Editor’s note: This essay continues a dialogue that started with Steve Pinker’s essay titled “The False Allure of Group Selection” published on Edge.org. All readers are invited to comment at the end of this essay. In addition, professional evolutionists are invited to provide more extensive comments on the Social Evolution Forum.
Thomas Kuhn (1970) forever changed the conception of science with his notion of paradigms. Before, science was often seen as a relatively straight path to the truth through the repeated formation and testing of hypotheses. What could be simpler?
Kuhn observed that scientists sometimes get stuck viewing a topic a certain way. Their particular configuration of ideas is capable of a limited degree of change through hypothesis formation and testing, but cannot escape from its own assumptions in other respects. This makes the replacement of one paradigm by another a complex and uncertain process.
A clash of paradigms is currently taking place for evolutionary theories of social behavior. In this corner, multilevel selection theory (MLST), a configuration of ideas that began with Darwin and has maintained a degree of continuity, in addition to a degree of change, up to the present. In that corner, inclusive fitness theory (IFT), which can also claim roots in Darwin and has also changed while remaining true to a core set of ideas.
The most recent battle between the two paradigms began when Edward O. Wilson, one of the most celebrated living evolutionists, became a vocal proponent of MLST and started to denounce the utility of IFT. Those who are familiar with Wilson’s work know that he has been receptive to MLST all along (read his chapter on group selection in Sociobiology (1975) for details). I should know, because he sponsored the publication of my first article on group selection in the Proceedings of the National Academy of Sciences in 1975 and we have co-authored a number of more recent articles together, including the comprehensive “Rethinking the Theoretical Foundation of Sociobiology” published in the Quarterly Review of Biology in 2007. Wilson’s more recent “conversion” was notable less for his acceptance of MLST than his rejection of IFT as a useful paradigm. He was joined by the eminent mathematical biologist Martin Nowak (along with his young-career colleague Corina Tarnita) in a major article in Nature, and elaborated on his views in his most recent book The Social Conquest of Earth.
Proponents of IFT could not take this assault on their paradigm lying down. The responses to the article in Nature included one with 137 co-authors (Abbott et al. 2010). Richard Dawkins wrote a spirited review of Wilson’s book in Prospect magazine and Steven Pinker wrote an essay for Edge.org titled “The False Allure of Group Selection”. The twenty commentaries (including one by myself) following Pinker’s essay provide a fascinating snapshot of the clash between the two paradigms.
Before continuing, I want to stress that the clash between MLST and IFT departs from the Kuhnian notion of paradigms in at least two respects. First, in the major examples discussed by Kuhn, one paradigm eventually collapses and is replaced by the other. Nobody talks about pre-Copernican views of the universe anymore. Even though proponents of MLST and IFT sometimes write as if the other paradigm has or will collapse, there is a strong sense in which a collapse of one paradigm shouldn’t be expected. Instead, the two paradigms are like different languages, such as Russian and English, a metaphor that I will elaborate upon below.
Second, Kuhnian paradigms are thought to be incommensurate, such that people who think in terms of one truly cannot see the world in terms of the other. In the case of MLST and IFT, some proponents fit this description but others can easily grasp both paradigms and acknowledge the utility of one, even if they have a preference for the other. I count myself among them as a proponent of MLST who acknowledges the utility of IFT, along with David Queller, a proponent of IFT who acknowledges the utility of MLST. The language metaphor is apt: For people who speak only a single language, another language appears confusing and redundant. People who have become bilingual can easily toggle between two languages and have no wish for one to replace the other.
Queller and I are not alone. There is a sizeable community of evolutionists who are bilingual with respect to MLST and IFT. If anything deserves to collapse in the clash between these two paradigms, it the unilingual position that only one paradigm deserves to exist. When unilinguals become bilingual, the so-called “group selection controversy” will be over.
Diagnosing the claim that one paradigm is confusing and unproductive
The current battle between proponents of MLST and IFT include claims that one’s non-preferred theory is confusing, inconsistent, unproductive and adds nothing to one’s preferred theory. Here is a sample of quotes from MLST proponents criticizing IFT.
“Yet, considering its position for four decades as the dominant paradigm in the theoretical study of eusociality, the production of inclusive fitness theory must be considered meagre….Inclusive fitness theory is only another method of accounting, one that works for very restrictive scenarios and where it works it makes the same predictions as standard natural selection theory. Hence, there are no predictions that are specific to inclusive fitness theory. (Nowak et al., 2010).”
