Contemporary Perspectives on the Meaning, Roles, and Implications of Chance in Evolution

1 Introduction and General Remarks

Since Darwin’s time, and especially since the modern synthesis in the first half of the twentieth century, researchers have recognized that chance plays some role in evolution. Just which role, what its relative importance is as an evolutionary factor and what its presence implies for our broader conception of the natural world have been matters of extensive empirical and conceptual debate.

To give just a few examples of the places where chance and contingency appear in evolutionary theory, consider the following: (1) Sometimes natural disasters occur (e.g., an earthquake, a forest fire) for which none of the organisms in a population is better adapted than any others. In those cases, the subset of organisms that survive and reproduce will be a random sample from the original population (this would be a case of drift); (2) On the other hand, mutations are also said to be random. This sometimes means that they are not biased towards genes or traits that confer an advantage to the organisms that possess them (in fact, most mutations are either neutral or deleterious for their possessors). It can also mean that they are random in a deeper sense (indeterministic, unpredictable, etc.); (3) Evolution on a more macro scale is sometimes said to be contingent. This view was popularized by Stephen Jay Gould, who posed that nothing in the process of evolution necessarily entails that it will end up with forms resembling the ones that currently exist. On the contrary, since contingencies and historicity shape diversity, if one were to “replay the tape of life”, the biosphere would look very different from what it does now.

In all these areas, debate has ensued. Again, to briefly illustrate: (1) Sometimes some organisms in the population do have a selective advantage over others, but by chance, they end up having lower reproductive success than those others. Should this count as a case of selection or drift (or both)? Should drift be thought of as a process or an outcome? Should it count as a cause? (2) Studies in the field of Evo-Devo now suggest mutations are biased towards some specific regions of the genotypic and phenotypic space. Can mutations still be called random? Are the constraints imposed by the availability of mutations important for shaping evolutionary trajectories? (3) The tree of life shows extensive convergence (i.e., acquisition of the same traits or genes in different lineages independently). Does this mean that evolution is not really as contingent as Gould claimed? How can we measure the degree to which contingency affects evolutionary outcomes?

Chance in Evolution, edited by Grant Ramsey and Charles Pence, is a collection of contributions centered on the meaning and import of chance in evolutionary theory. It contains chapters focused on the problems mentioned above (among many others) and does so from a rich diversity of perspectives, ranging from the philosophical, historical, and empirical to even the theological points of view. This stems from the fact that the contributors come from different disciplines (philosophy, history, biology, theology). This is sometimes noticeable in their proposed hypotheses, the methodologies used to examine them (from argumentation to experimentation to exegetical work) and the cited sources (from Aristotle to Darwin, the synthesis biologists, and contemporary biologists and philosophers). The book is thus an example of multidisciplinary work on a subject, with a breadth of perspectives that is not very usual to find. In this respect, it deserves commendation.

Overall, the book offers a comprehensive guide to anyone interested in the diverse outlooks from which experts are studying the issue of chance in evolution today. In consequence, I believe the book (in its entirety or parts of it) can be of interest to a wide variety of readers. Some contributions are more didactical and accessible to a wide audience (such as chapters 3, 4, and 5), while others are more technical and will be of interest mainly to specialists (e.g., chapters 6, 8, and 11). None of the chapters is totally introductory to their subject (except, perhaps, chapter 4), but educators can probably use most as secondary readings if they wish to dig deeper into some topic, especially at the university level.

The book is divided into three parts. In the remainder of the review, I give a brief overview of the contents of each part, and then offer some final thoughts.

2 Part One—The Historical Development and Implications of Chance in Evolution

The first part contains five chapters that are more historical in their outlook. All of them examine the role that some historical figure(s) attributed to chance in evolution and draw some conceptual conclusions from this analysis.

2.1 Chapter 1—Contingency, Chance, and Randomness in Ancient, Medieval, and Modern Biology (Depew)

The first chapter, by David Depew, has the ambitious goal of tracing the conceptualizations of biological contingency, chance, and randomness from antiquity to present times. By doing that, the author hopes to erase a simplistic narrative, according to which the pre-Darwinian worldview was typological, essentialist, fixist, determinist, etc. in a strong sense; that is, in which there was absolutely no place for contingency and chance. Setting this narrative aside, in turn, would help us gain a better understanding of what the Darwinian revolution, and the probabilistic revolution after it, implied.

