Books and Reviews


Debate on Larmarckism in Social Evolution

In June 2004, Geoffrey Hodgson and Thorbjorn Knudsen circulated a paper under a title of The Limits of Larmarckism Revisited. What followed was a very interesting debate provoked by a response from Richard Nelson. Geoffrey Hodgson, Thorbjorn Knudsen, John van Wyhe, Marion Blute, John Gowdy, and David Hull all participated in this high quality debate that took place within a few days. Anyone interested in evolutionary thought in the social sciences will profit from reading this exchange. Comments on the debate are welcome. (The Hodgson and Knudsen paper is now published in the Journal of Evolutionary Economics. Download here.)

Review by Richard Nelson

From: Richard R. Nelson
Sent: Friday, May 28, 2004 1:06 PM
Subject: Re: The Limits of Lamarckism

Dear Geoff and Thorbjorn,

I wish I could applaud and agree with the position you stake out in this essay [The Limits of Lamarckism], but I can’t (which I am sure you suspected, Geoff). I remarked about an earlier draft that the two of you seemed hung up on biological concepts and language.  You protested, Geoff, but I will say it again. You are still hung up.

First of all, I certainly agree that there is a category difference between a cultural element or trait, and the individuals, organizations, or social groups that adhere to that element or trait. But it is not all that clear that there always is a clear distinction between a cultural trait, and the behavior or action of an individual or organization. Let me stay with Veblen’s notion of widespread habits of action or thought as a definition of institutions, or culture. (Or do you want to make a clean distinction between these?). Consider the custom of shaking hands as a greeting as an element of culture (a habit of action). Do you really want to make a sharp distinction between a disposition to shake hands, and the fact that members of the society mostly shake hands, and are surprised when their gesture is not reciprocated? While there is a clear divide between a “habit of thought”, and the action that that habit tends to invoke under appropriate circumstances, I don’t think you can make such a clear divide regarding habits of action.

Regarding David Hull’s argument that a learned pattern of behavior, that improves performance in a particular environment, that is taught to or imitated by another person or organization, is not an example of a Lamarckian process, but rather involves “contagion”, that strikes me as word play, and blindness to what social scientists like me mean when we say that cultural evolution has Lamarckian elements..

Sure, if only actions or displays or explanations (behaviors) can be taught or imitated, and not dispositions per se, then there is not a direct route from disposition to disposition. But as I said above, the disposition–standard way of doing something difference strikes me as very forced. And even if you are committed to it, why do you have trouble with the notion that imitation of behaviors (displays of habits of action) or articulated beliefs (habits of thought) is not what most people (like me) mean when they say the process is Lamarckian? Or are you saying you know that is what we mean, but you dispute that that can be described as Lamarckian? If the latter, why do you think your use of the term is more legitimate than mine?

Look, those of us who have been arguing that cultural evolution has a Lamarckian element mean by that the following. First, innovations often come about from conscious, and intelligent, efforts to find better ways of doing something. Second, that the successful innovations get taught to or imitated by others in the broad society. No denying Darwinian mechanisms here. No arguing that anything biological is involved. No argument against the proportion that habits (routines) define propensities, rather than something that is always active.

But you know that. So what, then, is your argument against that kind of Lamarckianism, which is what the discussion is really about?

Best regards, along with my disagreement…..Dick

From: Geoffrey Hodgson
Sent: Friday, May 28, 2004 3:24 PM
Subject: The Limits of Lamarckism

Dear Dick

Many thanks for your comments on our “Limits to Lamarckism” paper. You accuse us of being “hung up on biological concepts”. I’d respond as follows.

You agree that socio–economic evolution is Darwinian. To understand this we need to unpack Darwinism. Thorbjorn and believe that the genotype–phenotype distinction is part of the package. This is not being “hung up on biological concepts”.

Furthermore, we argue in our paper that those that call socio–economic evolution “Lamarckian” are importing into the social sciences biological concepts that simply do not transfer. Accordingly, descriptions of socio–economic evolution as “Lamarckian” are more hung up on specifically biological concepts than those that use the “Darwinian” label in its core sense of variation–inheritance–selection.

Alternatively, if it is argued that the genotype–phenotype distinction does not apply to socio–economic evolution, then it is simply impossible to meaningfully describe such evolution as “Lamarckian” because this notion of evolution requires such a distinction to make sense.

We are fully aware that operationalizing the genotype–phenotype distinction in socio–economic evolution is difficult. Given that the required conceptual work is not yet fully done, the possibility success cannot yet in principle be ruled out or in. Attempts have to be made, before they can be shown to fail, if indeed they do fail.

I do want to make a very clear distinction between a disposition to do something and the action itself. The need for this distinction is very well established in philosophy since Aristotle. I do not see a big problem in enforcing it.

