Based on several recent discussions in the blogosphere, I decided to re-read Stephen Jay Gould’s paper entitled “Is a new and general theory of evolution emerging?”, published in 1980. This is one of Gould’s most (in)famous papers, but one that I am convinced most people have not read carefully, or at least have not understood. It’s the one that is usually cited when Gould is accused of calling for a rejection of natural selection or population genetics. Apparently it is also the one that caused many people to think that punctuated equilibria was a saltationist theory. In this post, I will quote extensively from the paper, with some thoughts and opinions interjected throughout.
To start, here is the abstract:
The modern synthesis, as an exclusive proposition, has broken down on both of its fundamental claims: extrapolationism (gradual allelic substitution as a model for all evolutionary change) and nearly exclusive reliance on selection leading to adaptation. Evolution is a hierarchical process with complementary, but different, modes of change at its three major levels: variation within populations, speciation, and patterns of macroevolution. Speciation is not always an extension of gradual, adaptive change to greater effect, but may represent, as Goldschmidt argued, a different style of genetic change–rapid reorganization of the genome, perhaps non-adaptive. Macroevolutionary trends do not arise from the gradual, adaptive transformation of populations, but usually from higher-order selection operating upon groups of species, while the individual species themselves generally do not change following their geologically instantaneous origin. I refer to these two discontinuities in the evolutionary hierarchy as the Goldschmidt break (between change in populations and speciation) and the Wright break (between speciation and trends as differential success among species).
A new and general evolutionary theory will embody this notion of hierarchy and stress a variety of themes either ignored or explicitly rejected by the modern synthesis: punctuational change at all levels, important non-adaptive change at all levels, control of evolution not only by selection, but equally by constraints of history, development and architecture–thus restoring to evolutionary theory a concept of organism. [Emphasis added]
The paper proper is divided into six parts. I will work through each in sequence, noting areas where I think confusion or misquotation have distorted the intended arguments.
I. The Modern Synthesis
Gould begins by quoting what he calls “one of the last skeptical books written before the Darwinian tide of the modern synthesis asserted its hegemony”, by Robson and Richards (1936):
The theory of Natural Selection … postulates that the evolutionary process is unitary, and that not only are groups formed by the multiplication of single variants having survival value, but also that such divergences are amplified to produce adaptations (both specializations and organization). It has been customary to admit that certain ancillary processes are operative (isolation, correlation), but the importance of these, as active principles, is subordinate to selection. [Robson and Richards 1936, p.370-371].
He then summarizes the core of the Modern Synthesis as follows:
Its foundation rests upon two major premises: (1) Point mutations (micromutations) are the ultimate source of variability. Evolutionary change is a process of gradual allelic substitution within a population. Events at broader scale, from the origin of new species to long-ranging evolutionary trends, represent the same process, extended in time and effect–large numbers of allelic substitutions incorporated sequentially over long periods of time. In short, gradualism, continuity, and evolutionary change by the transformation of populations. (2) Genetic variation is the raw material only. Natural selection directs evolutionary change. Rates and directions of change are controlled by selection with little constraint exerted by material (slow rates are due to weak selection, not insufficient variation). All genetic change is adaptive (though some phenotypic effects, due to pleiotropy, etc., may not be). In short, selection leading to adaptation. [Gould 1980, p.119-120.]
As anyone reading this surely knows, there are exceptions to all of these assumptions. Of course, proponents of the Modern Synthesis have always known this too. But, as Gould points out:
All the synthesists recognized exceptions and “ancillary processes,” but they attempted to both prescribe a low relative frequency for them and to limit their application to domains of little evolutionary importance. Thus, genetic drift certainly occurs–but only in populations so small and so near the brink that their rapid extinction will almost certainly ensue. And phenotypes include many non-adaptive features by allometry and pleiotropy, but all are epiphenomena of primarily adaptive genetic changes and none can have any marked effect upon the organism (for, if inadaptive, they will lead to negative selection and elimination and, if adaptive, will enter the model in their own right). Thus, a synthesist could always deny a charge of rigidity by invoking these official exceptions, even though their circumscription, both in frequency and effect, actually guaranteed the hegemony of the two cardinal principles. [Gould 1980, p.120].
Recognizing that he might be accused of erecting a straw man, he then quotes directly from Mayr (1963), one of the architects of the Modern Synthesis:
The proponents of the synthetic theory maintain that all evolution is due to the accumulation of small genetic changes, guided by natural selection, and that transspecific evolution is nothing but an extrapolation and magnification of the events that take place within populations and species. [Mayr 1963, p.586, emphasis added].
