Yesterday, I posted the first part of my critical review of Jason Rosenhouse’s 2022 book Failures of Mathematical Anti-Evolutionism, published by Cambridge University Press, in which I demonstrated how he badly misrepresented significant elements of the 1966 Wistar Symposium and how he falsely claimed that the biologists there had adequately answered the challenges posed to the Neo-Darwinian Modern Synthesis by the mathematicians.
Rosenhouse also makes three important assertions in a single paragraph on page 124 in Chapter 4: The Legacy of the Wistar Conference, each of which claims that subsequent work has vindicated the biologists and refuted the mathematicians. However, none of them is accompanied by any evidence within the chapter itself.
Assertion 1: “Molecular biology and computer science were both in a very rudimentary state when the conference was held in 1966. Both fields have blossomed spectacularly in the ensuing decades, with results that have not been kind to the perspectives of Eden and Schützenberger. Molecular biologists have learned a lot about the geometrical structure of protein space, and their results make Eden’s simplistic combinatorial calculations look hopelessly naive.”
Rosenhouse provides no citations. He names no studies nor molecular biologists. No results actually described. No quantifications are offered, just a general statement about how subsequent developments “have not been kind”, whatever that is actually supposed to mean. Rosenhouse does not tell the reader what the geometrical structure of protein space actually looks like, whether functional proteins are in fact densely connected by single-step mutations, or what fraction of single-residue changes preserve function in any specific protein family. He simply asserts that the results exist and implies that those results are devastating for the arguments presented by Eden and Schützenberger, then moves on. The reader is expected to take this on faith and, presumably, to wait for all of this to be shown to him in Chapter 6.
This failure to actually support his assertions with any details at all is particularly egregious because either those results exist or they do not. If molecular biologists have demonstrated that functional proteins are densely connected stepping stones in sequence space, that result would have a name, a date, and a journal. Rosenhouse does not provide any of these. The most relevant empirical work, deep mutational scanning studies, actually shows the opposite of what Rosenhouse implies. Single-residue changes to most proteins tend to reduce or destroy function. The fitness landscape is rugged, not smooth. Eden’s 1966 concerns about the rarity of functional proteins in sequence space have been confirmed by subsequent experimental work, not refuted by it.
Assertion 2: “Schützenberger’s arguments about computer simulations have likewise not held up, and computer simulations of the evolutionary process are now commonplace.”
This is a non sequitur disguised as a refutation. The fact that computer simulations of evolution are “commonplace” does not even begin to address Schützenberger’s argument. Schützenberger’s point was not that you cannot simulate selection on a computer. His point was that random typographic changes to a program, changes that are not guided by a fitness function designed by the programmer, do not produce functional programs. The existence of evolutionary algorithms does not refute this, because every evolutionary algorithm ever written has a programmer-designed fitness function that does exactly what Schützenberger said biology lacks: it provides the mapping between typographic changes and functional outcomes. The programmer is the intelligence that Schützenberger said was missing. Dawkins’s WEASEL program is the perfect example: it works because Dawkins specified the target string, the fitness function, and the selection rule. Remove any of these programmer-supplied components and his simulation produces nothing. Rosenhouse does not address this distinction at all. He simply notes that computer simulations exist and treats their existence as a refutation, as if Falcon 3.0 and Spore somehow proved Schützenberger wrong.
Assertion 3: The section heading “The Two Pillars of Mathematical Anti-Evolutionism” and the final section imply that subsequent chapters will deliver the killing blow. The chapter ends by promising that “We will discuss this work in Chapter 6.”
This is the most revealing move. Rosenhouse has spent an entire chapter arguing that the mathematicians at Wistar were wrong and that the biologists successfully answered them. But his own chapter provides no quantitative refutation of any of the three mathematical arguments provided by the biologists at the time or by anyone else since. Eden’s sequence-space calculation stands unchallenged by any number Rosenhouse provides. Ulam’s rate calculation stands unchallenged by any rate Rosenhouse computes. Schützenberger’s formal-language argument stands unchallenged by any mechanism Rosenhouse identifies. After 26 pages of rhetoric, historical misrepresentation, and posturing, the actual mathematical demonstration is deferred to a later chapter the reader has yet to read.
Which is why, in my third and final post on Rosenhouse’s book, we’ll examine whether he delivered on those three promises or not.
