Debunking the Selfish Gene

In this essay, I will explain the core error in Richard Dawkins’ notion of the selfish gene, and explain how it is related to group selection theory.

This is about biological purpose: what it is, and what has it. I will compare and contrast three theories of biological purpose:

  • Phenocentric Theory: The organism (more precisely, the reproducing unit) is the locus of biological purpose. The purpose of life is to reproduce. An organism is a reproducing machine.
  • Species-centric Theory: The species is the locus of biological purpose. The purpose of life is to perpetuate the species. An organism is a species-perpetuation machine.
  • Genocentric Theory: The gene (nucleotide sequence) is the locus of biological purpose. The purpose of life is to perpetuate genes. An organism is a gene-perpetuation machine.

The selfish gene concept is part of the genocentric theory, which is Dawkins’ view. He believes that organisms are gene-survival machines.

The species-centric theory is often called “group selection theory”. Dawkins correctly rejects it. However, the genocentric theory is based on a very similar misconception.

In this essay, I will argue that the phenocentric theory is correct, and the other two theories are incorrect.

Biological Relationships

There is an important distinction between part-whole relationships and instance-type relationships.

TypeHuman bodyHuman hand
InstanceMy bodyMy hand

A type is a category of objects with a common form. The human body is a type. Every human is an instance of that type. The human hand is also a type. An individual hand, such as my hand, is an instance of that type. Types are abstract. They consist only of information: the form that is actualized by the instances.

To be clear, type and form are essentially the same. A type is a category of instances that have a common form.

A whole is a form or object that is composed of different parts that are spatially, temporally and causally related. The human body is a whole, and the human hand is part of that whole. Likewise, my body is a whole, and my hand is part of my body. If the whole is an object, then it has an independent existence from other objects of the same type. By contrast, a part of a whole can only exist within the whole. One human body can exist without another. A human hand can only exist as part of a human body.

The species-centric and genocentric theories involve a confusion between instance-type and part-whole relationships. Both view an instance-type relationship as a part-whole relationship.

The Phenocentric Theory

In this section, I will briefly present the phenocentric theory of biological purpose. In the next two sections, I will explain the errors of the other two theories.

The phenocentric theory makes the part-whole and instance-type relationships clear. It locates purpose in the intersection of biological whole and instance, which is the organism.

TypeAbstract purposeAbstract function
InstanceActual purposeActual function

A type can have an abstract purpose, but not an actual purpose. The human body has an abstract purpose: to reproduce. This means that all of its instances have that actual purpose. The purpose of an individual human body is an instance of the abstract purpose. The purpose of an individual human body is to reproduce: to generate offspring.

Reproduction is not just creating, or helping to create, other human beings. It is generating your own offspring.

Parts of a whole have functions relative to the purpose of the whole. The human hand has an abstract function within the human body: to grasp and hold objects. This abstract function is instrumental to the abstract purpose of reproduction. An instance of the human hand has the actual function of grasping and holding objects, and is instrumental to the actual purpose of its body. An actual purpose/function is an instance of an abstract purpose/function.

The function of a part is instrumental to the purpose of the whole. Instrumentality is a relation between part and whole, not instance and type. The instance is not instrumental to the type in any way.

My hand does not serve the “interests” of the abstract human hand, of which it is an instance. It serves the interests of my body, of which it is a part.

Every part of the body is instrumental to the purpose of the body. Those parts inherit their forms and functions by reproduction, just as the body inherits its form and purpose by reproduction.

My purpose of reproduction is not your purpose of reproduction, nor anyone else’s. It is not just to create new instances of the human type. It is to produce my own offspring. Your purpose is to produce your own offspring. We each inherited an individual version of the abstract purpose.

Reproduction is the basis of natural selection, because it is the event through which biological forms are copied, and by which biological entities (organisms) are created.

A new mutation must be multiplied by reproduction to become frequent. To persist, a biological form must be copied by reproduction. Thus, biological forms are selected to be instrumental to reproduction.

The reproducing unit is the locus of purpose. In most cases, it is what we view as the organism. Reproduction creates new organisms. An organism is a reproducing machine.

✦ ✦ ✦

Now, let’s consider the other two theories. This table shows the relevant relationships.

TypeSpeciesAbstract gene
InstanceOrganismGene copy

The Species-centric Theory

In The Selfish Gene, Dawkins correctly rejects the species-centric theory: that natural selection is based on the perpetuation of the species.

The species-centric theory is based on a misconception about the relationship between the individual organism and the species. It views the relationship as part-whole, rather than instance-type. It views the species as a whole (entity) with an actual purpose, composed of organisms that are instrumental to that purpose.

The species is an abstract type, which consists only of information, and is somewhat in the eye of the beholder. It is just a type of organism. Organisms are not parts of the species, and they are not instrumental to the purpose of the species. The species does not have an actual purpose. It has an abstract purpose of reproduction, which means that each member of the species has the actual purpose of reproduction: generating its own offspring.

A simple thought experiment demonstrates that reproduction, not species-perpetuation, is the basis of selection. Suppose that a species contained some individuals who acted for the benefit of the species (defined as perpetuation or expansion) and some individuals who acted only toward the creation of their own offspring. So, some individuals are reproductively altruistic toward other members of their species, while others are reproductively selfish. The selfish individuals would receive a reproductive benefit from the altruists, while the altruists would pay a reproductive cost. So, on average, the selfish trait would be reproduced more than the altruistic trait, eventually replacing it.