Referring specifically to this characterization of the synthesis, Gould wrote the following paragraph which, perhaps more than any of his other writings, raised the ire of defenders of the synthesis:
I well remember how the synthetic theory beguiled me with its unifying power when I was a graduate student in the mid-1960’s. Since then I have been watching it slowly unravel as a universal description of evolution. The molecular assault came first, followed quickly by renewed attention to unorthodox theories of speciation and by challenges at the level of macroevolution itself. I have been reluctant to admit it–since beguiling is often forever–but if Mayr’s characterization of the synthetic theory is accurate, then that theory, as a general proposition, is effectively dead, despite its persistence as textbook orthodoxy. [Gould 1980, p.120].
So, was it textbook orthodoxy, or was Gould making this up? I don’t have any textbooks from 1980, but here’s what Freeman and Herron (2007, p.96) state in one of the most widely used evolution textbooks:
…the Modern Synthesis, or the Evolutionary Synthesis, was a consensus grounded in two propositions:
- Gradual evolution results from small genetic changes that are acted upon by natural selection.
- The origin of species and higher taxa, or macroevolution, can be explained in terms of natural selection acting on individuals, or microevolution.
Nevertheless, critics construed Gould’s statement about the death of the Modern Synthesis as characterized by Mayr to mean that he was advocating a total overthrow of evolutionary theory. Here is how Charlesworth et al. (1982) quoted the passage above:
I have been watching it [neo-Darwinism] slowly unravel as a universal description of evolution … I have been reluctant to admit it … but … that theory, as a general proposition, is effectively dead, despite its persistence as a text-book orthodoxy.
I suspect that this is the version most people remember. But here is the full quote again, with the cherry picked bits indicated:
I well remember how the synthetic theory beguiled me with its unifying power when I was a graduate student in the mid-1960’s. Since then I have been watching it slowly unravel as a universal description of evolution. The molecular assault came first, followed quickly by renewed attention to unorthodox theories of speciation and by challenges at the level of macroevolution itself. I have been reluctant to admit it–since beguiling is often forever–but if Mayr’s characterization of the synthetic theory is accurate, then that theory, as a general proposition, is effectively dead, despite its persistence as textbook orthodoxy.
What Gould argues in Part I is simply that the Modern Synthesis, as it had developed by 1980, had a rather narrow focus on small-scale variation and natural selection, and assumed that most major evolutionary patterns could be explained by extrapolating processes within populations into deep time. What he was rejecting was this very narrow interpretation, though later critics gave the impression that he was challenging all aspects of the synthesis and calling for a revolution in evolutionary theory.
II. Reduction and Hierarchy
In this section, Gould highlights the inherent conflict between reductionism/extrapolationism and a theory involving hierarchy and emergence. As he put it:
The general alternative to such reductionism is a concept of hierarchy–a world constructed not as a smooth and seamless continuum, permitting simple extrapolation from the lowest level to the highest, but as a series of ascending levels, each bound to the one below it in some ways and independent in others. Discontinuities and seams characterize the transitions; “emergent” features not implicit in the operation of proceses at lower levels, may control events at higher levels. The basic processes–mutation, selection, etc.–may enter into explanations at all scales (and in that sense we may still hope for a general theory of evolution), but they work in different ways on the characteristic material at divers[e] levels [Gould 1980, p.121].
He then gives an example from the molecular level, including the notion of gene regulation (versus purely reductionistic genetics), repetitive DNA, transposable elements, and developmental regulatory genes of large effect. Thus, he says:
We may find, for example, that structural gene substitutions control most small-scale, adaptive variation within local populations, while disruption of regulation lies behind most key innovations in macroevolution. [Gould 1980, p.121].
He concludes this section by noting:
The modern synthesis drew most of its direct conclusions from studies of local populations and their immediate adaptations. It then extrapolated the postulated mechanism of these adaptations–gradual, allelic substitution–to encompass all larger-scale events. The synthesis is now breaking down on both sides of this argument. Many evolutionists now doubt exclusive control by selection upon genetic change within local populations. Moreover, even if local populations alter as the synthesis maintains, we now doubt that the same style of change controls events at the two major higher levels: speciation and patterns of macroevolution. [Gould 1980, p.121].
The importance of structural mutations versus regulatory mutations is a topic of substantial debate in current evolutionary biology, so I think Gould was on the mark here. We also have learned a lot more about the nature of eukaryote genomes, which are far more complex than Gould imagined nearly 30 years ago.