Depew begins with Aristotle’s discussion of Empedocles’ proto-selection theory, according to which (in a previous era of the cosmos) parts self-assembled into organisms, and only those whose combinations of parts were stable remained. Aristotle objected, according to Depew, that organisms are not mere aggregates of separate parts, but integrated wholes, which come about by the progressive differentiation of an initially indeterminate matter (a process later called epigenesis). This process, even if it has a final cause (télos), and is therefore not random, is still contingent, since (for example) defects, early death and spontaneous abortion might occur.

In contrast, Depew argues that it was only in the Enlightenment (with authors such as John Ray) when essentialism and fixism in a stronger sense emerged, alongside preformationism.¹ Depew views preformationism as a determinist attempt to replace Aristotelian self-creative and goal-oriented nature with a more mechanistic worldview. It was during this time, according to the author, that necessity-purposiveness on the one hand, and contingency-chance on the other, were woven together.

This weaving, in turn, would explain some bad readings of the Origin, both in Darwin’s time and in the present (e.g., natural selection equals chance, because it is contingent). Additionally, according to Depew, Darwin’s genius consisted in showing how contingency is compatible with goal-orientedness. Moreover, natural selection (a contingent process) explains “the purposive characteristics of organisms”. That is, natural selection can generate adaptations with chancy inputs (variations that arise independently of the advantages or disadvantages they provide to the organisms, and a connection between fitness and reproductive success that is only probabilistic).

2.2 Chapter 2—Chance and Chances in Darwin’s Early Theorizing and in Darwinian Theory Today (Hodge)

In chapter 2, Jonathan Hodge aims to defend a (very particular) brand of causalist take on natural selection. Or at least to show that it has greater historical support than its alternatives (mainly statisticalism, but also other variants of causalism).² To do this, Hodge divides his chapter into two parts.

The first part traces at length Darwin’s conceptualizations of chance throughout his academic career (from their pre-Origin beginnings to his mature views). His conclusions from this section are two. Firstly, that Darwin’s views of chance were not influenced by the probabilistic revolution going on at the time. Instead, as was more common in his time, he held the ignorance-view of chance.³ Second, and more importantly, that (the mature) Darwin invoked chances in different places: in the origin of variations (which are chancy in the sense of fortuitous/accidental/with unknown causes) and in their fate (organisms that have advantageous variations have only an increased chance of having a greater differential reproductive success). Here, Darwin invoked chance in the sense of probabilistic causation, not of fortuitousness. That is, differential reproductive success is not due to sheer luck or accident; it is (probabilistically) caused by a relation between the traits the organisms possess and the environment in which they live.

Hodge uses this last point to argue for a causalist view of natural selection in the second part of the article. He sees a continuity between Darwin’s view of the relation between traits and reproductive success as probabilistic causation, and contemporary explanations of differential reproductive success. For example, according to him, what today allows us to differentiate between cases of selection and drift is the fact that the traits of the organisms in question are probabilistically causally relevant to their reproductive success (and not anything related to sampling, sampling errors, etc.). The main lesson he wants us to draw is that mathematical representations of natural selection are different from natural selection itself, and that the former did not replace the latter. The synthesis architects, Hodge argues, saw this very clearly. They viewed their mathematical results as offering statistical analyses, not replacements, of causation. Thus, statisticalists may argue that fitness does not play a causal role in the mathematical apparatus of population genetics, but this does not mean that natural selection itself is not causal.

Lastly, and polemically, Hodge warns other causalists (and non-causalists) against what he sees as confusing modes of speech: (i) force talk, (ii) fitness talk, and (iii) principle of natural selection talk. Particularly surprising are his claims that natural selection can “get along just fine” without the concept of fitness (p. 66), which only plays a role in the mathematical apparatus, and thus should not be employed in introductory texts.

2.3 Chapter 3—Chance in the Modern Synthesis (Plutynski, Vernon, Matthews, Molter)

The third chapter is written by Anya Plutynski, Kenneth Blake Vernon, Lucas John Matthews, and Daniel Molter. It aims to establish and document the uses of chance (and their cognates, such as random and accidental) in some of the representative authors and texts of the modern synthesis.

The chapter begins with a brief and useful introduction describing what the modern synthesis was, from both a conceptual and an institutional point of view. The authors then move on to distinguish five possible senses of chance, which are to be traced through the synthesis authors in the following sections. Chance can thus be viewed as (i) metaphysical indeterminism, (ii) randomness, (iii) a proxy for probability, (iv) contingent and rare (climatic/ecological) events, and (v) factors that oppose selection.

