A reply to Dembski's review of my book

[Image of cover of Rosenhouse book]

William Dembski has now reviewed my new book The Failures of Mathematical Anti-Evolutionism. His review comes to roughly 18000 words. According to at least one standard classification scheme, that makes it a novella. The full review can be found here. It is also appearing in sections at the Discovery Institute's site Evolution News (see here).

It is a person of rare talent who can write at such length without getting anything right. Dembski’s response consists mostly of lengthy, irrelevant digressions that have little to do with anything I actually wrote, out-of-context sentences and sentence fragments from my book that seem specifically intended to mislead his readers about my arguments, and his usual arrogance and self-puffery about his own brilliance. He obsesses over minor comments I include in a chapter’s endnotes, while ignoring the main argument I made in the chapter itself. When he does get around to responding to an argument, he seems in most cases not even to understand the point I was making. Frankly, I could easily have written a more cogent reply to my own book.

I will only respond to what I regard as the most important points and will simply ignore large swaths of his review. To anything that I do not specifically address, you can assume my response is to roll my eyes and say, “Whatever. Let him blather.”


Anti-evolutionists routinely present spurious probability calculations meant to refute evolution. In a lengthy chapter on probability, I explain that a proper calculation must take place in the context of what mathematicians refer to as a “probability space”. For our purposes, this means that you must have a good grasp on the range of possible outcomes, as well as an understanding of the probability distribution appropriate to those outcomes. In the context of the evolution of complex adaptations, we never have what we need to do this. As Harvard biologist Martin Nowak put it, “You cannot calculate the probability that an eye came about. We don’t have the information to make this calculation.”

I then considered several anti-evolutionist probability arguments, including the notorious flagellum calculation Dembski presented in his book No Free Lunch. I argued that his calculation was worthless for two main reasons: It was based on the false assumption that an “irreducibly complex” system could not evolve gradually, and it assumed a completely unrealistic model of how such structures arise.

How does Dembski respond to all this? Mostly by complaining that I described his work as an attempt at “mathematical proof” that the flagellum is the result of intelligent design:

Rosenhouse, as a mathematician, must at some level realize that he’s prevaricating. It’s one thing to use mathematics in an argument. It’s quite another to say that one is offering a mathematical proof. The latter is much much stronger than the former, and Rosenhouse knows the difference. I’ve never said that I’m offering a mathematical proof that systems like the flagellum are designed. Mathematical proofs leave no room for fallibility or error. Intelligent design arguments use mathematics, but like all empirical arguments they fall short of the deductive certainty of mathematical proof.

This appears in Section 9 of Dembski’s review. This section is more than 1500 words, and it consists of little else than variations on this complaint.

If Dembski really wants to cavil over the difference between “mathematical proof” and “a mathematical argument so powerful it should convince any reasonable person” then he is welcome to do so. None of my arguments are affected in the slightest by this sort of asinine rhetorical nitpicking. My criticism of Dembski’s flagellum calculation was not that it falls short of deductive certainty. Rather, my criticism was that the calculation was based on so many biologically ludicrous assumptions that it is completely worthless.

The relevant paragraph from No Free Lunch is this:

An irreducibly complex system is a discrete combinatorial object. Probabilities therefore naturally arise and attach to such objects. Such objects are invariably composed of building blocks. Moreover, these building blocks need to be [sic] converge on some location. Finally, once at this location the building blocks need to be configured to form the object. It follows that the probability of obtaining an irreducibly complex system is the probability of originating the building blocks required for the system, multiplied times the probability of locating them in one place once the building blocks are given, multiplied times the probability of configuring them once the building blocks are given and located in one place. (No Free Lunch pp. 290-291)

I replied that this model for understanding the evolution of the flagellum is flatly ridiculous. An irreducibly complex system cannot be treated as a discrete combinatorial object. Nor can we model the construction of the flagellum by multiplying three independent probabilities related to origination, localization, and configuration. All of this is discussed at length in my book.

In his review, Dembski makes no attempt to defend the biological legitimacy of his model. Instead, he says things like this (still in Section 9):

Thus, I’m supposed to be presupposing that irreducible complexity makes it impossible for a system to evolve by Darwinian means. . . . But that’s not what I’m doing. Instead, I’m using irreducible complexity as a signpost of where to look for biological improbability.