Only traits that are instrumental to reproduction can be selected for. Evolution creates reproducing machines, not species-perpetuation machines.

The Genocentric Theory

I shall argue that the fundamental unit of selection, and therefore of self-interest, is not the species, nor the group, nor even, strictly, the individual. It is the gene, the unit of heredity.

— Richard Dawkins, The Selfish Gene

Dawkins’ notion of the selfish gene is that individual copies of an abstract gene will somehow “act” (inside organisms) for the “benefit” of the abstract gene. In this view, the abstract gene is the locus of purpose, and the purpose of an individual organism somehow derives from the purposes of its genes. Dawkins believes that an organism is a gene-survival machine.

The genocentric theory views the abstract gene as an entity with an actual purpose, composed of gene copies that are instrumental to that purpose. It views the relationship between the gene copy and the abstract gene as part-whole, rather than instance-type.

What is the selfish gene? It is not just one single physical bit of DNA. Just as in the primeval soup, it is all replicas of a particular bit of DNA, distributed throughout the world. If we allow ourselves the licence of talking about genes as if they had conscious aims, always reassuring ourselves that we could translate our sloppy language back into respectable terms if we wanted to, we can ask the question, what is a single selfish gene trying to do? It is trying to get more numerous in the gene pool. Basically it does this by helping to program the bodies in which it finds itself to survive and to reproduce. But now we are emphasizing that ‘it’ is a distributed agency, existing in many different individuals at once. The key point of this chapter is that a gene might be able to assist replicas of itself that are sitting in other bodies. If so, this would appear as individual altruism but it would be brought about by gene selfishness.

— Richard Dawkins, The Selfish Gene

First, Dawkins says that the selfish gene is “all replicas of a particular bit of DNA, distributed throughout the world”. In other words, it is all instances of an abstract gene. That is a class of objects, not an object or entity. He describes it as “a distributed agency”. But a class of objects has no agency, even metaphorically. When he talks about a gene doing something (assisting replicas of itself), he has switched to talking about a single instance as the metaphorical agent, instead of the abstract gene.

This is terribly confused.

There is a clear analogy between group selection theory and Dawkins’ notion of the selfish gene. In group selection theory, the species is viewed as an entity with a purpose. The species is a distributed agency that “acts” through its instances (organisms). In Dawkins’ view, the abstract gene is viewed as an entity with a purpose. The abstract gene is a distributed agency that acts through its instances (gene copies).

Both view a type as an entity with a purpose. Both view the instances of the type as parts of a whole, and instrumental to the purpose of the whole.

The abstract gene is a type, which consists only of information (a nucleotide sequence). It is less subjective than species, because it could be defined in very precise terms. But, like the species, the abstract gene is purely information. It is just a sequence of nucleotides that can exist in many different molecules and organisms.

Individual copies of a gene are not parts of the abstract gene, and they are not instrumental to its purpose. The abstract gene does not have an actual purpose or function. It has an abstract function within an abstract whole. A gene copy has an actual function within an organism, and that function is instrumental to the organism’s purpose of reproduction.

The gene copy does not serve the purpose of the abstract gene, any more than the individual organism serves the purpose of the species. Instead, the gene copy is instrumental to the organism’s purpose of reproduction. It is selected to contribute to reproduction.

It is hard to imagine a gene that, as an instance in an organism, somehow “acts” for the benefit of the abstract gene. Perhaps it would cause reproductive altruism toward other organisms with the same gene. But that is a very complex effect for a single gene to have, and almost impossible to arise as a single mutation.

For this imaginary gene, the scope of altruism would change over time. Initially, the gene would exist only as a single mutation. If it was replicated (by reproduction) enough times, then eventually it might exist in many individuals, perhaps in all members of a species, or even in the members of multiple species. How could it “act” to benefit its abstract type under all of those conditions?

If genes were selected to serve the interests of the abstract gene, then every gene would have its own purpose, and its own scope of altruism. Some genes are shared by all members of a species. Some are shared by all members of a genus. Some vary within a species. Some exist only in a few individuals. How could an organism be coherent if its genes had different purposes and scopes?

Genes are selected to contribute to reproduction. If a gene exists in many copies, that is because it has been copied many times by reproduction, and presumably contributes to reproduction. If a gene is perpetuated over time, that is because it continues to be copied by reproduction, and presumably contributes to reproduction.

Evolution creates reproducing machines, not gene-perpetuation machines.

But, But, But…

But if a gene has the property of making itself more common, then it will become more common. So, genes are selected to become more common.

— Someone on the internet

This is a fundamental misconception about evolution. The theory of evolution is not “a form that makes itself more common will be more common”.

First, there is a conceptual error. A gene can’t have the property of making itself more common. An abstract gene doesn’t do anything. It is just information. Its instances have effects in organisms. Those effects can be summed up by us, but that aggregate effect is not the “action” of the abstract gene. It is just an aggregate of individual effects. For example, an increase in the frequency of a gene is just the aggregate effect of many individual acts of reproduction by organisms.

Second, the theory of evolution is not just a tautology. It is not just saying “if X perpetuates itself, then X will be perpetuated”, or something like that. The theory of evolution describes the process that creates and maintains biological forms and their populations. Reproduction is the crux of that process. The theory of evolution explains how differential reproduction causes aggregate effects, such as the evolution of different species, and the distribution of species in an ecosystem.