“Equations seemed to arise out of nowhere in kin selection… Moreover, the concept of “relatedness” seemed to morph and change over time …Casting a problem in terms of inclusive fitness is like having to undergo elaborate and time-consuming initiations to join an elite club, only to end up with nothing in the way of privileges (Nowak and Highfield 2011).”
“Much of the inadequacy of the theory comes from looseness in the definition of r, hence the very concept of kinship, in various interpretations of the Hamilton inequality….the only unifying theme seemed in time to be that r, originally defined by pedigree, is whatever it takes to make Hamilton’s inequality work. The inequality thereby lost meaning as a theoretical concept, and became all but useless as a tool for designing experiments or analyzing comparative data (E.O. Wilson 2012).”
Proponents of IFT protested en masse against these statements, but here is what some of them have to say about MLST.
“The first big problem with group selection is that the term itself sows so much confusion. People invoke it to refer to many distinct phenomena, so casual users may literally not know what they are talking about. (Pinker 2012).”
“The first and deepest problem with this debate is that the term group selection does not have any single fixed meaning, but has been used over the last half century to convey a huge and tangled thicket of different and conflicting meanings. The great majority of these are seriously defective as a way of describing reality. (Cosmides 2012).”
“‘Group selection’, even in the rare cases where it is not actually wrong, is a cumbersome, time-wasting, distracting impediment to what would otherwise be a clear and straightforward understanding of what is going on in natural selection. (Dawkins 2012).”
“Models of group selection are either mathematically equivalent to those based on kin selection but less tractable, or are so nebulous that they can’t be analyzed at all. Further, claims that kin selection is less useful than group selection in understanding nature are simply wrong. (Coyne 2012).”
The symmetry of these complaints strongly suggests the existence of two paradigms. Each paradigm is an internally consistent configuration of ideas with explanatory power for those who use the paradigm. But viewing the world from within one paradigm makes the other paradigm appear confusing, unproductive, and so on. The authors of these statements write as if their criticisms are true in some absolute sense, when in fact the criticisms are only true for the authors in a relational sense. That’s what the basic phenomenon of paradigms is all about.
Why the two paradigms are incommensurate
For readers who are unfamiliar with MLST and IFT, here is a brief description of why they are so incommensurate with each other. To begin, they define basic terms such as “individual selection” and “altruism” in different ways. For MLST, individual selection is defined in terms of relative fitness within groups. A trait is altruistic when it benefits others or the group as a whole and decreases the fitness of the altruist, relative to other members of its own group. For IFT, direct individual fitness is defined in absolute terms. A trait is altruistic when it benefits others or the group as a whole and decreases the absolute fitness of the altruist. The term “group selection”, which plays a key role in MLST, does not exist within IFT. The term “inclusive fitness”, which plays a key role in IFT, does not exist within MLST.
The differences between the two theories are sufficiently great that W.D. Hamilton regarded MLST as a failed theory and developed IFT to explain the evolution of altruism without invoking group selection. Briefly, his method involved calculating the effect of an altruistic act on the absolute number of copies of the altruistic gene that are identical by descent. The same altruistic act benefits copies of the non-altruistic gene and copies of the altruistic gene that are not identical by descent, but these effects can be ignored because they do not alter the frequency of the altruistic gene in the gene pool. Only effects on genes that are identical by descent produce evolutionary change. Given the assumptions of the model, Hamilton’s rule gives the conditions for the spread of the altruistic gene in the gene pool.
Given the same assumptions, it is possible to calculate the relative fitness of the altruistic gene within and among groups. Consider a group of relatives that are socially interacting with each other. Some are altruistic and others are not. We don’t keep track of genes that are identical by descent. Instead, we calculate the effect of the altruistic act on the frequency of the altruistic gene within the group of relatives. The altruistic gene is at a selective disadvantage within the group. The only way for the altruistic gene to evolve in the total gene pool is if groups with more altruists contribute more to the gene pool than groups with fewer altruists.