That is a pretty strange thing to write, considering that in No Free Lunch he wrote this:

Richard Dawkins has memorably described this gradualistic approach to achieving biological complexity as “climbing Mount Improbable.” . . . For irreducibly complex systems that have numerous diverse parts and that exhibit the minimal level of complexity needed to retain a minimal level of function, such a gradual ascent up Mount Improbable is no longer possible. (No Free Lunch p. 290)
“A gradual ascent up Mount Improbable is no longer possible.” Sure sounds like he is assuming that an irreducibly complex system cannot evolve gradually. Moreover, irreducible complexity is only a signpost of where to look for biological improbability if you accept the premise that such a system cannot evolve gradually. Since that premise is false, irreducible complexity is not a signpost for anything.

I think what is going on here is that Dembski feels stung by another argument I present in my book. I point out that in his discourse, it is really Michael Behe’s claims about irreducible complexity that are doing all the work. The probability calculations do nothing to strengthen the argument. Recall that Behe, in Darwin’s Black Box wrote:

The result of these cumulative efforts to investigate the cell---to investigate life at a molecular level---is a loud, clear, piercing cry of “design!” The result is so unambiguous and so significant that it must be ranked as one of the greatest achievements in the history of science. The discovery rivals those of Newton and Einstein, Lavoisier and Schrodinger, Pasteur and Darwin. (Darwin’s Black Box, p. 233)
How interesting that Behe did not say, “Irreducible complexity in the cell is interesting and all, but what we really need is a probability calculation to seal the deal.”

Returning to the review, after several paragraphs spent retreating from his own calculation, Dembski writes:

The fact is that they [Rosenhouse and his fellow Darwinists] have no probability estimates at all for the evolution of these systems. Worse yet, because they are so convinced that these systems evolved by Darwinian means, they know in advance, simply from their armchairs, that the probabilities must be high. The point of that section in No Free Lunch was less to do a definitive calculation for the flagellum as to lay out the techniques for calculating probabilities in such cases (such as the perturbation probabilities)”

Of course we have no probability estimates for the evolution of these systems. That is because probability theory is fundamentally the wrong tool for this particular job. That’s the whole point! Not only can you not rigorously calculate the probability of evolving a particular complex system, you cannot even estimate it in any reasonable way. Martin Nowak had it exactly right.

Dembski’s claim to have presented techniques for evaluating these probabilities is pure fantasy. His calculations are just biological numerology. He assigns numbers to objects in random ways and then, by manipulating the numbers, pretends that he has learned something about the objects. Actually, he has only learned something about the way he assigned the numbers.

And just to show that Dembski cannot be counted on to get even the simplest things right, let me also note that I certainly do not just assume that the probability of evolving a flagellum is high. I distinctly remember writing “[T]he course of evolution is affected by so many chance events that the probability of any specific modern outcome of the process could be extremely low, but this does not make us suspect design, because something had to happen. Therefore, some additional argument is needed to go from low probability to a conclusion of design” (p. 131). I revisit this issue again on page 132, where I note that in considering anti-evolutionist probability, we need to pay attention to how they propose to get around this point (that low probability by itself does not mean much). It arises again on page 139 in the context of Dembski’s flagellum calculation, and yet again on page 157 in the context of Behe’s probability calculations in The Edge of Evolution.

To review: I argued that Dembski’s flagellum calculation was biologically ridiculous. For obvious reasons, he is unable to offer any response to this simple point and does not even try. Instead, he just whines about phrasing, presents ludicrous caricatures of what I said, and misses every important point. This is representative of his entire review.


Dembski argues that it is not low probability by itself that implies design, but rather the combination of low probability with a recognizable pattern. He refers to such a pattern as a “specification”.

In simplistic examples it is easy to see what he means. Any sequence of Hs and Ts in 100 coin tosses is as unlikely as any other, but 100 Hs, or a perfect alternation of Hs and Ts, embody patterns that make us think that something other than pure chance is involved. Any pattern of crags and grooves on a mountain is incredibly unlikely, but the faces on Mt. Rushmore are such that we immediately infer design.