III. A Note on Local Populations and Neutrality
In this section, Gould takes a brief detour to emphasize the importance of neutral change at the molecular level–something that was largely lost from the Modern Synthesis as it grew to strongly emphasize selection. He begins by noting that there is more genetic variation in populations than was expected by the Modern Synthesis. One reason is that frequency dependent selection or selection in response to mildly fluctuating environments can maintain polymorphisms (versus directional or stabilizing selection, which both eliminate variation). More importantly, he argues that this reflects a much stronger than anticipated role for genetic drift and the predominance of neutral mutations. Again, he does not take this as a reason to overthrow all aspects of the synthesis:
None of this evidence, of course, negates the role of conventional selection and adaptation in molding parts of the phenotype with obvious importance for survival and reproduction. Still, it rather damps Mayr’s enthusiastic claim for “all evolution … guided by natural selection”. The question, as with so many issues in the complex science of natural history, becomes one of relative frequency. Are the Darwinian substitutions merely a surface skin on a sea of variation invisible to selection, or are the neutral substitutions merely a thin bottom layer underlying a Darwinian ocean above? Or where in between? [Gould 1980, p.122].
In summary, neutral processes are not a challenge to the importance of Darwinian adaptation, only to its exclusivity. Acknowledging a large role for neutral processes at the molecular and population levels would challenge the Modern Synthesis to the extent that it is based on an assumption that selection is the dominant mechanism at all scales and that chance enters only at the level of providing small mutations on which selection may act.
IV. The Level of Selection and the Goldschmidt Break
Gould begins this section by noting that speciation usually cannot be observed or replicated in the lab, and that therefore models of speciation are typically based on analogy and inference. He points out that Darwin drew an analogy with domestication, and considered subspecies as incipient species that gradually evolve into species through natural selection. Mayr’s allopatric speciation model challenged this, of course, but Gould suggests that it still maintained two major principles based on smooth extrapolationism:
(i) The accumulating changes that lead to speciation are adaptive. Reproductive isolation is a consequence of sufficient accumulation. (ii) Although aided by founder effects and even (possibly) by drift, although dependent upon isolation from gene flow, although proceeding more rapidly than local differentiation within large populations, successful speciation is still a cumulative and sequential process powered by selection through large numbers of generations. It is, if you will, Darwinism a little faster. [Gould 1980, p.122].
Gould then notes, “I do not doubt that many species originate in this way; but it now seems that many, perhaps most, do not.” He argues that several new (in 1980) models of speciation challenge the extrapolationist explanation. Some of these involve major chromosomal changes. I believe this is why he is often accused of proposing punctuated equilibria as a saltationist theory, but then of retreating when he was critcized for this view.
Gould’s list of challenges to “allopatric orthodoxy” relate to the amount of gene flow within species and especially to the plausibility of sympatric speciation. He argues that the assumption that gene flow homogenizes large populations (thereby making isolation necessary) may not be valid, and that instead local demes may be distinct enough for speciation. If that is the case, then the distinction between allopatric and sympatric speciation is oversimplified and sympatric speciation is much more likely than acknowledged. Strong selection or the rapid fixation of major chromosomal variations could underlie such processes. Here is where Gould first talks about rapid chromosome-level isolation:
The most exciting entry among punctuational models for speciation in ecological time is the emphasis, now coming from several quarters, on chromosomal alterations as isolating mechanisms–sometimes called chromosomal speciation. In certain population structures, particularly in very small and circumscribed groups with high degrees of inbreeding, major chromosomal changes can rise to fixation in less than a handful of generations [Gould 1980, p.123].
He uses the word “punctuational”, but he does not mention “punctuated equilibria” at all or cite any of his paper on the subject, and indeed what he is describing — mechanisms of rapid sympatric speciation — has little to do with punctuated equilibria, which is about peripheral isolates. As far as saltational mechanisms are concerned, their challenge to orthodoxy is that they undermine the emphasis on selection acting on small mutations as the dominant mechanism. He continues:
The control of evolution by selection leading to adaptation lies at the heart of the modern synthesis. Thus, reproductive isolation, the definition of speciation, is attained as a by-product of adaptation–that is, a population diverges by sequential adaptation and eventually becomes sufficiently different from its ancestor to foreclose interbreeding … But in saltational, chromosomal speciation, reproductive isolation comes first and cannot be considered as an adaptation at all. It is a stochastic event that establishes a species by the technical definition of reproductive isolation. To be sure, the later success of this species in competition may depend upon its subsequent acquisition of adaptations; but the origin itself may be non-adaptive. We can, in fact, reverse the conventional view and argue that speciation, by forming new entities stochastically, provides raw material for evolution. [Gould 1980, p.124].