They then examine the uses of chance by three early synthesis authors (Fisher, Haldane, and Wright) and four later synthesis authors (Dobzhansky, Mayr, Simpson, and Stebbins). From the first three, they show that Haldane ascribed a lesser role to chance, while Wright gave it the (empirically) greatest one. In the latter’s “shifting balance theory,” chance played role in fixing some mutations in temporally isolated subpopulations. When barriers to isolation broke, this would allow the population to move to new peaks in the adaptive landscape. Fisher, on the other hand, mentioned chance most substantially in connection to sense (i). He argued in favor of a new scientific worldview where genuine indeterminism took place, but where explanations were still possible at the level of aggregates and populations.

From the later authors, it is notable that Dobzhansky ascribed less of a role to chance in the third than in the first edition of Genetics and the Origin of Species. Stebbins, in the latest work of all those examined, also took drift to have a relatively lower importance than selection in general. Plutynski et al. see these two cases as confirming what Gould had called a “hardening” of the synthesis (a de-emphasizing of the importance of drift and an emphasizing of selection). On the other hand, Mayr, who was concerned with arguing in favor of the allopatric (vs. sympatric) model of speciation,⁴ put some emphasis in chance in sense (iv), in the formation “by chance” of geographical isolation barriers. Lastly, Simpson saw population size as affecting both tempo (rate of change per unit time) and mode (level at which change takes place) of evolution.

The chapter concludes by stating that, despite the differences mentioned above, there were large agreements about the roles of chance in evolution. For example, all agreed that mutation is by chance (not directed towards adaptation), that gamete sampling is a source of chance in sense (ii), that contingent events (such as the formation of isolating barriers, sense (iv)) play a role in macroevolution, etc. The authors also caution (wisely in my opinion) against drawing easy and immediate conclusions from their historiographical work to current philosophical debates, since the synthesis authors did not have those debates in mind, nor did they frame their positions thinking about them.

2.4 Chapter 4—Is It Providential, by Chance? Christian Objections to the Role of Chance in Darwinian Evolution (Ashley)

Chapter 4 is the odd one of the book. It is written by a theologist (J. Matthew Ashley). This author does not attempt to clarify the meaning or roles of chance within evolutionary theory (he leaves that to the rest of the contributors). Instead, what he wants to diagnose is the parts of Christian theology with which chance in evolution has been taken to be in conflict, and to propose a way to alleviate those tensions. Now, I have to say that, as an atheist, I have very little interest for matters of Christian (or any other religion) theology. However, as a believer in science, I believe it is smart to understand the reasoning of those who publicly oppose evolution, and to support those who, within the religious camp, endorse it. For this reason, I found the article interesting and instructive.

Ashley proceeds by considering two case studies, which are two theologians writing more than a century apart and coming from different Christian traditions. On the one hand, he examines an 1874 review of Darwin’s Origin by the Presbyterian Charles Hodge, and on the other, he looks into a 2005 op-ed piece in the New York Times by the Catholic Cristoph Schönborn (as well as a 2007 book, where he expands on those ideas and responds to critics). Despite their differences, the author finds that these two theologians echo similar concerns and find problems with the same two parts of theology.

Ashley shows that both encountered tensions between chance in evolution and (i) purposiveness or goal-orientedness in nature (which is part of the argument from design from natural theology) and (ii) the central claim of Christianity that God is intimately involved with (creates, cares about, maintains and governs) his creation, both as a whole and in its particulars. He suggests that they did not adopt any available solutions to any of these two points of tension because they thought that it would raise problems with the other.

According to Ashley, this is because, deep down, both theologians were committed to “a clear and transparent reading of God’s purposes and providential guidance in natural and historical processes” (p. 113). In contrast, he proposes to understand God’s providential governance of the world in a less transparent way, such that it becomes compatible with “the apparent messiness and contingency of history” (which would include chance events in evolutionary history). The analogy would be between providence and a narrative, where the authors’ intended plot is realized through the messy and contingent actions of the protagonists.

2.5 Chapter 5—Does Darwinian Evolution Mean We Are Here by Chance? (Ruse)

The fifth chapter, by Michael Ruse, continues on related subjects, albeit from a different perspective. Ruse seeks to explore the relation between evolutionary biology and human uniqueness, to see if modern biology can provide us with what Christianity gave us before.