But there is an obvious problem with applying this notion to evolutionary questions: How do we distinguish design-suggesting patterns from those we impose on nature through excessive imagination? Dembski claims that we can use the function of a complex adaptation as its specification, but biologists claim that natural selection crafts functional structures as a matter of course. Given this, does the function of a system suggest design or just natural selection? How can we be confident that in using function as a specification we are not doing the equivalent of looking at a fluffy, cumulus cloud and seeing a dragon?

In practical situations, we always make such judgements with the benefit of extensive background knowledge. We know what usually happens when coins are tossed, and we know what mountains look like when people do not carve faces into them. But we have no such background knowledge for evolution. When considering the endpoints of billions of years of evolution, we have no base of experience for distinguishing design-suggesting patterns from those that can arise through natural selection.

How does Dembski respond to all of this? Mostly by complaining that I do not discuss “specification” as an abstract concept. He writes (in Section 10 of his review):

Events become probabilistically more complex as they become more improbable (this is consistent with, as pointed out earlier, longer, more improbable sequences of coin tosses requiring longer bit strings to be recorded). At the same time, descriptive complexity characterizes patterns that describe events via a descriptive language. Descriptive complexity differs from probabilistic complexity and denotes the shortest description that will describe an event. The specification in specified complexity thus refers to patterns with short descriptions, and specified complexity refers to events that have high probabilistic complexity but whose identifying patterns have low descriptive complexity.
On and on he goes in this manner, for paragraph after tedious paragraph.

Dembski accuses me of being “especially hung up on [his] notion of specification.” This is a weird charge since actually I barely discuss the notion at all. My only extended consideration of the concept occupies a mere seven pages (pp. 140-146), and the only reason the section is that long is that I spend several pages trying to bring clarity to Dembski’s frequently murky and overly-technical prose.

Those seven pages appear in a section called, “Is the Flagellum Complex and Specified?” and that should give you a clue as to why I do not dwell on Dembski’s argle-bargle about complexity measures. I did not discuss them because Dembski himself does not discuss them when he applies his ideas to the bacterial flagellum. I was not discussing “complex, specified information” as a general approach to design-detection. Instead, I was quite explicitly restricting my discussion to how Dembski applies his machinery to a biological case. (More on this in the next section).

With that in mind, let me show you Dembski’s treatment of the specification of the flagellum:

[I]n the case of the bacterial flagellum, humans developed outboard rotary motors well before they figured out that the flagellum was such a machine. This is not to say that for the biological function of a system to constitute a specification humans must have independently invented a system that performs the same function. Nevertheless, independent invention makes the detachability of a pattern from an event or object all the more stark. (No Free Lunch, p. 289)

That is his entire discussion, and there is no mention of complexity measures at all. When it comes time to apply his notions to a given biological system, Dembski apparently considers it sufficient just to refer vaguely to its function. If Dembski does not think complexity measures are important in this context, then why should I?

The concern I raised was that saying of a flagellum that it resembles an outboard motor is comparable to saying of a cloud that it resembles a dragon. There is nothing to allay this concern in Dembski’s review, but he does mention something called “detachability” in the quotation above. And that is why I discuss the notion of detachability at some length in my book. I present Dembski’s incredibly technical definition in full (p. 141-142), carefully explain what it means, and then discuss why it does not address the concern I raised.

Dembski’s definition includes numerous abstract objects: a reference class of possibilities, a chance hypothesis, a probability measure, a rejection function, and a notion of background knowledge that is defined in terms of certain conditional probabilities. It is basically a bowdlerized version of Fisherian hypothesis testing, for those with some statistical background.

But if we have little hope of defining a probability space sufficient for carrying out calculations about flagella, we have even less hope of finding real-world counterparts for any of the abstract objects in Dembski’s definition of “detachable”. This is especially true for the part about background information, as I have noted.

In No Free Lunch, Dembski does not even try to define the real-world counterparts of his abstract constructs. He presents no reference classes, probability measures, rejection functions, or relevant background knowledge. He just looks at the flagellum and says, “Kinda looks like an outboard motor.” Dembski’s liberal use of notation and jargon is useful for creating the illusion that something deep has been said, but all of that machinery goes clean out the window when it is time to apply it to evolution.