Gould then notes that Richard Goldschmidt argued, while the synthesis was being constructed, that there is a break between what happens in populations and what happens when new species originate. He calls this the “Goldschmidt break”. Gould ends this section by quoting Goldschmidt (1940), though not in relation to any particular ideas about mutation or saltationism, only to say that extrapolation from short-term population processes is not enough to explain all speciation:
The characters of subspecies are of a gradient type, the species limit is characterized by a gap, an unbridged difference in many characters. This gap cannot be bridged by theoretically continuing the subspecific gradient or cline beyond its actually existing limits. The subspecies do not merge into the species either actually or ideally … Microevolution by accumulation of micromutations–we may also say neo-Darwinian evolution–is a process which leads to diversification strictly within the species, usually, if not exclusively, for the sake of adaptation of the species to specific conditions within the area which it is able to occupy … Subspecies are actually, therefore, neither incipient species nor models for the origin of species. They are more or less diversified blind alleys within the species. The decisive step in evolution, the first step towards macroevolution, the step from one species to another, requires another evolutionary method than that of the sheer accumulation of micromutations. [Goldschmidt 1940, p.183].
One can certainly disagree with Goldschmidt that subspecies are never incipient species, or with Gould for quoting this, but one cannot argue that Gould endorsed Goldschmidt’s particular genetic views nor linked these to punctuated equilibria in any way in this section.
V. Macroevolution and the Wright Break
In this section, Gould moves from mechanisms of speciation to large-scale macroevolutionary trends. He opens by summarizing the interpretation of macroevolutionary patterns given by the Modern Synthesis, namely that they occur gradually and with branching as an afterthought (i.e., cladogenesis subordinate to anagenesis). Moreover, most of this gradual transformation is an adaptive outcome of population-level selection.
Gould presents several counterpoints to this extrapolationist model. First, he notes that models of speciation emphasizing branching and stasis suggest that higher-level trends may actually result from sorting at the species level rather than as an extension of selection at the population level. This is where punctuated equilibria comes in, and I think it is worth noting that he does not link this to saltational mechanisms. In fact, he explicitly notes that any model of speciation that emphasized comparatively rapid speciation followed by stasis would have the same implications:
We regard stasis and discontinuity as an expression of how evolution works when translated into geological time. Large, successful central populations undergo minor adaptive modifications of fluctuating effect through time (Goldschmidt’s “diversified blind alleys”), but they will rarely transform in toto to somethign fundamentally new. Gradual change is not the normal state of a species. Speciation, the basis of macroevolution, is a process of branching. And this branching, under any current model of speciation–conventional allopatry to chromosomal saltation–is so rapid in geological translation (thousands of years at most compared with millions for the duration of most fossil species) that its results should generally lie on a bedding plane, not through the thick sedimentary sequence of a long hillslope. (The expectation of gradualism emerges as a kind of double illusion. It represents, first of all, an incorrect translation of conventional allopatry. Allopatric speciation seems so slow and gradual in ecological time that most paleontologists never recognized it as a challenge to the style of gradualism–steady change over millions of years–promulgated by custom as a model for the history of life. But it now appears that “slow” allopatry itself may be less important than a host of alternatives that yield new species rapidly even in ecological time). Thus, our model of “punctuated equilibria” holds that evolution is concentrated in events of speciation and that successful speciation is an infrequent event punctuating the statis of large populations that do not alter in fundamental ways during the millions of years that they endure. [Gould 1980, p.125].
Again, the main point is that trends at the macroevolutionary scale may represent the outcome of differential survival and reproduction (speciation) of species. This would make macroevolutionary trends at least partly distinct from microevolutionary processes. Gould notes that Sewall Wright emphasized this, which is why he calls this the “Wright break”. This is the one place that punctuated equilibria is discussed, specifically in light of its implications for species selection.
The second major issue that Gould then raises is that sometimes “saltational” changes probably do happen. Not at the level of normal speciation events, but in major innovations. (Again, this would have nothing to do with punctuated equilibria, which is strictly about the patterns of speciation). He also does not mean that whole new structures appear suddenly. As he explains:
I do not refer to the saltational origin of entire new designs, complete in all their complex and integrated features–a fantasy that would be truly anti-Darwinian in denying any creativity to selection and relegating it to the role of eliminating old models. Instead, I envisage a potential saltational origin for the essential features of key adaptations. [Gould 1980, p.127].