He begins with a section comparing Christian thought with pre-Darwinian evolutionary accounts. A central difference between them, Ruse tells us, was that evolutionary thinking at the time was tied with doctrines of social progress, while Christian thinkers tended to be more politically conservative. But beyond this difference was a fundamental agreement: that we humans are somehow special. In the evolutionary accounts of the time, we humans were the end point to which evolution had always pointed, and this natural progression reflected the social progress which we, unaided by God, were capable of achieving (through education, science, politics, etc.).

“Humans were special. Then came Charles Darwin, who spoiled everything” (p. 125). Darwin was a firm believer in social progress.⁵ However, his theories “spoiled” things in various ways. Firstly, Darwin argued for them without appealing to social progress itself (as his predecessors had done), thus showing that it was logically possible to be an evolutionist without being a progressist. Secondly, to him, traits could be said to be adaptive only with respect to a given environment. That is, a trait (such as high intelligence) could be adaptive in some environments but neutral, or even detrimental in others. Therefore, there is no absolute notion of perfection/adaptation which humans could claim for themselves. And thirdly, Darwin saw variations as arising randomly (i.e., not according to the needs of their possessors), which makes evolution seem less oriented to achieve some predetermined goal, such as the existence of humans.⁶

The article then moves on to discuss other, more recent, accounts that have tried to make sense of the idea of progress and human uniqueness. Some see progress in natural selection when later forms beat out earlier ones. Julian Huxley appealed to arms races: as a predator or prey makes a move “towards greater independence” from the other, the other will either go extinct or respond with another move, achieving “a higher plane of independence” (p. 134). Others appealed to the presumed superiority of some ecological niches over others, which would ground the superiority of the organisms that occupy them. Since organisms will tend to diversify and occupy all available niches in the ecosystem, progress would ensue as more and better niches become occupied.⁷ A third group appeals to the increase in complexity, differentiation, specialization, etc. (either by selective or non-selective mechanisms) to characterize progress.⁸

These themes are also mixed (somewhat confusingly, in my opinion) with discussions over the contingency or directionality of the evolutionary process and the inevitability of the appearance of humans (or human-like, intelligent) creatures.

3 Part Two—Chance in the Processes of Evolution

The second part of the book has three chapters, more centered on conceptual issues and the study of microevolutionary processes (selection, drift, and mutation).

3.1 Chapter 6—The Reference Class Problem in Evolutionary Biology: Distinguishing Selection from Drift (Strevens)

Chapter 6, by Michael Strevens, is an attempt to solve the reference class problem (RCP), applied to the probabilities used in evolutionary explanations. The RCP is the problem of distinguishing the factors that are taken into account when measuring a probability from those that are not (the parameters from the variables, in technical jargon). For example, when spinning a wheel of fortune that is painted half black and half red, if detailed factors like the initial translational and angular velocity are taken into account (are considered as parameters), then, at each particular spin, the probability of red would equal either 1 or 0 (and conversely, 0 or 1 for black). If they are not taken into account (are considered as variables), then the probability equals 0.5.

Strevens shows that a solution of the RCP in the context of evolutionary explanations is important for distinguishing cases of natural selection from cases of drift. In the classical account, drift is conceptualized as the sampling error (i.e., the deviation of actual outcomes from expected ones) of some evolutionary sampling process (e.g., parental or gamete sampling). For instance, if two genetic variants in a population have the same probability of being eaten by a predator but, “by accident,” one of them encounters more predators and is eaten more than the other, then its decrease in frequency is said to be due to drift. But this, in turn, depends on the previous estimation of the probability of being eaten (from which the expected frequencies were derived). If factors like the initial locations of the predators and prey had been taken into account (as parameters instead of variables) then perhaps the probabilities would not have been equal, and thus the decrease in frequency would not be said to be due to drift. Therefore, the choice of parameters influences whether we attribute an evolutionary outcome to selection or drift.

The authors’ solution to the evolutionary RCP consists in counting as parameters only those that are explanatorily relevant to the outcomes in question. This, of course, requires an account of explanatory relevance that is independent of the distinction between parameters and variables itself. The key to finding this is to note that the outcomes in question are the changes in frequencies (of genotypes or phenotypes) of the population, not the fates of particular individuals within it. So, the explanatorily relevant facts (the ones that determine the probabilities) will be those that make a difference to the frequencies.