Another aspect of Dembski’s discourse is his use of the “No Free Lunch” theorems. These are actual mathematical results that say, roughly, that the average performance of a search algorithm over all possible landscapes is no better than blind search. This might make us wonder why nature presents us with just the sort of fitness landscapes on which evolution is effective.

Dembski, Robert Marks, and Winston Ewert claim to have extended these theorems. They refer to their results as “conservation of information” theorems. They define a quantity called “active information”, which roughly refers to the background information brought by a researcher to a specific search problem. Their main theorem then shows that within their formalism, the amount of information outputted by the algorithm cannot exceed the active information brought by the researcher to the problem.

In this way, they argue that evolutionary algorithms cannot create novel information. They also spend a lot of time criticizing “artificial life” experiments on the grounds that they achieve such results as they do only because the researchers bring illicit active information to their work.

In my book, I offer many points in reply. Among other things, I note that it is very unclear that evolution really is a search in the precise technical sense assumed by the No Free Lunch theorems. We have here another instance of Dembski bringing difficult mathematical formalism into his discussions, but then not really using the formalism to do anything. This is why David Wolpert, one of the discoverers of the original No Free Lunch theorems, described Dembski’s arguments as “fatally informal and imprecise” and as being “written in jello.” He elaborated:

There simply is not enough that is firm in his text, not sufficient precision of formulation, to allow one to declare unambiguously “right” or “wrong”. . . .The values of the factors arising in the NFL theorems are never properly specified in his analysis. . . .[T]hroughout, there is a marked elision of the formal details of the biological processes under consideration.

Dembski’s arguments do not improve when he starts extolling the virtues of his so-called conservation theorems. As applied to biology, his argument is nothing more than the claim that nature has to be a certain way for evolution to work. Most of us did not need difficult mathematical theorems to understand this. The fitness landscapes confronted by evolving organisms arise ultimately from the laws of physics, and Dembski and his collaborators are really just asking why the universe is as it is. It's a perfectly good question, but hardly one within biology’s domain.

How does Dembski respond to this? Mostly by complaining that I did not cite his technical papers laying out the theoretical apparatus underlying his theorems (we are now in Section 13 of the review):

As with specified complexity, in proving conservation of information theorems, we have taken a largely pre-theoretic notion and turned it into a full-fledged theoretic notion. In the idiom of Rosenhouse, we have moved the concept from level 1 to level 2. A reasonably extensive technical literature on conservation of information theorems now exists. Here are three seminal peer-reviewed articles addressing these theorems on which I’ve collaborated. . .
He now presents the bibliographic information for three of his papers and writes “Rosenhouse cites none of this literature.”

Now, first of all, my idiom was “track 1” and “track 2”. Just saying.

Secondly, do take a moment to savor the planet-sized ego it takes for anyone to describe his own work as “seminal”. For those unfamiliar with this term, academics use it to refer to work that was not only original and important when it first appeared, but which also inspired new avenues of investigation for other researchers. It is a term of respect bestowed by others years after a work is published and has proven its fruitfulness. Big egos are an occupational hazard in academe, but absolutely nobody describes his own work as seminal, especially when few other researchers seem to have paid any attention to it. See my earlier comment regarding arrogance and self-puffery.

Now let us see if we can come up with a reason why I did not cite those papers in my book. In fact, maybe we should check to see if I explained precisely this point in the book itself:

[W]e find ourselves in the same position we were in back in Section 5.6. We saw that Dembski claimed to have produced a piece of mathematical machinery---complex, specified information---that could be used to distinguish things that must have been designed from things that could be explained in some other way. He then proposed to apply this machinery to biology and claimed thereby to show that intelligent design is in some way implicated in natural history. We decided that we did not need to examine every piece of the machinery since we only cared about its claimed application to biology.

And so it is here. It would take many pages to explain and analyze the mathematical formalism proposed by [Marks, Dembski, and Ewert], but it would not further our agenda to do so. If they want to apply their techniques to abstract problems in combinatorial search, then they are welcome to do so. Whether such proposed applications have any merit would be a topic for a different book. Our interest here is solely in the manner in which they apply their machinery to biology. (pp. 212-213)

I based my discussion of “conservation of information” on their book Introduction to Evolutionary Informatics, published in 2017. They rather pompously describe their book as a monograph, which I took to mean they regarded it as primarily a work of scholarship as opposed to mere popularization. The book is far more recent than the papers Dembski mentions, and it is specifically about the proposed applications of their ideas to evolution.