In general, this sort of “saltation” would involve major developmental changes that could be caused by small mutations of regulatory genes. Gould rejects Goldschmidt’s hypothesis of “systemic mutations”, but suggests that developmental regulatory mutations could sometimes lead to the saltational origin of basic forms of new features. This is a topic that is widely discussed today, so again, I think Gould was basically right to raise this point. In fact, the example he gives is of the origin of the jaw from gill arches, which he argues could have been “saltational” in the sense that it occurred quickly through a developmental mutation. Interestingly, a recent paper by Shigetani et al. (2002) suggests that this transition did indeed involve a developmental mutation, specifically a heterotopic shift in gene expression.
The third major challenge to extrapolationism in explaining macroevolutionary patterns that Gould presents has to do with non-adaptive processes. This follows from his classic paper with Richard Lewontin, published the year before (Gould and Lewontin 1979). Here, he revisits the points about there being more variation in populations than adaptationist models should expect, and that speciation by chromosomal mechanisms could be non-adaptive. He then moves on to discuss the relevance of large-scale trends that are not caused by population-level selection. As he says:
… if trends represent a higher-level process of differential origin and mortality among species, then a suite of potentially non-adaptive explanations must be considered. Trends, for example, may occur because some kinds of species tend to speciate more often than others. This tendency may reside in the character of environments or in attributes of behavior and population structure bearing no relationship to morphologies that spread through lineages as a result of higher speciation rates among some of their members. Or trends may arise from the greater longevity of certain kinds of species. Again, this greater persistence may have little to do with the morphologies that come to prevail as a result. [Gould 1980, p.128].
The point, yet again, is that the Modern Synthesis as described by Mayr — which emphasized gradual, adaptive change as the cause of macroevolution — is undermined by other processes at both higher and lower levels. Gould is not arguing that the synthesis is incorrect or that it needs to be overthrown, only that it is unduly narrow and needs to be expanded. Many of the issues he raises here are still under debate, though they are still dismissed by some who argue that the synthesis is more or less complete.
VI. Quo Vadis?
In this concluding section, Gould is clear about what he is arguing:
I think I can see what is breaking down in evolutionary theory–the strict construction of the modern synthesis with its belief in pervasive adaptation, gradualism, and extrapolation by smooth continuity from causes of change in local populations to major trends and transitions in the history of life. [Gould 1980, p.128].
Here is what he sees as the features of a revised evolutionary theory:
- It will be hierarchical in nature. This doesn’t mean that each level is totally disconnected, only that we will have to recognize the operation of (similar) mechanisms at each level.
- It will include punctuational patterns at all levels, not only gradual change.
- It will emphasize the role of constraints in shaping phenotypes.
- It will properly recognize the input of non-adaptive processes at all levels.
- It will link insights from paleontology, developmental genetics, and molecular biology with those derived from population studies.
Overall, the paper seems quite mild in retrospect. Gould did not call for a revolution in evolutionary biology, nor did he transform punctuated equilibria into a fundamentally radical mode of speciation. He merely argued against the features of the Modern Synthesis that its proponents themselves outlined. Many of the arguments he raised were ahead of their time but are now common in evolutionary discourse. This is especially true given the enormous growth of biological information in the nearly three decades since this paper was published.
Charlesworth, B., R. Lande, and M. Slatkin. 1982. A neo-Darwinian commentary on macroevolution. Evolution 36: 474-498.
Freeman, S. and J.C. Herron. 2007. Evolutionary Analysis (Fourth Edn). Prentice Hall, Upper Saddle River, NJ.
Goldschmidt, R. 1940. The Material Basis of Evolution. Yale University Press, New Haven.
Gould, S.J. and R.C. Lewontin. 1979. The spandrels of San Marco and the Panglossian paradigm: a critique of the adaptationist programme. Proceedings of the Royal Society of London B 205: 581-598.
Gould, S.J. 1980. Is a new and general theory of evolution emerging? Paleobiology 6: 119-130.
Mayr, E. 1963. Animal Species and Evolution. Harvard University Press, Cambridge, MA.
Robson, G.C. and O.W. Richards. 1936. The Variation of Animals in Nature. Longmans, Green, and Co., London.
Shigatani, Y., F. Sugahara, Y. Kawakami, Y. Murakami, S. Hirano, and S. Kuratani. 2002. Heterotopic shift of epithelial-mesenchymal interactions in vertebrate jaw evolution. Science 296: 1316-1319.
Update — See also Gould (1982).