Skip navigation

About IDEA Center

News & Events



IDEA Student Clubs


Contact Us



Evolution by Intelligent Design: A Response to Lenski et al.

Lenski et al., "The Evolution of Biological Complexity, " Nature, 423:139-144 (May 8, 2003)

by Casey Luskin (with a little help from my friends) (Updated November, 2006)

On May 8, 2003, Nature published an article, "The Evolutionary Origin of Complex Features" (by Richard E. Lenski, Charles Ofria, Robert T. Pennock, and Christoph Adami, Nature, 423:139-144; herein the "Lenski article," or "Lenski paper"), which uses computer simulations of evolution via a program called Avida, to purport to "demonstrate the validity of the hypothesis, first articulated by Darwin and supported today by comparative and experimental evidence, that complex features generally evolve by modifying existing structures and functions" (internal citations removed). Not only does the paper (and some of its subsequent evolutionist commentators) vastly overstate the implications for the creative power of Darwin's mechanism, it also unwittingly gives credibility to the intelligent design hypothesis.

  • Part 1: Response to the Substantive Claims of Lenski, et al.
  • Part 2: How Lenski et al. Lends Credibility to Intelligent Design
  • Part 1: A Response to the Claims of Lenski et al.

    A. Some insightful comments from William Dembski:
    The Lenski paper has been cited by some evolutionists as providing a refutation of the claims of ID proponents. William Dembski recounts that biologist Andrea Bottaro cites to the Lenski paper as one example of a refutation of the claims of ID. Below is an excerpt of Dembski's response to the Lenski article: "What about [Bottaro's] reference #7? This reference is to Richard Lenski et al.'s May 8, 2003 paper in Nature titled "The Evolutionary Origin of Complex Features." This paper describes a computer simulation and thus contains no actual biology. Go to the discussion section, and you'll read "Some readers might suggest that we 'stacked the deck' by studying the evolution of a complex feature that will be built on simpler features that were also useful. However, that is precisely what evolutionary theory requires..." In other words, the computer programmers built into the simulation what they thought evolution needed to make it work. The validity of this study depends on whether the simulation faithfully models biological reality.

    "Unfortunately the simulation presupposes the very point at issue. It therefore begs the question and doesn't prove a thing about real-life biological evolution. The Lenski simulation requires that complex systems exhibiting complex functions can always be built up from (or decomposed into) simpler system exhibiting simpler function. This is a much stronger assumption than merely allowing that complex systems may include functioning subsystems. Just because a complex system can include functioning subsystems doesn't mean that it decomposes into a collection of subsystems each of which is presently functional or vestigial of past function and thus amenable to shaping by natural selection.

    "The simulation by Lenski et al. assumes that all functioning biological systems are evolutionary kludges of subsystems that presently have function or previously had function. But there's no evidence that real-life irreducibly complex biological machines, for instance, can be decomposed in this way. If there were, the Lenski et al. computer simulation would be unnecessary. Without it, their demonstration is an exercise in irrelevance. Bottaro's 'fatal blow' against irreducible complexity is nothing of the sort."
    (William Dembski, Introduction to Uncommon Dissent, pg. xxv, xxix (ISI Books, 2004).)
    B. Some Musings on the Lenski Paper by Casey Luskin:
    When I originally read the Lenski paper a couple years ago, I felt that they did make a good faith effort to model some aspects of biological reality. For example, the simulations began with a population of "computer programs" or "digital organisms" which could reproduce but occasionally made mistakes (mutations) when copying their code. Fitness was also increased by having a mutation which added a "logic operation" to its code. I recommend reading the original paper to see how the logic operations can build upon one another to produce the target logic operation, called "EQU."