See the point? I was writing a book about mathematical anti-evolutionism (note the title). The three papers Dembski faults me for not citing do not discuss biological questions. That is why I did not cite them. I cited recent writing that addresses the subject of my book, and I ignored older work that was irrelevant to my subject. How does this reflect poorly on me?

We are now in Sections 12 and 13 of Dembski’s review. True to form, he goes on and on about his own brilliance and technical sophistication. He hurls plenty of personal insults, out-of-context sentence fragments, and silly caricatures of my arguments. But it is worth wading through this because when he eventually turns his attention to what I actually said, the result is wonderfully demented. Demsbki writes:

True, he [Rosenhouse] dismisses conservation of information theorems as in the end “merely asking why the universe is as it is.” (p. 217) But when discussing artificial life, he admits, in line with the conservation of information theorems, the crucial information is not just in the algorithm but also in the environment. (p. 214)

Let us have a look at what I actually said on page 214. I had just quoted a paragraph (a whole paragraph, mind you, in contrast to Dembski’s habit of just quoting sentence fragments from me) from their book in which they were critical of the well-known Avida experiment. I replied thusly:

It is on the basis of such arguments that [Marks, Dembski, and Ewert] dismiss computer simulations of evolution as unrealistic, but their logic is hard to follow. It is self-evident that the Avida organisms found evolutionary success in part because the researchers created an environment in which success was possible. However, it is equally self-evident that the algorithm plays a big role in the success. That is, Avida’s organisms achieved success because a particular algorithm interacted with a particular environment. The algorithm and the environment are both critical, and therefore it is plainly wrong to say, “This active information source is the reason for Avida’s success.” (p. 214)

Some admission. What could be more obvious than that the environment and the algorithm are both important in evolution? Who needs difficult mathematical theorems to grasp that point?

Dembski now writes:

Yet if the crucial information for biological evolution (as opposed to artificial life evolution) is built into the environment, where exactly is it and how exactly is it structured? It does no good to say, as Rosenhouse does, that “natural selection serves as a conduit for transmitting environmental information into the genomes of organisms.” (p. 215) That’s simply an article of faith.

He now presents a somewhat cryptic quote from Holmes Rolston. Picking up the action on the other side:

So no, the information was not always there. And no, Darwinian evolution cannot, according to the conservation of information theorems, create information from scratch.

Dembski is descending into gibberish here, and his point is becoming hard to discern. I do not know what he means by “the information was not always there”. What information was not always where? Since evolution requires an interaction between the genetic systems of organisms and the environments in which they find themselves, I do not know what it means to say that the “crucial information . . . is built into the environment.” The environment and the organisms are both crucial; why single out one or the other? And since Darwinian evolution assumes that we already have both organisms and environments, I do not know what it means to suggest that Darwinian evolution “creates information from scratch”.

So let us back up a bit and try to bring some blessed clarity to the discussion.

First, there is no mystery as to how standard evolutionary processes can create new genes and adaptations. Every step in the process has been observed and is well-understood. Genes really do mutate, sometimes leading to new functionalities. Natural selection really can string together several such mutations, leading to changes in gene pools and to better adapted organisms. Gene duplication with subsequent divergence plainly leads to increased genome sizes and to new opportunities for innovation. There is no article of faith here, just simple empirical facts.

Plainly, such processes lead to the creation of novel genetic information by any reasonable definition of the term “information”. When biologists say that natural selection is a mechanism through which new genetic information can be created, this is what they mean.

Now let us turn to that sentence fragment about natural selection being an information conduit. Had Dembski quoted even a little of the surrounding context, it would have been clear that I was referring back to a discussion from earlier in the book in which I explored these issues at length. Dembski asks for where nature’s information is and for how it is structured. I addressed that question thusly:

[T]here is nothing wrong with viewing natural selection as an information conduit between the environment and a population’s gene pool. Recall that according to Shannon’s conception, information content is something possessed by an event in a probability space. Seen in that way, any physical system that can exist in more than one state can contain information. This is because if the system can exist in more than one state, then there must, in principle, be a probability distribution that describes the likelihood of being in one state versus another. And since the local environments in which gene pools find themselves can certainly exist in many states, it is not an abuse of language to say the environment contains information.