Both methods employ the same assumptions and make the same predictions about when the altruistic gene evolves in the total population, but they seem so different that even their developers were confused about their equivalence. The history of their confusion is well known to scholars but deserves to be more widely known among the general public. The first MLST model of altruism in family groups was by George and Doris Williams and was published in the journal Evolution in 1957. Hamilton published his IFT model in 1963 in The American Naturalist and 1964 in the Journal of Theoretical Biology. All of these are top journals and it is silly to claim that either formulation is confusing in some absolute sense. Hamilton was unaware of the Williams & Williams paper in the development of his own theory. George Price developed a MLST model that partitioned natural selection into within- and between-group components, which was first published in the journal Nature in 1970. It was more general and elegant than the Williams & Williams model and could be applied to family groups in addition to many other kinds of groups. Only when Hamilton encountered the Price equation did he fully appreciate the equivalence between IFT and MLST, as he described in a paper published in 1975. This history is recounted for a general audience by the historian of science Oren Harman in his book The Price of Altruism (2010), and recently summarized by Eric Michael Johnson (2012), who blogs for Scientific American, in the context of the current clash.
1975 was also the year that I published my first article on group selection in the Proceedings of the National Academy of Sciences, with E.O. Wilson as the sponsoring editor. One of my contributions was to define the concept of a trait-group. Groups do not exist in any absolute sense; instead, they are typically defined in reference to particular activities or traits. This is true in everyday language (e.g., my nation, my school, my church, my bowling club) and it is also true for evolutionary models of social behavior. A model for the evolution of warning calls will define the group as the set of individuals within earshot of each other. A model for the evolution of resource conservation will define groups as the set of individuals who compete for the same resource. Mathematically, whenever the fitness of an individual is influenced by other individuals, it must be defined by a function f(N,p), where N and p reflect the number and genetic composition of the group influencing the focal individual’s fitness. All models of social behavior must define the grouping that is appropriate for a particular trait, which is the trait-group for that model.
My second contribution was to show that when trait-groups are examined from a multilevel perspective, the core claim of group selection is often correct. In other words, traits can evolve in the total population even when they are selectively disadvantageous within groups. This result was in contrast to the prevailing belief that higher-level selection is almost invariably weak compared to lower-level selection. Fancy math was not required to draw this conclusion. A simple algebraic model was sufficient. My model was less general than the Price equation, but both models made the same point.
Ed Wilson was cautious about publishing my paper in PNAS. After all, acceptance of group selection was at its low ebb and he would appear foolish if there was something wrong with my argument. He therefore had it thoroughly reviewed by theoretical biologists, who examined my argument from stem to stern before deciding that it was correct. That was 37 years ago, and still the clash between the two paradigms continues. To give a sense of proportion, 37 years represents 24% of the history of Darwinian thought, starting with the publication of Origin of Species. If that’s not an example of incommensurability, what would be?
How Equivalence Changes the Concept of Paradigms
MLST and IFT provide outstanding examples of paradigms in terms of their incommensurability, but now it is important to clarify why they depart from the concept of paradigms in other respects. In the examples that Kuhn described, such as the Copernican revolution, one paradigm (the view that the earth is the center of the solar system) is eventually replaced by the other (the view that the sun is the center of the solar system). There was a right and wrong about the matter, even though the rejection of the false paradigm was not smooth.
In the case of MLST and IFT, neither paradigm is wrong in the way that the pre-Copernican view of the solar system was wrong. Instead, they parse the conceptual space in different ways but reach the same conclusion on right-and-wrong issues such as when a given altruistic behavior evolves in the gene pool. Different metaphors are employed to describe their equivalence, such as “accounting methods”, “languages”, or “perspectives”.
All three of these metaphors are apt in their own ways. Imagine creating a spreadsheet to organize your finances. In some ways it is useful to organize your expenditures in terms of date. In other ways it is useful to organize them in terms of categories such as food, business, and so on. Both are equally correct and deserve to coexist because they are useful for different purposes. In the same way, MLST accounts for gene frequency change in the total population by creating the categories “within-group selection” and “between-group selection”. The IFT method creates the categories “genes identical by descent” vs. “genes that represent a random draw from the gene pool”. Neither is wrong and they deserve to coexist if they are useful for different purposes.
Different languages are not equivalent at the level of single words; if they were, then translating between languages would be easy. Instead, different languages parse the world in different ways. The concept represented by a single word in one language can require a treatise to understand in a different language. If the entire world spoke a single language, then a precious form of diversity would be lost. Nevertheless, it is possible to translate among languages and gifted translators are admired almost as much as the authors that they translate. If MLST and IFT are like different languages, then each can offer unique insights by parsing the evolutionary process in different ways, even though the insights from one can be translated into the other language.