Third, an abstract gene does not have the aggregate effect of becoming more common, most of the time. To be viewed as a gene, a mutation must have initially had that aggregate effect. Otherwise, it would not exist in multiple instances, and it would not be a type. So, if we view something as a gene (or allele of a gene), then it must have become common in the past. However, no gene increases in frequency forever. Eventually, it reaches a natural limit and stops increasing. Then it could have the aggregate effect of persisting for some time. Both of those aggregate effects are due to reproductive fitness. To become common and then persist for some time, a gene must positively contribute to the reproduction of individual organisms, because that is how it becomes common and persists.

Finally, if that argument were correct, then it would also apply to the species-centric theory. A species that has the aggregate effect of making itself more common (or perpetuating itself) would become more common (or perpetuate itself).

✦ ✦ ✦

This essay was very short and somewhat incomplete. I left out a lot of details, arguments and explanations. For the complete version, see the book Debunking the Selfish Gene.

Life is complex, and perhaps you can think of some possible exceptions to what I have said. People might think of bees, for example, or altruism within families. I cover some of that complexity in blog posts (Bees are Not Social, Kin Selection Theory is Wrong) and also in the book. In this essay, I focused on the core issues, not with every possible objection or counter-argument that might come up.

If you found this essay interesting, then please take the time to read the book.


  1. You might've conflated "abstract type" with concrete collections of instances. For example,
    Type: "integers", instance: "5", collection - "{5, 12, 8}".
    Type: "humans", instance: "you", collection: "your clan".
    Type: "ants", instance: "this ant", collection: "this ant colony".

    An abstract type can't have a specific purpose, but a collection of instances certainly can. Likewise, as you said, abstract organisms wouldn't have any evolutionary machinery to participate in some abstract species selection. But a concrete group of conspecifics in a particular time and place may have the evolutionary machinery to perceive each other as a collective, and to drive the organisms to participate in specific group selection mechanisms relative to that collective.

    Personally, I'm deeply skeptical of all three theories. It seems intuitive that evolution happens in some complicated tension between different loci of selection, rather than being neatly confined to one. Nature doesn't owe us any neat explanations for anything.

    I apologize if I'm missing some context. I watch your Youtube channel, but haven't read most of your posts.

    1. No, I haven't conflated anything.

      As you are using "collection", it seems to mean a physical organization of things, that are held together by a container or by bonds. I think you are conflating that with a type, such as a species or an abstract gene.

      A clan is linked by social bonds, so it has a structure. A society is a structure. A society can have a collective purpose, but it emerges from the individual purposes of its members, and it depends on the structure of society. Incentives align the interests of individuals with the interests of society. A clan is somewhere between society and family.

      Of course, an ant colony has a group purpose, because it is formed around the queen. I explain the biology of ants and bees in

      So, no, I'm not conflating anything. What you are calling a "collection" is an organized group. It is an analogous to a physical object, and it has a physical structure of some kind. The individual elements are causally linked in a way that creates a coherent whole.

      That is not the case for an abstract type, or the instances of an abstract type. The instances are simply defined by a common form (the type). They are not in a container or linked by bonds. A species is not a physical object. Its instances are not physically connected. Likewise for an abstract gene.

      Both group selection and the selfish gene concept are based on viewing a type as an entity, via a fallacious metaphor, and then viewing instances are parts of that entity.

    2. "Both group selection and the selfish gene concept are based on viewing a type as an entity, via a fallacious metaphor, and then viewing instances are parts of that entity."

      They're based on viewing a concrete population of organisms as an entity. (Which is what I meant by "collection" in this case). Viewing the population's instances as parts of that entity may be valid, if their DNA encodes them to perceive and respond to their concrete population as some sort of entity.

      i.e. I agree that viewing wolves as parts of some "wolf species" is invalid, but it's valid to view them as members of their wolf pack, which may or may not be the object of some group selection mechanisms vis-a-vis other wolves.

    3. No, they're not based on viewing a "concrete population" as an entity. Group selection views the species or subspecies as an entity. The selfish gene concept views the abstract gene as an entity.

      By "concrete population", you mean some kind of organized structure, such as a society. That is an extremely rare adaptation. Family units are more common, organized around male-female cooperation. Most species have neither. Very few species have anything like a wolf pack, let alone human society. Some have loose aggregations, such as herds. Some have strange quasi-family units, such as the hives of ants and bees. Each can be understood in terms of how it is organized and how it evolved. None of these structures require group selection theory to explain them (it can't explain anything), and group selection theory is not a theory of social structures.

      A "group selection mechanism" would not apply to something like a wolf pack. A wolf pack exists very briefly and is very small. The individual members have some freedom to leave or join, and they succeed or fail at reproduction as individuals. A pack is a social-family structure that emerges from the behavior of individual wolves.

      You're conflating the species-centric view with some vague notion that relates to social units, which you don't distinguish from other structures, such as hives.

      I suggest that you read the book.