This can actually be an illuminating metaphor. There is a strong sense in which the gene pools of modern organisms can be said to record information about the ancestral environments in which they evolved. The process is not much different from receiving medical information from your doctor and then making lifestyle changes as a result. Just as your doctor gives you information on how to live a healthier life, so too does the environment give information to a gene pool about how better to survive. In the language of information theory, we would say this is communication through a noisy channel because natural selection is not the only mechanism of change, and evolution is not always adaptive. But it is an interesting way of looking at things nevertheless. (p. 204)

Now, does this constitute evolution creating information “from scratch”? Since Dembski and his coauthors are the only ones who talk like that, I am not sure what the question means. But in my book, I make the following point:

Standard evolutionary mechanisms have the ability both to increase the information storage capacity of the genome through gene duplication, and to refine, via mutation and natural selection, the information stored there, resulting in better-adapted organisms. This is what is meant when scientists attribute to evolution the ability to create new information.

If you now want to play gotcha, and argue that evolution did not really create information but only transformed preexisting information in the environment, then you are welcome to do so. However, it is no great accomplishment to make this observation. If you are just saying that nature has to be a certain way for evolution to work then you can just assert it as obvious without further argument. You do not need to write lengthy books to defend this claim, or to deploy difficult mathematical theorems in support of it. (p. 205)

Finally, what of Dembski’s assertion that “the information was not always there”? Well, what was “there” was a source of information that had the potential to be modified by organisms into functional structures. That information is found in the collection of possible environmental states, which in turn determine what constitutes fitness for the organisms living in those environments.

Does that preexisting environmental information need to be explained by reference to intelligent design? According to Dembski, that would only follow if environmental information were an instance of complex, specified information. Good luck showing that it is. You will run into all the same problems we have already seen.


I have shown that Dembski has not replied seriously to any of the major points I raised about his work. But I do not think I have really done justice to just how poorly done his review really is. You can start reading at any random point, and you will very quickly come to something stupid.

In Section 4, Dembski writes, “But Rosenhouse’s Darwinism plays to the lowest common denominator. Throughout the book, he hammers on the primacy of natural selection and random variation, entirely omitting such factors as symbiosis, gene transfer, genetic drift, the action of regulatory genes in development, to say nothing of self-organizational processes.”

I do not discuss those things because I was not writing a general treatise on evolutionary biology. The way you could tell is that in the Preface I wrote, “I am not writing a general treatise on evolutionary biology.” I elaborate on this as follows:

Let me be clear that this is a mathematics book that also discusses biology, as opposed to a biology book that also discusses mathematics. Inevitably, there are places where we must get our hands dirty by digging into the biological details, but my central points are mathematical and not biological. Biologists will rightly criticize me for presenting a simplistic version of evolutionary theory. I focus almost entirely on natural selection acting at the level of genes, but everyone understands that there is far more to evolution than this . . . My argument is essentially that even if we take this narrow understanding of evolution as out starting point, we still have more than enough resources to refute any gambit coming from the other side. (p. xiv)

Moving on, in my book I often use the image of “track one” and “track two” mathematics. Track one is our intuitive understanding of abstract objects, while track two is the rigorous technical detail found in textbooks and journal articles. I use this distinction in part as a convenient way of describing different sorts of errors in mathematical anti-evolutionism. Sometimes they argue vaguely at a track one level, but the rigorous details needed to make the argument convincing are just completely lacking. Other times they provide very technical track two arguments, but when you try to understand what is really going on you find that it is all just meaningless gobbledygook. All I am really saying is that a good mathematical argument should be both precise and comprehensible.