These, according to Strevens, are properties like (what he calls) the microconstancy and the smoothness of the initial condition distribution. To illustrate briefly, the microconstancy of the wheel of fortune refers to the fact that a function plotting the outcomes for small changes in the initial conditions (for example, in the initial velocities) results in a distribution where half of the results are red, and half are black. The smoothness of the initial condition distribution refers to the fact that people do not tend to throw the wheel with velocities that are within the micro-regions that result in red outcomes more often than they do so for black regions. According to the author, these two facts suffice to explain the long-term frequencies of red and black (i.e., the fact that they tend to equal the proportions of the wheel painted in that color). The velocity of any particular throw of the wheel (or the particular locations of the organisms subject to predation) are thus not explanatorily relevant facts, since they add nothing to an explanation that contains only the two properties mentioned above.

3.2 Chapter 7—Weak Randomness at the Origin of Biological Variation: the Case of Genetic Mutations (Merlin)

Chapter 7, by Francesca Merlin, analyzes the proper way to characterize randomness in genetic mutations. She does this not from the point of view of them being (non-)biased towards adaptation (what she calls the “evolutionary perspective”). Instead, the author looks at genetic mutations independently from their phenotypic effects and thus, their evolutionary consequences. Merlin also does not discuss the issue from a metaphysical perspective (e.g., determinism vs “pure” indeterminism), focusing instead on a more metatheoretical outlook (i.e., looking at the models used to describe them, not looking at the structure of the real world).

From this perspective, she first distinguishes between a chance (stochastic) process and a chance outcome. The first are processes that can be modeled probabilistically, while the second refers to series of outcomes characterized by irregularity, unpredictability, etc. (what the mathematicians call random). One can also speak of random outcomes (especially of unique outcomes) as those that are the result of stochastic processes.

After this, a distinction between strong and weak randomness for processes is made. Strongly random processes are those that (i) are indiscriminate sampling processes (the probability of being sampled is the same for all elements) and (ii) are invariant over time (sampling is with replacement, independent of other events).⁹ A process is weakly random if at least one of the conditions above fails. Genetic mutations satisfy condition (i) if the probability of a mutation is independent of the site of the nucleotides. They satisfy condition (ii) if that probability is independent of the changes in the physico-chemical conditions inside and outside the cell during the time period studied.

Merlin argues that genetic mutations should be considered to be the result of only weakly random processes. To argue for this, she describes various biases that have been known for a long time (e.g., the existence of hotspots of mutation) and other more recently discovered ones (populations with a temporary increase of individuals with two or more mutations), which violate assumptions (i) and (ii). The author also shows that the model that molecular geneticists use to estimate mutation rates (the Lea-Coulson model) assumes that mutations are strongly random. Surely, every model employs some idealizations, but in this particular case, according to Merlin, the impact of these idealizations in the predictions obtained from evolutionary models using the mutation rates obtained from the Lea-Coulson model have not been studied. She urges for empirical research on the subject, as well as for the formulation of better, less idealized models (better methods exist today to develop them).

3.3 Chapter 8—Parallel Evolution: What Does It (Not) Tell Us and Why Is It (Still) Interesting? (Lenormand, Chevin, Bataillon)

Chapter 8, by Thomas Lenormand, Luis-Miguel Chevin and Thomas Bataillon, aims to contrast selectionist and mutationist views on evolution, within the field of parallel evolution. The authors characterize the selectionist view as one where variation is abundant in all directions (not limiting) and selection is the main driver of evolution. Contrarily, the mutationist view holds that identifying the selection pressures is not enough to understand evolution, because variation may be available only in some specific regions of the phenotypic space. The first view is associated with the modern synthesis, and the second with the field of Evo-Devo.

Parallelism (“the repeated occurrence of similar phenotypic or genotypic features in independently evolving populations”, p. 196) is important in this discussion because many have understood it to be explainable by both views. Moreover, parallelism bears on the issue of contingency in evolution (see the next section); so, the authors hope that shedding light into the patterns, causes and mechanisms that explain parallelism can have some impact on that debate (the connection with the book’s topic is a bit thin, though).