  2. "Group selection views the species or subspecies as an entity. The selfish gene concept views the abstract gene as an entity."
    In every context in which I've seen the term "group selection" (I've worked in computational genomics for a large chunk of my career), it was in the context of a given population, e.g. a certain colony of microbes or mice, not all microbes or mice of that species - that would make zero sense of course. If someone defined group selection by "species or subspecies", they chose their words very poorly, or used it as a pop sci shorthand, as biologists often do. Likewise, viewing some single "gene" as a vehicle of evolution is another pop sci shorthand. Genes and groups of genes co-evolve in stupidly complicated ways, this is taken for granted by any geneticist.

    "By "concrete population", you mean some kind of organized structure, such as a society."
    I mean exactly those social units which the given organism is genetically programmed to form and function within, and which they naturally create. Exactly that - nothing more, nothing less. Depending on the creature, this may be a family, a pack, a herd, a hive, a colony. As you said, many organisms have nothing at all.

    Each member of such a group is keyed into it by genetic mechanisms that govern these group dynamics. Insofar as these mechanisms are under selection pressure, it's self-evident that "group selection" occurs.

    Obviously, it would be absurd to claim that "group selection" accounts for all, or most, of evolution. As you said, groups just as easily fall apart, recombine, etc. A huge amount of selection takes place on other levels of existence as well. Likewise, it would be obscenely reductionist to approach evolution solely through individual reproduction or by viewing particular genes as vehicles of evolution. All three approaches are insufficient to explain the fullness of the data by themselves.

    1. First, I clearly defined the three views of biological purpose. If you want to talk about some other notion of group selection that is not at the level of the species, it's off-topic.

      Organisms can have traits that make them form social or family structures. (A hive is not a social structure.) Regardless, those traits will always be selected to contribute to reproduction, and almost always at the level of what we view as the organism. For example, wolves have the innate ability to form packs. They do this because it benefits them reproductively as individuals. Young wolves leave their natal pack. The dogs in a neighborhood can form a pack. This has nothing to do with shared genes or group selection.

      If your hollow-man version of group selection theory is simply "some organisms form groups and those groups affect reproduction", then it's just meaningless. Yes, humans evolved to live in social group. That doesn't mean that the group is the unit of selection. Humans did not evolve to work toward the perpetuation or replication of the group. They evolved to be selfish reproducers, and their social behavior is based on individual selfishness.

      If you want, we could debate in voice, as long as I can record it for others to listen to.

    2. "First, I clearly defined the three views of biological purpose. If you want to talk about some other notion of group selection that is not at the level of the species, it's off-topic."
      That's fair enough. I confess myself confused though - who defined a notion group selection at species level? I never heard of this notion being widely held. When you argue against this view, who do you argue against? AFAIK, "Group selection" always pertained to "groups", not species.

      "The dogs in a neighborhood can form a pack. This has nothing to do with shared genes or group selection."
      It requires "shared genes" with regard to collective function; an analogy would be an internet protocol, where both sides need to have very similar code in order to mutually understand their formats and relate to each other. A dog can't really form a pack with a cow or a tree. This necessarily requires such shared machinery to undergo selection, at least in part, at the group level, because its adaptiveness for survival is tested in a "multiplayer" setting.

      "That doesn't mean that the group is the unit of selection."
      The group isn't the unit of selection, neither is the individual or the gene. There is no unit of selection. "Unit of selection" is a reductionist human abstraction that nature doesn't care about.

      "Humans did not evolve to work toward the perpetuation or replication of the group. They evolved to be selfish reproducers, and their social behavior is based on individual selfishness."
      If I understand correctly, this is based on your contention that altruism can't evolve because it would be automatically selected against in favor of the selfish. I disagree with this claim, because it doesn't account for all the other behaviors that co-evolve with altruism. As much as people do this for simplicity, the adaptiveness of a behavior can't be considered in isolation: every behavior co-evolves with other behaviors in absurdly complex ways. However, that whole can of worms is probably off-topic. As regards to the topic at hand, I would definitely assert that human social instincts can't be reduced to reproductive selfishness.

      "If you want, we could debate in voice, as long as I can record it for others to listen to."
      With gratitude, I'll decline. Live debates are not my thing - I'm not verbally articulate or quick on my feet, so I generally trip over myself. But out of respect for your time, we can draw a line under this discussion.

    3. See:

      The original notion of group selection theory was at the level of the species, and Dawkins spends some time talking about that view and debunking it in the first chapter of "The Selfish Gene". Since it has been discredited, maybe some academic biologists have tried to revive the term with another meaning, or with no meaning, to protect it from falsification. There's a lot of BS in academic biology.

      We are talking about the basis of selection. You are conflating social behavior with group selection. Those are totally different things. Social behavior does not require group selection. Cooperation is a social behavior, which has benefits for individuals. The form of the group is a manifestation of the individual-level traits, which were selected to make individuals reproduce, not to make the group "succeed".

      Sexual cooperation is a simple example. Males and females evolved to fit together, but not to be altruistic toward each other. Male and female traits did not evolve to contribute to the fitness of the family, but to contribute to individual reproductive fitness. That's why both sides can evolve mechanisms to defect on sexual cooperation.

      The unit of selection is as meaningful as "star", "electron", f = ma and other theoretical abstractions. It has explanatory power. The form of an organism has been selected to have an effect -- what effect? Reproduction, which is an individual-level effect. That's the fundamental point.

      Are you making an argument against it? No. I see no argument, no alternative theory. Just some vague appeal to complexity/ignorance.

      "I disagree with this claim, because it doesn't account for all the other behaviors that co-evolve with altruism."