    I break my criticisms into the following categories:
    1. Stacking the Deck: It was pre-ordained that the complex function can be created from the less complex functions (they hand-coded a solution before even running the simulation)--but there is no such guarantee in biology that subsystems can be so easily combined to produce anything useful! The complexity gap between the smaller functions (NAND, etc.) and the target functions (EQU) is not very big. In fact, they were able to create EQU using only 5 of the more primitive logic operation subsystems. This means that as far as logic is concerned, only 5 of the basic logic functions used in the programs are needed to evolve EQU. They created a simulation which they knew could evolve the target function through the subsystems. (This is why I have titled this critique "Evolution by Intelligent Design.")
    2. Too Much Selective Advantage: Selective advantage was given to literally every single addition of logic functions in the organisms which evolved EQU. Additionally, every mutation which added code, always added functional line(s) of code, while in nature mutations are never guaranteed to have any meaning or functionality in the environment. This makes the evolution of EQU essentially inevitable, and it does not test irreducible complexity. In a true irreducibly complex system, there will be no selective advantage along an evolutionary pathway. In real world, there is no guarantee that the subsystems you need will necessarily give you a selective advantage along your evolutionary pathway.
    3. Illustrating that Irreducible Complexity is Unevolvable: When the aforementioned "selective advantage" was taken away, and fitness only increased when the target function EQU appeared, EQU NEVER EVOLVED in their simulations! This is very significant because it shows that they modeled true irreducible complexity, and that when they did, irreducible complexity could not evolve!
    C. "Stacking the Deck"
    As Dembski notes, the simulation did indeed "stack the deck" in favor of evolution. The subsystems in the Lenski article were guaranteed to be capable of producing the target function from the very beginning. In fact, the authors were able to hand-code (i.e. intelligently find) a solution to create the target system ("EQU") from the simpler functions which required only 5 of the logic operations. This means that the target system is, logically speaking, just a combination of a few subsystems! It is not clear this models biological reality because the authors knew the subsystems were eminently capable of capable of creating the target system completely apart from running the simulations. The evolution seems to be guaranteed to be possible--which does not test Darwin's theory--it assumes it.

    D. Too Much Selective Advantage
    A selective advantage was granted with the "evolution" (i.e. addition) of any additional logic operations. These logic operations were the very tools by which the target function could be built. It thus seems unsurprising and inevitable that the target function evolved with such stepwise selection pressure in a population. This is like evolution which has been pre-programmed by intelligent design. Darwinian processes are just the vehicle to achieve the inevitable simple "evolution" of the easily-evolvable target system (EQU). Whether or not this is the biological reality was not tested. In the real world biological function is much more difficult to come across. For example, a paper in Journal of Molecular Biology calculates that the "prevalence of sequences performing a specific function by any domain-sized fold may be as low as 1 in 1077, adding to the body of evidence that functional folds require highly extraordinary sequences." (Axe, JMB, 341:1295-1315 (2004).) Every mutation had meaning within the environment, distancing the simulation from biological reality.

    Regarding biological reality, one critique of Avida published in a biology journal recognized the unrealistic ease by which Avida allowed evolution to proceed: One of the 26 possible instructions in a creature’s ‘genome’ is a logic operation (NAND), whilst the others perform various manipulations: copying, input/output, and so on. Composite logic operations are valued according to the number of elementary NAND operations needed to perform them. The most valuable is EQU (‘equal’), which returns a 1 only if both input bits are the same. This requires five NAND operations, as well as other operations which move intermediate results between registers. A hand-written program required 19 operations to achieve EQU; a digital organism needs additional code for replication. ... In Lenski et al.’s artificial organisms, the mutation rate per site is quite high (0.0025), so that favourable pairs can be picked up by selection at an appreciable rate; this would be unlikely in most real organisms because, in these, mutation rates at each locus are low.

    (Nick Barton and Willem Zuidema, "Evolution: The Erratic Path Towards Complexity," Current Biology, Vol. 13, R649–R651, August 19, 2003)
    Another way the study granted too much selective advantage was that every mutation inserted a line of code which had pre-set meaning and functionality in the computer-environment. The biological analogy might be inserting a fully functional protein via the duplication of a gene. Such gene-duplication events do occur in biology, but biochemical evolution at the gene-level must also via point mutations (or insertions or deletions of segments of DNA within a gene). Point mutations, insertions, or deletions need not add such discrete and meaningful functionality, however every mutation in the Lenski paper which added code represented the addition a of a complete and discrete function. The paper thus granted the addition of too much functionality for each "mutation" and did not accurately model biology in this respect. Were it to accurately model biology, the study might have randomly substituted, inserted, or deleted code at the lower level of words or even letters into the code of the digital organism. While this might have had the effect of creating many syntax errors in the code of the digital organism, it might have been closer to what really happens in biology when a gene experiences substitutions, insertions, or deletions of nucleotides.

    By allowing for a selective advantage every time a new logic operator was formed, their fitness function may look something like this:


    Diagram 1: In the above diagram, a selective advantage is essentially given every time you get closer to the target function, achieved at peak EQU. Thus, it is unsurprising that EQU evolved when an increased selective advantage was granted with the addition of each logic operation.

