Chance favors the minute animalcule: John Tyler Bonner on randomness


A colleague recently (well, not that recently; sorry, Art) lent me a copy of John Tyler Bonner’s latest bookRandomness in Evolution. Dr. Bonner is emeritus faculty at Princeton University, where he has been since 1947, shortly after World War II interrupted his Ph.D. studies. Among many other contributions, Bonner was a pioneer in the development of the social amoeba (or cellular slime mold) Dictyostelium discoideum as a model system for multicellular development and cell-cell signaling. A member of the National Academy of Sciences and a fellow of the American Association for the Advancement of Science, he has published over twenty books and mountains of peer-reviewed papers.

As much as David Kirk’s Volvox, Bonner’s books The Evolution of Complexity and First Signals: The Evolution of Multicellular Development influenced my decision to study Volvox in grad school. I had the pleasure of meeting Dr. Bonner in 2009 when, as a graduate student, I invited him to give a departmental seminar at the University of Arizona. It really was a pleasure; this is someone who thinks deeply about big questions and has made important contributions to understanding many of the answers.

The central argument of the new book is that randomness plays a larger role, relative to natural selection, in the morphology of small organisms than that of large ones. Typically of Bonner’s work, the book is coherent, readable, and full of fascinating examples. Although the cellular slime molds are his primary study organism, Bonner has long had an interest in, and interesting things to say about, Volvox, so I was excited to read his most recent thoughts.

The logic that dictates a larger role for randomness in smaller organisms is as follows: in large organisms, development requires a large number of steps. If any of these is disrupted by mutation, especially early in development, the embryo is likely to die. In smaller organisms, by contrast, all developmental steps are near the end of development (since development is short), so mutations are less likely to be fatal and mutant offspring more likely to survive:

These [developmental] steps are interconnected and sequential, building one upon another. In a small organism, with a minimum number of steps, a mutation will almost immediately be at the end of the chain and therefore able to affect the morphology of the whole organism…In a larger organism the interconnected…steps now become more numerous, corresponding with the degree of the increase in size…

Now here is the key point. A mutation could arise at any point in the thicket of connected lines, but if a step were altered through a debilitating mutation, the sequence of steps that would have followed to complete development would be blocked, and the result would be the death of the embryo…Only the mutations that appear at the end of the chain would have a chance of survival.

In larger organisms with complicated developmental programs,

If a mutation adversely affects a step during the course of development, everything stops and the embryo dies.

I can see a couple of problems with this argument. First, it addresses the magnitude of the phenotypic effects of mutations, not the fitness effects of phenotypic changes. We can assume that any given mutation with a phenotypic effect has a fitness effect that lies on a spectrum from strongly positive to fatal (including zero for truly neutral mutations). The magnitude of the phenotypic effect presumably bears some relation to the fitness effect: in most cases, mutations with large phenotypic effects will have large (usually negative) fitness effects. It may be true that early-acting mutations in large organisms generally have larger phenotypic and fitness effects than late-acting mutations, and it may be true that early-acting mutations in large organisms generally have larger phenotypic and fitness effects than typical mutations in microbes. But that is no reason to suppose that the relationship between phenotypic effect and fitness effect is fundamentally different between large and small organisms, as Bonner’s argument seems to require.

We’re interested here in a pair of relationships: timing of mutation to magnitude of phenotypic effect and magnitude of phenotypic effect to magnitude of fitness effect. Bonner’s argument about the length of development affects the first relationship: early-acting mutations have larger phenotypic effects, on average, than late-acting mutations. Let’s stipulate this. For Bonner’s argument to be right, though, selection has to tolerate larger phenotypic effect mutations in small organisms than large, and this is all about the second relationship. Even if all microbe mutations are effectively late-acting, and therefore of small effect, there is no reason given to suppose that selection will better tolerate mutations of similar phenotypic effect in small organisms than in large ones.

More critically, Bonner’s argument seems to contradict a well-established principle of population genetics, namely that the efficacy of natural selection increases with population size, because genetic drift overpowers natural selection in small populations. We know that small organisms tend to have larger populations than big ones (think E. coli versus elephants). This means that natural selection of a given strength is generally more effective in microbes. Bonner does acknowledge this:

For drift to lead to a genetic change in a population, the population size needs to be small. It is a well-known fact that the population density of an organism is inversely related to its size. For instance, in Africa there are many fewer elephants than small rodents per square mile, and there are infinitely more microorganisms than rodents in the same space. This contradicts my argument that random changes are more abundant in small organisms than in large ones. However, the generation of novel gene arrangements in a population that result from drift is no doubt very small compared to the generation of novelty by mutation and the absence or reduction of selection found in small organisms.