      Again, no argument.

      There are no human behaviors that would be explained by altruism.

    4. "You are conflating social behavior with group selection."
      No, I'm claiming that social behavior requires group selection.
      For example, a group of animals communicate by some specific set of signals. Even though the mutations that shape this machinery spread from individuals, it is tested for adaptiveness at the group level. The survival of the communication system's participants is affected by the collective effectiveness of the system itself within a given environment, not just their individual experience of it.

      "Are you making an argument against it? No. I see no argument, no alternative theory. Just some vague appeal to complexity/ignorance."
      First, falsification of a theory doesn't require an alternative theory. I'm happy to admit ignorance.
      Secondly, no - a "unit of selection" is not as meaningful as a "star", "electron" or "f=ma". A more apt analogy would be someone claiming that there is a meaningful "gene for intelligence". No, there is a tangled many-to-many relationship between hundreds of genes and hundreds of phenotypic traits that bear on cognitive ability in some convoluted way. If someone blithely says they're just gonna cut this Gordian knot by claiming that these 200 base pairs on chromosome 5 are the "IQ gene", this would not be a meaningful theory. So yes, I'm making an argument against it. An "appeal to complexity" is perfectly valid when the proposed theory is far too reductionist to be useful, and likely to produce nonsense when used in any downstream application.

      "There are no human behaviors that would be explained by altruism."
      Given that humans display altruistic behavior, I assume you meant that all such behaviors are actually explained by some consequent increase to the individual's reproductive fitness, as opposed to some impulse towards altruism itself.
      Alternatively, such behaviors can simply be explained by altruism. Given that humans display a wide range of behaviors - from obvious reproductive selfishness to self-sacrifice with absolutely no realistic expectation of ever reproducing - it's reasonable to assume that both reproductive selfishness and altruism co-exist in tension within individual humans, and their expression is highly situational. They can co-exist just fine, because neither level represents a "unit of selection" at the exclusion of the other.
      This multivariate model explains the full range of human behavior without any weird mental gymnastics. I don't understand the need to shoehorn everything into a univariate model - some sort of "choose one out of three" - which requires tortured rationalizations for all cases that don't fit that model. Selection happens just fine on multiple levels. You're not under any obligation to choose between them, and their combination allows for models with far more explanatory power.

    5. Social behavior does not require group selection, any more than male-female sexual cooperation requires "family selection". The genes for social behavior are not tested at the group level, just at the individual level. You are conflating the two, because you're just leaping from "social behavior exists" to "group selection exists", when there is no logical relationship between the two. The genes are tested by reproduction. They only replicate if the individual reproduces. If a gene contributes to group success but causes individual reproductive failure, it will be eliminated by selection.

      You are also contradicting yourself, because you reject the "unit of selection", which is a presupposition of group selection. You could believe that there are multiple units of selection, but that's not the same as rejecting the notion.

      Wrt the unit of selection, it is at the level of abstraction of "electron" or "f = ma" and it has similar explanatory power. I make the argument that there is a single unit of selection and a single biological purpose. I show how that explains the data of biology. You seem to believe in some kind of inverse Occam's razor, by which the more complex explanation is superior, even if it has less explanatory power. Simpler theories have more explanatory power, because they are more constrained. They are easier to falsify.

      Vague hand-waving at complexity/ignorance has no explanatory power. It's just a lazy criticism that you can throw at anything. It's a debating tactic. You certainly haven't falsified what I am saying.

      Humans exhibit lots of different behaviors, many of which are not adaptive in the modern world. However, all of those behaviors can be explained as caused by mechanisms that evolved to contribute to reproductive fitness. Those mechanisms are not perfect. E.g. the emotions are heuristic, ad hoc and stimulus dependent. People use birth control to prevent reproduction. That can be explained by us having the instinct of lust, but no "reproduction drive".

      I talk about this a lot. See:

      There is no general pattern of altruism among human beings, which requires explanation. Again, there is social behavior, which is explained by mechanisms that evolved to generate reproductively selfish behavior. Society does not require altruism.

      I discuss altruism and selfishness in detail here:

      I explain the game-theoretic basis of society here:

      I explain the distinction between family and society here:

      A theory that predicts nothing has no explanatory power. You haven't actually presented a theory, however, just waved your hand at complexity. My theory predicts that biological forms will be explicable in terms of reproductive fitness, and I back that up by providing such explanations. I go through many examples in the book, provide explanations, and show how they are not explained by the other theories.

      I debunk kin selection theory here:

  3. It says my comment is too long :)
    Let me address the abstraction/theory question first. Then I'll do a bit of a deep dive on group selection, since there seems to be some genuine confusion there.

    Here's a more direct analogy. Consider the age-old question: is some human behavior or outcome caused by nature or nurture? The serious answer, in practically every case, is "almost certainly a combination of both. In highly variable proportion depending on the situation".
    Now, suppose someone turns up and frames this discussion as follows: There are two theories: 1) fate is driven by nature 2) fate is driven by nurture. He introduces an abstraction that he calls "locus of fate", which is the driver. He then claims that the "locus of fate" is nature, not nurture (or vice versa). He then defends his theory by debunking the other half of his false dichotomy.