The chapter has sections on patterns, mechanisms, and models of parallel evolution. The discussion of patterns centers mostly on the levels at which it can be observed (genetic, protein and cell function, macroscopic phenotype, ecological function, etc.). The authors note that parallelism may occur at one level but not at levels below it, because the mapping between the two levels may be many-to-one. For example, parallelism in ecological function is compatible with different traits at its base (the same goes with phenotypic parallelism with genetic differences). Regarding the mechanisms, the authors distinguish between phenotypic and genotypic parallelism. On the one hand, phenotypic parallelism is typically interpreted as being the result of selection (exposure to similar environmental pressures), although they emphasize that constraints could have a role as well. On the other hand, genotypic parallelism is most often seen as the result of constrained variation (e.g., the existence of limited building blocks, mutational biases). Lastly, in the section on models, the authors present their own model of mutational parallelism (which is based on a 2010 model by Chevin, Martin, and Lenormand), and draw some heuristic consequences from it. They also discuss some phenotypic models of parallel evolution.

4 Part Three—Chance and Contingency in the History of Life

Finally, the third part of the book contains four contributions. All of them center around the issue of contingency in evolution.

4.1 Chapter 9—Contingent Evolution: Not by Chance Alone (Desjardins)

In chapter 9, Eric Desjardins aims to characterize what it means for evolution to be a historical process, and to show how this differs from the assertion that it is (merely) chancy. To do this, he appeals to the notion of path dependence.

Path dependence relies on a branching conception of history. A process will be chancy if, from every event, at least two different events may follow, each with some positive, non-negligible probability. Actual history will take some specific path (sequence of events) of the branching tree. Additionally, chance processes will be path-dependent if following different paths affects the probability of reaching the diverse outcomes. For instance, consider an initial situation S₀ from which two events E₁ and E₂ can follow. Additionally, let it be the case that, from both E₁ and E₂, two events E₃ and E₄ are reachable. If the probability of reaching E₃ and E₄ from both E₁ and E₂ is the same, then the process will be chancy but not path-dependent (both final outcomes will have the same probability regardless of the path followed thus far). Alternatively, if the probabilities of reaching E₃ and E₄ are different if one is in E₁ than if one is in E₂, then the process will be path-dependent.

The author then moves on to examine the ways in which one could empirically distinguish between mere randomness and historicity (path-dependence). The issue is especially difficult for several reasons. Most obviously, because we cannot observe most species during evolutionary significant times. But also, because even where we can do this (with certain bacteria, for example), the experimental design needed to decide between these two hypotheses would be rather complicated. For example, if we could somehow do Gould’s experiment and “replay the tape of life” from a given point, the “observational” consequences from both hypotheses would be the same—Gould’s thought experiment is designed to distinguish contingency from necessity/determinism, not randomness from path-dependence. Fortunately, some studies in experimental evolution allow us to do the latter. For example, Desjardins mentions the Long-Term Experimental Evolution (LTEE) study by Richard Lenski and his collaborators—described in greater detail in the next chapter of the book. This experiment has two characteristics that Gould’s does not. First, the tape is replayed many times (there are 12 clonal populations evolving independently). And second, samples are taken every 500 generations. Thus, it becomes possible to know the paths followed by each population, and even to replay the tape again from a specific point in the path. Both of these facts allow us to estimate the branching probabilities, and therefore, to distinguish chance from path-dependence. Desjardins then reviews a case within the LTEE (the acquisition of the ability to metabolize citrate in one of the populations) where historicity was most likely at play.

In the final section, the author identifies “generative entrenchment” (a notion developed by Wimsatt) as a source of path dependence in macroevolution. Generatively entrenched features are those that arise early in development, and thus have cascading consequences for the rest of it. In that way they become “entrenched” and act as phylogenetic constraints. The idea is that the different prior histories (paths) taken by different species will form different generatively entrenched traits, and thus lead to different probabilities of following other paths from there on.

4.2 Chapter 10—History’s Windings in a Flask: Microbial Experiments into Evolutionary Contingency (Blount)

Chapter 10, by Zachary D. Blount, continues on some of the topics from the chapter preceding it. Its goal is to review in detail the kinds of microbial experiments used to test various hypotheses of contingency in evolution. After an introduction, the author explains the reasons for using microbes for these purposes (short generation times, high population sizes that can be bred in small volumes, asexual reproduction which permits them to be cloned to start parallel evolution experiments, etc.)