Finally, imagine that you are planning to ascend a mountain with a friend. As you survey its contours, certain features are difficult to make out from where you are standing. It makes sense for your friend to stand in a different spot so that you can collectively get a better view of the mountain’s contours. MLST and IFT might offer different perspectives on the same evolutionary process, in the same way that you and your friend offer different perspectives on the same mountain.
All of these metaphors help to explain how two scientific paradigms such as MLST and IFT can be different yet worthy of coexistence, a combination that goes beyond the traditional concept of paradigms. Against this background, it is interesting to revisit some of the arguments quoted above, which are intended by proponents of one paradigm to drive the other paradigm extinct.
Consider the argument that a given paradigm is confusing. I am confused by Russian, but that’s my problem, not a problem with the Russian language. My confusion will cease as soon as I learn Russian. When we view paradigms as like languages, the “confusion” criticism becomes no criticism at all.
Or consider the argument that every result derived from one paradigm can also be derived from the other paradigm. This is expected for different accounting systems, languages, and perspectives, but it ignores the advantages inherent in their differences. If I need to tally my business expenses, an accounting system organized by date will be cumbersome, even though it might be possible. A given result that emerges easily from one perspective might be opaque from another perspective and only derivable in retrospect. The inter-translatability criticism becomes no criticism at all.
Finally, consider the argument that a given paradigm changes over time. Both paradigms have changed profoundly over the decades. For IFT, the interpretation of r became generalized beyond genealogical relatedness. For MLST the interpretation of groups became generalized. These changes are regarded as progress within each paradigm and are only confusing to those who don’t speak the language. The “always changing” criticism becomes no criticism at all.
In short, the current clash between proponents of MLST and IFT is primarily a clash between people who are fluent in one paradigm and confused by the other paradigm, which they falsely attribute to the other paradigm in some absolute sense. This is like saying “Russian is confusing”, rather than “Russian is confusing for a non-Russian speaker such as myself.” They also expect that their paradigm will triumph over the other paradigm, in the same way that the Copernican view triumphed over the pre-Copernican view. Replacement is not expected for accounting systems, languages, or perspectives, and it shouldn’t be expected for MLST and IFT.
Listen to the Bilinguals
The metaphors of paradigms as like accounting systems, languages, and perspectives are apt in another respect. They all imply that a single person can master more than one paradigm. It is possible to organize your financial records by date and by category. It is possible to be bilingual. And a single climber can view a mountain from different locations. These stand in contrast to the standard view of paradigms, which implies that incommensurability is inescapable.
Many evolutionists are fluent in both MLST and IFT, a fact that is sometimes obscured by the chest-thumping rhetoric of unilinguals. Here is a sample of comments by bilinguals who responded to Steven Pinker’s essay:
“Modern group selection theory is as mathematically rigorous as individual selection or inclusive fitness theory. I say this despite being someone who favors the inclusive fitness approach and whose entire career has been based on it. I think of these less as alternative theories that make different predictions than as two different languages describing the same world. They simply divide up fitness in slightly different ways – inclusive fitness into effects on self versus others, and multilevel selection into between-group and within-group parts – and a simple partition of fitness should not alter predictions (Queller 2012).”
“As with other kinds of dynamic processes, models of genetic evolution can be built many different ways. The issue of “group selection” revolves around the choice of an accounting system: how does one wish to track fitness or changes in gene frequencies. In many cases (though not all), the exact same process can be represented and developed using quite different evolutionary accounting systems [1-3]. These accounting systems include (1) individual fitness, (2) inclusive fitness, and (3) multi-level or “group” selection (Henrich 2012).”
“The first misconception here is the view that group selection is incompatible with kin selection. It is not. Kin selection says that the fitness of an individual depends on the genes of his kin and not just his own genes. Group selection says the fitness of an individual depends on the characteristics of the group he is in, not just his own genes (Gintis 2012).”
For these and many other evolutionists, toggling back and forth between MLST and IFT has become normal science, with no need for chest-thumping rhetoric. The choice of theory is simply a matter of choosing the best tool for the job.
The coexistence of paradigms does not mean that everyone is entitled to their own opinion on important empirical claims. As with the standard view of science, many hypotheses are just plain wrong and both paradigms must come to an agreement in rejecting them. Consider the central claim of MLST, which is that prosocial traits can evolve even when they are selectively disadvantageous within groups. The only way to evaluate this claim is by comparing gene frequency change within groups to gene frequency change in the total population. MLST is designed to make this comparison, like an accounting system designed for a certain purpose. IFT is not designed to make this comparison. Even though it includes the same information, it is like an accounting system organized by a different set of categories than those called for. Work is required to re-organize the information, but when this is done, IFT must come to the same conclusion as MLST, or else one of them is wrong.