I thought I was stating the obvious, but my track one/track two distinction seems to have driven Dembski completely insane. Section 7 of his review consists of ten full paragraphs of bizarre ranting about it, all of it nonsensical. He declares that in making this distinction I am setting myself up as a “math cop, stipulating the rules by which design proponents may use mathematics against Darwinism and for intelligent design.” He then accuses me of not playing by own rules:

Even so, having set the standard, Rosenhouse should meet it. But he doesn’t. For instance, when he describes a standard statistical mechanical set up of gas molecules in a box, he remarks: “We are far more likely than not to find the molecules evenly distributed,” (p. 234) I would agree, but what exactly is the probability space and probability distribution here. . . . But given the weight he puts on statistical mechanics in refuting appeals by creationists to the Second Law of Thermodynamics, once could argue that he had no business confining himself to track 1.

Of course my description of statistical mechanics took place at a track one level. That is because I was simply trying to explain a few basic points to a non-scientific audience. But if instead I was arguing that physicists from Boltzmann onward were hopelessly incompetent at their discipline, and that they clung to their views only out of morbid anti-religious bias, then I think the physicists could reasonably request that some track two details be provided.

And just as another example of Dembski being unable to get even the simplest things right, I should point out that actually I barely discuss statistical mechanics at all. The only reason I even mention it is that certain specific anti-evolutionists bring it up in their own writing. But the fact is that I devote far more space to the classical understanding of the second law than I do to statistical mechanics.

I am not setting myself up as a cop, I am not making arbitrary rules, and I do not just make random declarations about what is track one and what is track two. Dembski just made up all of that. I use the track one/track two distinction simply as a way of organizing broad classes of errors that anti-evolutionists make. When they equate entropy to randomness, or when they model the formation of a protein as a simple problem in discrete probability, then they are making track one arguments that fail as soon as you try to formulate them with track two rigor. When they produce elaborate technical models with copious jargon and notation, but then fail to link up their abstract constructs to anything in reality, they are presenting track two arguments with no track one interpretation.

Finally, I cannot let this one go by. In Section 13 Dembski writes:

What this means is that conservation of information is not tied to uniform probability or equiprobability. Rosenhouse, by contrast, claims that all mathematical intelligent design arguments follow what he calls the Basic Argument from Improbability, which he abbreviates BAI (p. 126). BAI attributes to design proponents the most simple-minded assignment of probabilities (namely uniform probability or equiprobability).
There is a reason Dembski does not quote me actually saying that all mathematical intelligent design arguments follow the pattern of the BAI. He does not even provide an out-of-context sentence fragment to give the false impression that I am claiming such a thing.

My discussion of the BAI occurs in Section 5.5 of my book, in the chapter on probability theory. If I really believed that all mathematical ID arguments fall under the BAI, then one wonders why this chapter went on for another 30 pages after this section was finished. Not to mention that the book goes on for more than 150 pages past this point. I even conclude this section by explicitly contrasting the extreme simple-mindedness of the BAI with the superficially more sophisticated probability arguments of certain ID proponents. Dembski’s characterization of this section is just completely made up.


We have barely scratched the surface of all that is wrong with Dembski’s review. He has not responded effectively to any of the main arguments of my book, and he has made no attempt even to characterize them with any honesty or integrity.

Ultimately, for all his interminable ranting, blather, and self-admiration, the bottom line is this: His notion of complex, specified information is a complete non-starter in the context of evolution. We cannot establish complexity because we cannot even approximate, much less calculate, the probabilities his model requires, and we cannot establish specificity because we have nothing like the background knowledge we would need to distinguish design-suggesting patterns from those that can be explained by natural selection. His arguments about the No Free Lunch/Conservation of Information theorems likewise have no relevance to evolutionary questions. It is not even clear that these theorems apply straightforwardly to biological evolution, but regardless, they imply nothing more than that nature has to be a certain way for evolution to be viable.

Looks like I could have written a shorter book! Dembski does not even lay a finger on any of these points. And he never will be successful in doing so, even if he writes another 18000 words.

Dembski understandably focused on that portion of my book that dealt with his own writing, but this is only a very small portion of the book overall. Have a look for yourself!

Related Posts

Jerry Coyne reviews the book at his site Why Evolution Is True here.

Matt Young reviews the book at Panda's Thumb here.

Joe Felsenstein reviews the book at Panda's Thumb here.

I announced the book and offer some general thoughts about mathematical anti-evolutionism at Panda's Thumb here. I also provided some additional commentary on this subject in an article for Skeptical Inquirer.