    Clearly, this is nonsense: both nature and nurture are drivers, he's being reductionist, and his "locus of fate" abstraction is a contrivance that only serves to confuse the issue - it's not a neutron or a gravitational constant, it doesn't grasp at any underlying reality.

    He might object that you haven't offered an alternative theory, you're just appealing to complexity and ignorance. He might claim that his univariate analysis complies with Occam's razor, while your bivariate muddling violates it.
    If he's a blank slate lunatic, he will contrive an environmental explanation for each data point, then claim that his theory has explanatory power.
    If he's a biological determinism lunatic, he will contrive a genetic explanation for each data point, then claim that his theory has explanatory power.
    However, as you can imagine (in fact, you don't have to imagine), either one of these two theories results in a torrent of absurdities if widely accepted.
    Anyway, I think you get the point. Occam's razor doesn't favor the simplest theory - it favors the simplest theory that adequately explains the data, and no simpler.

    1. Human behavior is not caused by a mixture of nature and nurture. That's incredibly naive. All expressed traits depend on the environment and genes. So, anyone framing any discussion in that way is just naive.

      When it comes to "nature versus nurture", no intelligent person proposes that genes cause outcomes in isolation. The question is about whether differences in outcomes are best explained by differences in genes. There are principled ways to answer that question.

      You are trying to analogize my theory to an absurd position, rather than actually making a counter argument, or even trying to understand it.

      Here's a better analogy. Newton claims that the orbit of the moon can be explained entirely in terms of gravity and his laws of motion. Someone else says "Oh no, we need different theories to explain heavenly motions and the motions of objects on the Earth. Your theory is too reductionistic."

      Of course, Newton's theory is superior, because it explains both types of phenomena by reducing them to common laws. It reduces the complexity of the phenomena, as a scientific theory should. That's the point of a scientific theory: to explain, not to generate verbiage.

      If you can pick and choose which "law" to apply, then there is no law. A mixed theory would not explain anything, unless there was a principled way (a higher law) that defined when one of the lesser principles applies.

      All biological data fit the phenocentric theory. There is no need to posit multiple bases of selection. I explain that in detail in the book.

  4. Regarding group selection.
    "Selection at the group level" and "group is a unit of selection" are two fundamentally different things, and the first does not imply the second. This will become more clear as I go on.

    Let me illustrate the general concepts with a toy example.
    Suppose you are an organism in a given environment, which selects for aggression at the individual level. All else being equal, a more aggressive specimen can expect higher reproductive fitness.
    Suppose you have two offspring. One is more aggressive, thus individually more fit. The other more docile, less fit.
    Consequently, the aggressive one produces 10 offspring, the docile one produces 5.
    Naturally, just as before, the fitness of these 15 individuals is also affected at the individual level by their respective aggressiveness.

    Here's where the nuance comes in.
    Each individual's fitness is also affected by the aggressiveness of the individuals it interacts with, and this "interaction neighborhood" varies between individuals.
    This is because organisms within a population almost universally experience more interaction with closer relatives than more distant relatives - they don't experience interactions uniformly at random across the whole population. The most trivial driver of this gradient is spatial proximity.
    Probabilistically, the 10 aggressive siblings interact more densely with each other than with their cousins, and the same goes for the 5 docile siblings.

    This produces a counterintuitive result: an aggressive's fitness is improved by his heightened aggression at the individual level, but simultaneously reduced by his similarly aggressive brothers in his interaction neighborhood. The opposite is true for the dociles. As a result, the 5 dociles will now experience better relative fitness than the 10:5 of the previous generation.
    Thus, we say that there is individual-level selection for aggressiveness, and group-level selection against aggressiveness.
    This prevents the naive scenario where the population just keeps getting more and more aggressive with each generation. These two opposing forces, acting at different levels of selection, bring the overall population aggressiveness towards some optimal distribution for that environment.

    Paradoxically, even though more aggression always improves any given individual's fitness and is always favored by selection at the individual level, it is countered in the population by the selection forces it generates at the group level.

    1. > "Selection at the group level" and "group is a unit of selection" are two fundamentally different things, and the first does not imply the second.

      No, they are the same. Selection is for some outcome. Is it for a group outcome or an individual outcome? That's what the unit of selection defines.

      Your toy example is one that I cover in detail in the book. It is also explained in the blog post "Kin Selection Theory is Wrong":

      Some level of sibling altruism can evolve, and I explain that in the book. It is not based on group selection. It is based on the extended phenotype. The genes of the parent are expressed in the children, so the children are part of the extended phenotype of the parent, especially if there is a nest.

      Group selection cannot explain this, and "group selection" is not about the family, but about larger aggregates, such as species or subspecies.

      You also assume that "aggressiveness" is a generic trait, when in fact most species, from bees to humans, have mechanisms to recognize family members. Children know their parents, and vice versa. Incest avoidance is a good example in humans. People did not evolve to be "nice" or "less aggressive", but we did evolve to be nice to our children, and to tolerate our siblings. The phenocentric theory explains why such mechanisms are necessary.

  5. This is what is generally meant by "group-level selection effects", as it is used in genomic research, at least where I've been.
    Important points:
    First, these group-level selection effects do not require any segregation between groups, or any form of organization within groups. In fact, they don't require any "groups" to even exist. They only require the condition that, on average, individuals experience denser interaction with genetically closer individuals, which is true for almost any species. This condition is sufficient for these group-level selection effects to come into play, even absent any meaningfully articulated "groups". Thus, as I said, notions of "group as a unit of selection" have nothing to do with any of this. The word "group" is somewhat misleading here. One might simply think of this as any selection effects happening above the individual level.