    It should be noted that the Lenski study found that sometimes fitness would decrease (i.e. they might lose some key logic functions after a mutation), only to then later increase by a subsequent mutation. Thus, they might claim their fitness function looks something like this:


    Diagram 2: Sometimes during the simulation, fitness decreased (or failed to increase), but yet the simulation still managed come down from relative adaptive maximum "r" and find its way to the higher adaptive peak EQU. The most striking example was one mutation which decreased fitness by 50%, only to allow EQU to evolve on the subsequent mutation. The fact that it took only one more mutation for fitness to go back up again shows this wasn't a very wide gulf to bridge. Thus, as shown in the above diagram 2, this is a fairly narrow gulf to be bridged, and the red line is meant to represent a "narrow" gulf that could be bridged by only one mutation in the simulation.

    The authors felt this was an important finding, because "[t]he evolution of a complex feature, such as EQU, is not always an inexorably upward climb toward a fitness peak, but instead may involve sideways and even backward steps, some of which are important." But when only one mutation is required to overcome the drop in fitness, this is not so impressive. Furthermore, as noted below, when irreducible complexity is accurately modeled, the simulation could not overcome the lack of selection pressure for evolving additional logic operations.

    E. Modeling Irreducible Complexity
    The paper made one profound finding when it accurately modeled true irreducible complexity (first full paragraph, pg. 143). Michael Behe has defined irreducible complexity as: "An irreducibly complex evolutionary pathway is one that contains one or more unselected steps (that is, one or more necessary-but-unselected mutations). The degree of irreducible complexity is the number of unselected steps in the pathway." (A Response to Critics of Darwin’s Black Box, by Michael Behe, PCID, Volume 1.1, January February March, 2002; When Lenski et al. created a simulation with high irreducible complexity, i.e. there was no selective advantage until the target function arose, EQU never evolved! Consider this quote from the Lenski paper: "At the other extreme, 50 populations evolved in an environment where only EQU was rewarded, and no simpler function yielded energy. We expected that EQU would evolve much less often because selection would not preserve the simpler functions that provide foundations to build more complex features. Indeed, none of these populations evolved EQU, a highly significant difference from the fraction that did so in the reward-all environment (P ~= 4.3 x 10-9, Fisher's exact test)." (emphasis added) In other words, when there is no selective advantage until you get the final function, the final function doesn't evolve. In this case, their simulation probably DID model biological reality because irreducible complexity claims that there is no advantage until you get the final function. In fact in such a scenario, it found that the evolution of such a structure was impossible. In other words, they just proved that irreducible complexity is unevolvable. Unfortunately, the authors failed to recognize this fact in their conclusion and continued to state that the creative power of Darwin's mechanism had been validated. This scenario is depicted in the diagram below:

    Diagram 3: Here there is no selective advantage until a high level of complexity is achieved. If a level of complexity yields no fitness, then there is no Darwinian evolutionary reason why it should exist. Evolution thus cannot produce a function unless it confers some advantage upon the organism. If F <= 0, then there is no function to select. If dF/dG = 0, as is the case if attempting to evolve from point A to point B, then there is no selection pressure to move the genome in the direction of what is necessary to acquire the target function, EQU. This is how a fitness function looks for true irreducible complexity where there functionality is nonexistent until one achieves all the parts necessary for the function itself. Significantly, when Lenski et al. modeled irreducible complexity in this manner, EQU never evolved.

    The Lenski article makes a bold claim: "Our experiments demonstrate the validity of the hypothesis, first articulated by Darwin and supported today by comparative and experimental evidence, that complex features generally evolve by modifying existing structures and functions." But it seems clear that this claim is based upon an inadequate test of the model. Furthermore, this claim seems to contradict the finding of the Lenski paper that when no selective advantage exists until the target function is achieved, evolution cannot produce that final function. If anything, this paper confirms a prediction made by Charles Darwin in 1859: "If it could be demonstrated that any complex organ existed which could not possibly have been formed by numerous, successive, slight modifications, my theory would absolutely break down." Perhaps this quote from Darwin would have made a more appropriate conclusion to the Lenski paper.

    Part 2: How Lenski et al. Lends Credibility to Intelligent Design
    The Lenski paper can only be seen as a scientific response to the claims of ID proponents, published in a high profile journal such as Nature. Despite the fact that the authors of the Lenski paper would likely deny this fact, there are many clues which show that the article is intended as a rebuttal to the claims of ID proponents. Not only does this validate the work of ID proponents as posing a legitimate challenge to Darwin's theory, but it also indicates that the claims of ID proponents are eminently testable, falsifiable (though as discussed above, not yet falsified), and therefore also scientific in nature.

    Exhibit A: A notable co-author of the Lenski paper is philosopher Robert T. Pennock. Pennock is the author / editor of two books critiquing intelligent design (Tower of Babel [author] and Intelligent Design Creationism and its Critics [editor]) and is also a public activist, having testified in favor of evolution before the Texas State Board of Education. While Pennock is surely an accomplished academic, his co-authoring of a Nature article on evolving complexity is very odd for someone who teaches nothing but courses in philosophy, humanities, and history at Michigan State University. Pennock is surely a qualified academic, but he is not trained in the fields from which one should be publishing science articles in Nature? This odd fact turns out to have a simple explanation.