He then divides microbial evolutionary-contingency experiments into two kinds, which he calls Parallel Replay experiments (PREs) and Historical Difference experiments (HDEs). In PRE’s, clonal populations are founded and evolved under identical conditions. The objective is to check the amount of parallelism and divergence that emerges, in order to discriminate the importance of contingency and determinism in place. On the other hand, HDEs attempt to examine the impact of different prior histories between populations on their subsequent evolution. They typically have a two-step design. In the first step, clonal populations evolve in parallel to generate a different history for each. In the second step, the populations evolve under new conditions. The researchers then measure the impact that first-step differences had in the evolutionary trajectories taken in step 2.

As an example of a PRE, Blount reviews the LTEE by Lenski (his doctoral advisor) in some detail. After describing the model organism (E. coli) and the experimental setup, the author examines the results. He concludes that it reveals extensive parallelism (e.g., similar curves of fitness, growth rates, lag phases, cell and population sizes, among many others) with some divergencies (e.g., sets of mutations fixed, mutation rates). He also characterizes these results as typical of other PREs. Among the differences observed, Blount describes in detail two major ones. The first is a population that evolved a long-term polymorphism, with two monophyletic cell lineages (with markedly different cell sizes) coexisting for more than 50,000 generations. The other is the aforementioned acquisition of an ability to metabolize citrate by another population.

As an example of an HDE, Blount looks into a 1995 experiment by Travisano et al. In this experiment, 12 populations were founded from each of the LTEE’s replicates, after 2000 generations of evolution under a glucose environment (step 1). They were then introduced into an environment with maltose for nourishment and evolved for 1000 more generations (step 2). The results were that, for traits that are subject to strong selection, adaptation tended to swamp history’s effects, while history left its mark in selectively neutral traits. He concludes the section on HDEs by analyzing some of their limitations (possibly insufficient history in phase 1, differences reflecting contingency in phase 2 rather than differences due to history) and some possible solutions to these problems, which should be taken into account in the design of future experiments.

In a final section, Blount draws some conclusions and implications from the discussion above. Most notable are the connections between the discussion of contingency and the concept of evolvability, and a call to further interdisciplinary work among scientists of different fields, as well as with philosophers, in order to advance the field (something the book itself does rather well).

4.3 Chapter 11—Rolling the Dice Twice: Evolving Reconstructed Ancient Proteins in Extant Organisms (Kacar)

The eleventh chapter, written by Betul Kacar, reports some very recent findings in the field of experimental evolution and draws some potential consequences from them. These advances consist in a mixture of ancestral sequence reconstruction with experimental evolution. In the experiment they performed, Kacar and her team took an inferred past state of a bacterial protein and genetically engineered an extant organism to exhibit it. The reconstruction of the past state of a character (morphological or molecular) dates back to, at least, the beginning of cladistics. To briefly illustrate how this works and allow the reader to better understand what Kacar and her team did, let me introduce a brief and simple example. Say we know that the diversification pattern between three species, call them S₁, S₂, and S₃, was as in Fig. 1.

Fig. 1

Example branching pattern between three species

Suppose that, additionally, we knew that for a given character C, S₂ possesses state c₂, while S₁ and S₃ both exhibit c₁. One could infer from this that the ancestor of these three species possessed the state c₁ for the character C, since this hypothesis makes us postulate only one evolutionary transformation, against two of the rival hypothesis (c₂)¹⁰—see Fig. 2.

Fig. 2

Same as Fig. 1 with character C superimposed. At the left, the hypothesized character state for the ancestor is c₁, at the right it is c₂. a The preferred hypothesis because it forces us to postulate only one evolutionary transformation event, against two for b

In this chapter, Kacar describes a 2012 experiment where she took an inferred reconstruction of a 500 million-year-old amino acid sequence specifying an Elongation Factor-Tu protein (EF-Tu for short) and genetically engineered modern E. coli bacteria to exhibit it (unlike in the simple example above, the past state was rather different from those exhibited by extant bacteria). According to the author, this experiment was the first to produce a viable organism exhibiting an old protein (although the E. coli exhibiting the old EF-Tu had markedly lower fitness).

While fascinating on its own, this, by itself, has no immediate consequences for the issue of contingency in evolution. Kacar describes some additional experiments that she intends to do, which could yield some interesting consequences for that topic. They mainly involve letting the newly engineered bacteria evolve and tracking the changes that occur in the EF-Tu protein. She characterizes this as “replaying the tape” for a particular protein instead of for the whole organism. More precisely, and acknowledging the limitations that this type of study would have, she calls this making “a mix-tape,” since the ancient protein would be evolving in a modern cellular environment.

