental refinement of the incomplete system cannot be selected for. (Behe notes that an irreducible complex system can arise by chance increments but correctly concludes that the likelihood of this occurring is infinitesimal.)

An essential part of this argument which Behe does not properly appreciate is that the components of the system are discrete components. It is the discreteness of the components that forces them to be added as a whole.

The weakness of Behe’s argument is that he (of necessity) he relies heavily on an appeal to the inadequacies of the naive model of Darwinian evolution. In effect, he says “method D does not work; ergo such systems cannot be evolved”. To do this he rather cavalierly ignores non-D possibilities.

Evolution of systems

How can such systems evolve? That is, given a suite of systems S1, S2,… how can new systems arise. There are a number of possibilities for cascades. They include:

(a) Extension by adding a component. This may or may not change the function of the system.

(b) Composition of two systems. The output of system S1 becomes the input to S2; the result is a composite system S3.

(c) Replacement of a component or subsequence by a component or subsequence from another cascade.

Note that these are rather general possibilities. They do not necessarily require initial modification of the genome; alternate biochemical pathways can arise naturally. Once they have arisen, however, and have functional value they can be streamlined by natural selection operating on the genome in the manner predicated by the neo-Darwinian synthesis.

These are not the only kinds ways in which new systems can arise or existing systems be modified. A cascade is a linear sequence in time of operations O1->O2->O3… effected by components C1, C2, C3…. In general, however, the biochemistry of the cell is a matter of multiple processes operating concurrently and with subsystems having multiple inputs and outputs. One can have the merging of entire subsystems, e.g. in the establishment of symbiotic relationships as explored by Margulis.

Most enzymes are used to perform a specific task, such as cleave lactose into galactose and glucose (the constituent parts). It does this by cleaving beta-d-galactoside linkages (hence its name: beta- d-galactosidase, which is really nothing more than lacZ, the most studied gene in bacteria). But, although its natural target is lactose, it can also cleave any other beta-d-galactoside linkage with varying efficiency. This is used to good effect by providing a chemical called X-gal, where the ‘X’ part is linked to galactose through a b-d-linkage. When cleaved, the ‘X’ part becomes an insoluble blue dye.

Most enzymes are optimized on their main target, but can also use other closely related chemicals at lower to much lower efficiences or rates. For another example, an enzyme that adds a methyl to adenine may also add it to guanine at a much, much lower rate (this would be a side reaction of no necessary general consequence).

Needless to say, these ‘inconsequential’ side reactions provide ideal starting points for evolving a ‘new’ enzymatic activity that can efficiently perform the previously ‘inconsequential’ side activity. Most evolutionary change doesn’t involve radically new enzymatic activities (there are only about six classes of enzymatic reactions), just more efficient utilization of different substrates.

This page was last updated October 14, 2004.

site map
October 2004