Leaving aside that drift is determined by population size, not density, the first part is solid. But the last sentence seems to conflate the generation of mutations with their fixation. All novelty is ultimately generated by mutation. Selection and drift determine which mutations go to fixation or extinction. What I think Bonner means here is that drift in large organisms is no match for the reduced efficacy of selection in microbes (which allows multiple phenotypes to coexist), but this has not been established.

Whether or not Bonner’s thesis is right is an empirical question, but he is pessimistic about testing it:

In a crude sort of way, small organisms are more likely to be involved in randomness than large ones. Because internal selection will play little or no role in their short development, they will produce adult morphological variants in large numbers, and there will be an increased chance that some of them are untouched by natural selection. The big problem is that their neutrality cannot be proved, a difficulty that ignites the passions of many committed adaptationists. [emphasis added]

Strictly speaking, he’s right: neutrality, defined as the absence of selection, can’t be proved, simply because you can’t prove a negative. But it can be tested. I can see no reason, in principle, that an experimenter couldn’t track the survival of a large number of putatively neutral phenotypes and compare their representation in later generations to a null distribution. As Bonner points out, such an experiment would not settle the question if we can’t distinguish lack of selection from stabilizing selection:

Or if an organism remains unchanged morphologically through successive generations without being altered by selection, it might possibly not change because it is neutral. Neutral morphologies must be distinguished from unchanging morphologies due to stabilizing selection, where deviations from the norm are selected against: the morphological constancy over many generations is not a matter of chance, but of continuous selection.

However, I would argue that ‘morphological constancy over many generations’ is evidence of stabilizing selection as opposed to drift. The mathematics of population genetics is unforgiving on this point: drift in any finite population is a certainty, so the frequency of truly neutral variants cannot help but change over time. In the kind of population that Bonner envisions, in which a large population of microbes consists of coexisting, equally fit phenotypes, the frequencies of those phenotypes are certain to drift.

Bonner’s central thesis, that randomness affects the morphology of small organisms more than that of large ones, could still be right. If so, the relationship between phenotypic effect and fitness effect must be fundamentally different between large and small organisms. Bonner seems to be suggesting something like this in his examples of radiolaria (pictured below), foraminifera, and diatoms, which do indeed have a staggering array of morphologies:

So here are three unrelated groups of organisms that have independently shown an incredible variety of forms that not only exist today, but many have remained unchanged, or little changed, for millions of years. It is difficult for me to imagine that each of those many, many thousands of species, each with its distinctive sculptured shell, is maintained by a specific act of natural selection.


Plate 117 from Haeckel 1887, which is reproduced as Bonner’s Figure 4a.

If he is right, this book will be considered groundbreaking, and I’ll feel a fool. As Dan McShea is quoted as saying on the dust cover,

He could be right. But right or wrong, this claim will be hugely controversial. It doesn’t just ‘approach heresy,’ as he puts it. It is heresy.


Stable links, or the best I could find:

Bonner JT (1988) The Evolution of Complexity. Princeton University Press, Princeton, NJ.

Bonner JT (2000) First Signals: The Evolution of Multicellular Development. Princeton University Press, Princeton, NJ.

Bonner JT (2013) Randomness in Evolution. Princeton University Press, Princeton, NJ.

Haeckel E (1887) Report on the SCIENTIFIC RESULTS of the Voyage of H.M.S. Challenger during the years 1873-1876. Published by order of Her Majesty’s Government.

Kirk DL (1998) Volvox: Molecular-Genetic Origins of Multicellularity. Cambridge University Press, Cambridge.


  1. grahamjones says

    Interesting post. My first reaction is that you’re right and he’s wrong. I like what you said about population genetics. But there seem so many confounding factors, I don’t know what he’s actually claiming.

    If his argument is about developmental complexity, why does he keep talking about size? What are his predictions for nematode worms vs fruit flies? Or elephants vs mice?

    How do we compare the degree of phenotypic differences for different organisms? Suppose a fly gets three hairs instead of four on its legs. What is a similar amount of change in an ameoba or a mouse?

    How do generation times and mutation rates interact with his hypothesis?

    What about developmental plasticity? It may be that organisms with many stages in their development are able to compensate for a small ‘error’ early on, meaning the mutation is effectively neutral.

    On testing his idea (and a minor disagreement with you). You can never prove exact neutrality. It’s not because it is a negative, but the exactness that makes it impossible. But that’s not important. You can demonstrate effective neutrality which is all anyone should care about. A more careful statement of his idea would mean specifying something about the distribution of fitness effects.

    Graham Jones

Leave a Reply

Your email address will not be published. Required fields are marked *