Since these kinds of experiments have not been carried out yet, not much can be said a priori about their import to the issue of contingency. But, as an example, the author specifies some possible scenarios that could be stimulating. For instance, it would be interesting if going further back in time meant that more adaptive zones would be accessible (reaching different adaptive peaks).

4.4 Chapter 12—Wonderful Life Revisited: Chance and Contingency in the Ediacaran-Cambrian Radiation (Erwin)

In the final chapter, Douglas H. Erwin re-examines Gould’s arguments from Wonderful Life (1989) about contingency and the fossils of the Burgess Shale, to see how well they hold up against more recent findings. Within Gould’s text, the author identifies three arguments (which could also be seen as parts of one large argument) for the importance of contingency in the history of life. The first is that, rather than a “cone of increasing diversity” model (which, according to Erwin’s Gould, was the most widespread view), the fossils of the Burgess Shale show a rapid increase in morphological diversity, followed by the elimination of many lineages and diversification of the remaining ones. The extent of the diversity in morphology was such that, according to Gould, many (15 to 20) of the organisms had to be classified in separate phyla. Secondly, that selection did not play a prominent role in determining which clades made it past extinction and which did not. For instance, the fossil Pikaia, which was (at the time) the earliest known chordate, was not very abundant within the fossils of the Burgess Shale. If the abundance of a fossil can be taken as a measure of the likelihood of its extinction, our own ancestors did not have very good chances in the Cambrian era. Lastly, from the last two points, it follows that should we “replay the tape”, things could turn out very differently (i.e., contingency is a fundamental feature of macroevolution).

Erwin then reviews various empirical and conceptual advances that took place since Gould’s 1989 publication, which show that his arguments did not hold very well. Among the empirical discoveries is the finding of another Cambrian fossil site at Chengjiang, China. The site had already been discovered in 1984, but most publications of the new taxa took place during the 1990s. At this second site, for example, chordates are much more abundantly represented.

Among the theoretical advances, two are most significant. The first is the establishment of cladistic methods of phylogenetic reconstruction and criteria of taxonomic grouping.¹¹ Gould’s claim that many of the “strange” fossils of Burgess belonged to their own phyla was based on traditional, morphological criteria of classification. However, cladistic methods showed that many of these fossils were in fact stem groups of extant clades (early extinct representatives of the clade, which branched before the ancestor of the living members of it), and thus should not be grouped in separate phyla. The second is the development of molecular clock techniques, which showed that many Metazoan lineages had been diverging for long before the Cambrian explosion (despite the fact that morphological diversity, or disparity, did increase quite rapidly during this event).

In the following section, the author distinguishes five senses of contingency (drift, unpredictability, causal dependence, sensitivity to external disturbance and unbiased species sorting) and reviews some attempts to test the applicability of these notions independently of the fossils of Burgess. Among these attempts, he mentions the convergence of morphologies in the Anolis lizard, the LTEE, the presence of unique versus shared evolutionary innovations and the ubiquity of convergence. Erwin also mentions some conceptual issues regarding these senses of contingency, for example, the need to determine the level at which the claim is being evaluated (e.g., parallelism at one level is compatible with divergence at a lower one, see chapter 8).

5 Final Thoughts

In sum, Chance in Evolution is a good overview of the current research into the roles of chance in evolutionary theory. As the precedent sections have shown, the book illustrates very well the variety in the kinds of problems involved (from empirical to conceptual to historical ones) and, in consequence, the variety of disciplines from which this topic has been studied. I suspect that most readers will not be interested in every chapter of the book—this is normal though, for a collection of essays of this amplitude. On the flipside, given its wide scope, the book will very likely have something to offer for everyone interested in the topic.

As was said in the introduction of this review, educators will likely find use for some of the material as secondary reading, especially at the university and graduate levels. This is especially so for courses focusing on the historical or conceptual aspects of the subject, while not so much for those looking for a more technical/mathematical description of the apparatus currently used in evolutionary biology to model these phenomena.

The issue of chance in evolution is a complex one, and this book succeeds in showing its many facets. The wide variety of themes are generally well treated. Overall, I think it will be a valuable resource for researchers and students alike.