    Secondly, given a genomic dataset, individual and group-level effects at some locus can be delineated mathematically, the latter being shown to exist. The presence of selection effects above the individual level is a demonstrable empirical fact. It's not a subject of theoretical speculation or a product of toy scenarios. Even theoretically, their existence is trivially obvious: by reproduction, an individual behavior becomes a group-level selection force acting in the vicinity of the reproducers. It would be extraordinary for this to somehow not happen. Thus, your fitness and evolution is not just shaped by your individual adaptations, but also by the emergent echoes of adaptations bouncing around in your "group", however shifting and amorphous that may be. Furthermore, group-level selection can be easily demonstrated experimentally, and even shown to overpower individual-level selection in some cases. Here are some examples, even though the paper is very old:

    Thirdly, as mentioned before, this mechanism explains how prosocial behavior can persist, without being automatically destroyed by the costs it inflicts on any individual trying to manifest it. Generally speaking, antisocial behaviors are selected for at the individual level, and selected against at the group level. Conversely, prosocial behaviors are selected against at the individual level, but selected for at the group level. This naturally produces a convoluted tension between all sorts of behaviors.

    In conclusion. "Is it nature or nurture?" Almost certainly both. "Is it individual selection or group-level selection?" Almost certainly both. You pressed me for an "alternative theory", and it's simply this: it's both. tied together in all sorts of bizarre feedback loops. Occam's razor is overstepped by admitting only one of them, because both empirically happen and a serious theory must account for both.

    In actual practice, this is all a world of messy PCAs and regressions. There's no elegant simplicity to look for, as nice as that would be.

    I apologize for the run-on post. I hope it was all vaguely understandable. I'll wrap it up there.

    1. Your vague notions of group selection are simply bogus, and protected by their lack of predictive content. I'll give you a simple example. Kin selection theory "predicts" sibling altruism, right? So it explains why worker bees work altruistically for their sisters. That's what some people might claim. However, queen bees in the same hive try to kill each other. They are identically related as workers, yet they try to kill each other. If kin selection theory "explains" the worker altruism, it would be falsified by the queens killing each other. As a theory, it has no predictive or explanatory power. It cannot explain sibling altruism unless it is falsified by sibling competition. In fact, it explains nothing.

      A theory that says "something will happen" has no predictive power. A theory that says "individuals will sometimes act for the good of the species/group/whatever and sometimes act in their own reproductive interests" has no predictive power. It is meaningless.

      I make clear theoretical claims and defend them. I don't wave my hand at "messy PCA and regressions", etc. That's an appeal to ignorance/complexity. It's not the intellectual high ground.

  6. I am reading your book about this topic, it is amazing. Strictly nominalistic evolutionary theory.
    Do you think that species are simply a bunch of individuals that resemble each other, and similarity also comes from subjective attribution?Have you read work done by philosopher David K. Lewis?

    1. I hope you find it interesting. It's not "nominalistic". That's a bogus concept from a medieval dispute ("the problem of universals"). The dispute is based on a false dichotomy and naive assumptions about knowledge. So, if you think it's nominalistic, then maybe you haven't read very far into it, or understood what I'm saying.

      > Do you think that species are simply a bunch of individuals that resemble each other, and similarity also comes from subjective attribution?

      How could individuals resemble each other, if their similarity came from only subjective attribution? If that were the case, their only "similarity" would be the name that is applied to them.

      If types/forms were just names, there'd be no way to recognize them. No, there is real order to reality, and our brains induce knowledge that reflects that order. The patterns that we recognize are not purely objective, nor purely subjective. There must be an order to reality, or our brains could not compress the information of experience and generate abstractions. But those abstractions are not reality itself. See "Theories of Knowledge" on this blog.

      No, I haven't read anything by David Lewis.

  7. My native language is Chinese, so my English expression may not be clear.

    I think I understand your core position in the book. I do not support the anti-realist version of nominalism. I also maintain that both order and pattern are real. There is no disagreement on this point.

    What I say is nominalism mainly means that it advocates that the discussion of types can ultimately be reduced to the discussion of individuals/instances. The former depends on the latter. In principle, we can discuss evolutionary issues without resorting to the concept of types. Although it does not necessary.

    Evolution occurs at the level of instances/particularities/individuals, we can think of instances as a unified whole at all levels of our discussion, Dawkins makes a mistake by pitting genetic instances against the individuals/instances that carry them.And our discovery of order also seems to be grounded in the cognitive relationships we have with a sequence of individuals in experience.

    1. I replied to your earlier comment, but it was deleted. I think it was the same as the current one. My reply:

      Ah, sorry if I was abrupt.

      I use the term "representationalism" to describe my theory of knowledge. In that theory, we induce conceptual knowledge from experience. Concepts reflect real order, but the concept is not the order itself, any more than a perception of a tree is the tree itself.

      I don't think we can discuss things at the level of instances -- we can only think about things via ideas, and ideas depend on abstract concepts, which are (essentially) patterns induced from experience. E.g. in the classic example of a horse, we can talk about "a horse" (an instance of the type horse). There is no way to talk or think about an instance without invoking a type. So, I don't think we can talk about evolution without talking about types. Evolution is a theory of how types emerge, and it involves the inheritance of type/form.