In the current battle among unilinguals, this kind of empirical inquiry is a civilian casualty. To pick E.O. Wilson’s Social Conquest of Earth as an example, he makes a number of important empirical claims about the evolution of eusociality, in general and in relation to our own species, which he frames in terms of MLST. Bilinguals know that the same claims can be examined from an IFT perspective and that the two paradigms must ultimately come to the same conclusions. Yet, this inquiry has been largely halted by the chest-thumping rhetoric over which paradigm is the right one.
Beyond Group Selection
This essay is about two paradigms but it also stretches the concept of paradigms to include some that deserve to coexist rather than replacing each other. The metaphors of paradigms as accounting systems, languages, and perspectives are relevant to other topics in science, in addition to the group selection. In fact, Pinker’s essay is notable for addressing another topic—cultural evolution—in ways that have nothing to do with group selection. Once again, he defends a certain configuration of ideas as the only right one and disparages other configurations as confusing, unproductive, and changing. The possibility that he might be like a non-Russian speaker who blames the language for being confusing doesn’t occur to him. All scientists and intellectuals, indeed all people, need to become multi-lingual to avoid this kind of error.
Abbott, P., Abe, J., Alcock, J., & al., et. (2010). Inclusive fitness theory and eusociality. Nature, 471, E1-E4. doi:doi:10.1038/nature09831
Coyne, J. (2012). The problem with group selection. http://edge.org/conversation/the-false-allure-of-group-selection
Dawkins, R. (2012). The Descent of Edward Wilson. Prospect. http://www.prospectmagazine.co.uk/magazine/edward-wilson-social-conquest-earth-evolutionary-errors-origin-species/
Dawkins, R. (2012). “Group selection” is a cumbersome, time-wasting distraction. http://edge.org/conversation/the-false-allure-of-group-selection
Gintis, H. (2012). On the evolution of human morality. http://edge.org/conversation/the-false-allure-of-group-selection
Hamilton, W. D. (1963). The Evolution of Altruistic Behavior. American Naturalist, 97, 354-356.
Hamilton, W. D. (1963). The Evolution of Altruistic Behavior. American Naturalist, 97, 354-356.
Hamilton, W. D. (1964). The genetical evolution of social behavior: I and II. Journal of Theoretical Biology, 7, 1-52.
Hamilton, W. D. (1963). The Evolution of Altruistic Behavior. American Naturalist, 97, 354-356.
Harman, O. (2010). The Price of Altruism: George Price and the Search for the Origins of Kindness. New York: Norton.
Henrich, J. (2012). Too late: models of cultural evolution and group selection have already proved useful. http://edge.org/conversation/the-false-allure-of-group-selection
Johnson, E.M. (2012). The Good Fight. http://blogs.scientificamerican.com/primate-diaries/2012/07/09/the-good-fight/
Kuhn, T. S. (1970). The structure of scientific revolutions (2nd ed.). Chicago : University of Chicago Press.
Nowak, M. A., Tarnita, C. E., & Wilson, E. O. (2010). The Evolution of Eusociality. Nature, 466, 1057-1062.
Nowak, M.A., & Highfield, R. (2011). SuperCooperators: Altruism, Evolution, and Why We Need Each Other to Succeed. New York: Free Press.
Pinker, S. (2012). The False Allure of Group Selection. http://edge.org/conversation/the-false-allure-of-group-selection
Queller, D. (2012). Two languages, one reality. http://edge.org/conversation/the-false-allure-of-group-selection
Tooby, J. (2012) Genic selection and adaptationism: are we moving forward or back? http://edge.org/conversation/the-false-allure-of-group-selection
Williams, G. C., & Williams, D. C. (1957). Natural selection of individually harmful social adaptations among sibs with special reference to social insects, 11, 32-39.
Wilson, D. S. (1975). A Theory of Group Selection A Theory of Group Selection, 72(1), 143-146.
Wilson, D. S., & Wilson, E. O. (2007). Rethinking the theoretical foundation of sociobiology. Quarterly Review of Biology, 82, 327-348.
Wilson, E. O. (1975). Sociobiology: the new synthesis. Cambridge, Mass: Harvard University Press .
Wilson, E. O. (2012). The Social Conquest of Earth. New York: Norton.