    Given the odd presence of a an anti-ID co-author such as Pennock, the Lenski paper can be viewed as nothing other than a response to the claims of intelligent design proponents. While the Lenski article is careful to avoid legitimization of intelligent design by avoiding any direct citations to ID proponents, the article is couched as nothing less than a rebuttal to the claims of ID proponents. This brings us to Exhibit B: the abstract of the paper begins "A long-standing challenge to evolutionary theory has been whether it can explain the origin of complex organismal features" and ends "These findings show how complex functions can originate by random mutation and natural selection." Thus the article is framed as a rebuttal to those who "challenge" evolutionary theory.

    The article even attempts to address irreducible complexity without using the term. "Thus, although more than two dozen mutations were used to build EQU, undoing any one of them destroyed this function." They are stating that EQU was irreducibly complex, but yet it evolved. Thus, Exhibit C is as follows: the article directly purports to test the evolution of irreducible complexity but yet never uses the phrase. (Note: It is arguable that their stated conclusions about the evolution of irreducible complexity do not match the findings of their simulations. When EQU evolved, the study did not truly model irreducible complexity because it employed a "reward-all" environment where some function could be gained by adding parts which could also contribute to the final function. When the article properly modeled irreducible complexity, where only EQU was rewarded, EQU never evolved!)

    Exhibit D: To ice this cake, consider this statement from the February, 2005 of Discover Magazine (where references to the true point of the article were apparently not censored): "Their work hit a nerve in the antievolution movement and hit it hard. A popular claim of creationists is that life shows signs of intelligent design, especially in its complexity. They argue that complex things could never have evolved, because they don't work unless all their parts are in place. But as Adami points out, if creationists were right, then Avida wouldn't be able to produce complex digital organisms." (text available at Thus, we have co-authors of the paper and other prominent evolutionist advocates publicly proclaiming that the paper rebuts the claims of creationists. While the author Carl Zimmer quotes co-author Christopher Adami to sing an over-inflated victory song, one important point should not be lost: this study implicitly proves that the claims of ID proponents, such the claim that some biological features are irreducible complexity, are eminently testable via the methods of science. Apparently Nature saw the claim of irreducible complexity as such a threat to evolution that it saw fit to publish a study which attempted to model the evolution of irreducible complexity.

    The publication of this paper presented a dilemma to the authors because they wanted to simultaneously refute ID, and validate the faith of the scientific community in Darwin's magic mechanism, but yet not validate ID in the process through any references to proponents of ID-concepts. This strategic tactic has been employed numerous times in the Darwinist community, as NCSE Director Barbara Forrest has implied that Darwinists should not share the stage with ID proponents at conferences because it "lends [ID proponents] an undeserved academic legitimacy" (see "Barbara Forrest's Letter to Simon Blackburn"). Perhaps she gave similar advice to Pennock. (Pennock and Forrest think alike and have worked together against ID: After she encourages academics to avoid ID proponents, Forrest recommends Pennock's book in the very next paragraph of her Letter to Simon Blackburn; Pennock has returned such favors as he contributed reviews of the pre-print manuscript of Forrest's book against ID [see pg. viii].) In the case of the Lenski paper, the careful balancing act failed in its attempt to refute ID without legitimizing ID literature.

    All in all, it is a great thing that a study in a journal like Nature took the challenges of ID proponents seriously and tried to test them through a computer simulation. ID proponents should laud any scientist who tests and critique the claims of ID, regardless of the outcome of the study. However, it is unfortunate that here, the study was done published in such a manner which avoided acknowledgment that it was actually testing the claims of ID proponents. This covert "refutation without citation" does not do service to unbiased scientists who want to really investigate and understand the challenges posed to Darwin's mechanism by ID proponents.

    In the future, when scientific journals want to attempt to publish scientific refutations of intelligent design, but walk-the-tightrope and in-doing-so, avoid legitimizing the claims of intelligent design proponents, they may want to consider doing the following:
  • Do not list co-authors who are well-known anti-ID advocates who have no qualifications to author such a paper other than that they have extensive knowledge about ID theory.
  • Don't couch the article as an answer to "challeng[es]" to evolutionary theory.
  • Don't permit prominent evolution advocates to interview co-authors of the paper so they can be quoted in popular magazines discussing the article as if it is a rebuttal to ID proponents.