      Our discovery of order (induction of concepts) does arise from our experiences of things, and how they interact with our senses, emotions and muscles (what I call "semex"). It is ultimately grounded in experience and how the brain processes experience. So, I think we agree on that, but your view of knowledge might be somewhat different. I have a number of blog posts on knowledge, if you're interested.

      Clearly, you have spent some time thinking about these questions. If you ever want to chat in voice, let me know.

    2. This comment has been removed by the author.

  8. I think in principle it can, if the abstract concept comes from the induction of instances.We can translate the abstract concept into the relationship between individual instances from which we get this abstract concept.

    Of course this would become so cumbersome as to be completely unnecessary. But I think this possibility in principle indicates the ontological priority of the individual,so it's not trivial, although it's irrelevant to a discussion of any specific issues.

    Thanks for the invitation, but my English speaking and listening skills lag far behind my reading skills and don't yet support a good oral communication, let alone a discussion of theoretical biology .And I'm also a person who prefers and is good at communicating with words

    1. I noticed that you were having some trouble with the comments, so I thought voice might be easier.

      To be clear, the phenocentric theory doesn't depend on any philosophical claims about knowledge. It's in the domain of science, which presupposes our ability to recognize both instances and types. The errors of the other theories arise from sloppy, metaphorical thinking, not from a philosophical position. When I say "a type is not an object", I mean it is not a causal object, such as a rock, organism or society. It's just a category. In Platonism, a category is an abstract object that exists in some other realm of existence, the Platonic realm. I think that's nonsense, but I'm not referring to the Platonic form by "object" when I say "a type is not an object".

      People often think metaphorically about types/forms as objects, because we naturally use more concrete metaphors to think about more abstract things. So, people metaphorically project the properties of objects onto types, and thus confuse themselves. It is easier to think about a single wolf, or a wolf pack, than the abstract type "wolf". It is easy to project the properties of an organism onto a species, or to view a species as a super-organism consisting of its individual members. Those are both conceptual errors. (And scientific errors, not philosophical errors.)

      Dawkins rejects that notion that a species is a super-organism, but he makes almost exactly the same error in viewing the abstract gene as a super-organism consisting of its copies.

      Back to the theory of knowledge (another interest of mine), we can't reduce or replace abstract concepts with the relationship between instances, because the abstract concept is our representation of their relationship. An abstract concept is a regularity of appearance/interaction. It is caused by an objective regularity. We only know that regularity through the concept, just as we only know individual objects through their appearance/interaction, and how we categorize them based on that appearance/interaction.

      For example, there is a regularity of nature that causes us to induce the concept "horse" from our interactions with the world. We have no direct access to that objective regularity. Our minds only access the concept. The brain uses that stored concept to interpret interactions. E.g., when you see a horse, the brain recognizes a pattern, activates the concept "horse", and uses that concept to interpret your experience and drive action.

      We are not communing with the Platonic realm of forms when we recognize a horse, nor are we arbitrarily applying a label. The label is applied based on pattern-recognition, and the pattern was induced from past experience. We have machine-learning algorithms that do this, so we have a pretty good idea of how it works.

      I present this view of knowledge in "Theories of Knowledge" (on this blog), and I contrast it with other views. I'm trying to finish up a book on philosophical topics, which will cover it in more detail.

  9. I think your theory might also be called repro(duction)centric theory.

    I find it difficult to distinguish which evolutionary theory is more correct from the written description, and it can even give the illusion that the pluralist explanation is correct, but with math we can tell which theory is right.

    Another thing is to avoid anthropomorphic thinking. In fact, the evolutionary explanation is a retrospective one (how things in the present have been preserved by natural selection sifting through history), and from a retrospective perspective, subject with psychology/consciousness/free will can also be seen as mechanistic/deterministic: psychology/consciousness/free will exists only in the first-person perspective of the present moment. The psychology/consciousness/free will we experience in the present moment can also be described mechanistically/deterministically in the eyes of future evolutionary biologists.

    Retroactivity is a key issue that doesn't seem to be taken seriously by the philosophy of biology .

    1. It's not trivial to understand evolution, because it is an abstract process that operates on information, and we can't visualize it. So, people often use misleading metaphors and arrive at false conclusions.

      And of course, if the phenocentric theory was obvious, then I wouldn't have to write this essay. Everyone would already believe it. Any important idea that is not widely accepted must be hard to understand or accept in some way.

      Having said that, it isn't that hard to determine which theory is correct. It just requires careful thinking about processes. I give a relatively simple thought experiment which proves that the species-centric theory is incorrect. The same thought experiment rules out gene-based altruism as well.

      Math can't help us -- evolutionary theory isn't a mathematical problem. Evolution is a process, and requires causal reasoning. Math depends on conceptual understanding. There's no point doing math based on false assumptions. Math is something you use once you have the right conceptual foundation, if the concepts are quantifiable and measurable.

      I have an essay on free will:

      The free will | determinism paradox is caused by subject | object dissonance. When we think of ourselves as objects, subjectivity seems to disappear, because we are viewing ourselves from "outside", through abstractions. I don't think it has anything to do with retroactivity. The paradox exists for past and future choices. It is one of a family of paradoxes due to subject | object dissonance.


Post a Comment