You may wish to describe imitation in some loose sense as Lamarckian. The point here is that imitation is not the INHERITANCE of acquired characteristics. So it is not Lamarckian in the sense that the term is widely used. If imitation is Lamarckian (in your loose sense) then Lamarckism applies to biology as well as to social evolution. Animals imitate each other. Like you, Thorbjorn and I believe that imitation and learning are very important in society, but we are unconvinced that such a broad definition and use of the term Lamarckism serves any useful purpose. On the contrary, it seems to add to the confusion.

Nevertheless, your thoughtful remarks are very much appreciated.

Best wishes


From: Thorbjorn Knudsen
Sent: Friday, May 28, 2004 8:33 PM
Subject: Svar: Re: The Limits of Lamarckism

Dear Dick

Many thanks for your comments!

I think there is a lot gained by understanding exactly the details of replication processes in society and the economy. Exactly how are habits and routines transmitted? What transmission processes promote or hinder their reliable transfer? What is the nature of selection processes that operate on habits and routines? Such questions are driving what Geoff and I have been trying to do. I am concerned that concepts such as phenotype/genotype and perhaps even interactor/replicator cloud the message. For some reason the concept of selection (even natural selection) seems less problematic to most in the social sciences!

It seems highly useful in thinking about economic and social evolution to have a concept of selection. It also seems useful to have concepts of the properties of entities that are critical to understanding selection (phenotype/ genotype and interactor/replicator). These concepts should be refined and substituted for better alternatives if they were available. Perhaps you are right that Geoff and I have got hung up on biological concepts and language. The problem is that there seem to be no useful alternative concepts around that cover the properties of entities that are critical to understanding selection in the social realm.

Regarding the distinction between disposition and behaviour (or thought), I see it as a useful way to think about an instruction set that produces contingent behaviour (or thought). From an empirical perspective, dispositions are bothersome, however. It would be nice to find a way to tap their content.

I certainly agree that innovations often come about from conscious, and intelligent, efforts to find better ways of doing something, Also, that the successful innovations get taught to or imitated by others in the broad society. We might call this Lamarckism in a very broad sense, but why? Rather than borrowing the term Lamarckism (from biology) or perhaps associating this term with new meaning, we might call it something else. Perhaps we should consider the nature of the underlying transmission and selection processes before we agree what is going on and what to call it.

All the best

From: Richard R. Nelson
Sent: Saturday, May 29, 2004 7:40 AM
To: Thorbjorn Knudsen; Geoffrey Hodgson
Subject: Re: Re: The Limits of Lamarckism

Dear Tjorbjorn,

Thanks for your note. I am in accord with much in it.

I certainly agree that to develop a serious theory of cultural, or social, or economic, evolution, we must get into the details, and get them right, I agree that the concept of selection is key here, and that we need to understand just what is selected for or against, and how. Indeed, I have staked my whole professional career on the belief that building a solid evolutionary theory was the way towards progress in economics, and the other social sciences.

However, when the term “evolution” is invoked, most people think of biological analogies. To make that progress on an evolutionary social science therefore requires that we come to grips with how cultural evolutionary processes are different from, as well as similar to, Darwinian biological evolutionary processes. One way to proceed on that is to try to understand what is general about a “Darwinian” process, and what may be special to biology.

It is clear that one can specify an evolutionary process, in the spirit of Darwin, without building in a distinction between a genotype and a phenotype, or a cultural trait and the entities that operate influenced by that trait. All one needs is a mechanism by which entities with different durable or semi–durable characteristics act in an environment in which those with certain characteristics are more fit, tend to expand more or contract less, than those with other characteristics. I developed exactly such a model in my 1968 AER piece, where I proposed that the industrial development process in underdeveloped economies was one in which “modern” firms entered the industry, were more profitable than “traditional” firms, and drove the latter from the market. I later found out that in that paper I had rediscovered Fisher’s equation. No need in Fisher’s formulation for a distinction between trait and entity; the entities are defined by their traits. Similarly in many models in evolutionary biology, Maynard Smith’s game theory formulation, for example.

A distinction between trait and organism was forced on Darwin empirically, not logically. Organisms die off. Their traits appear to be carried forward in their offspring. Thus to make evolution work, given these empirical facts, there must be some mechanism of trait transmission, or put another way, traits need to be seen as entities in their own right, not locked into particular organisms. Later, biologists came to understand genetics, and the presence of the Weismann barrier. They built this into their theories of biological evolution. But they are there in modern theory not because they are a logical necessity in a general Darwinian evolutionary theory, but to fit with the empirical facts that are known about biological evolution.

Because of the empirical facts of the matter, not logical necessity, in a theory of cultural evolution we need to distinguish cultural traits from the entities that operate under their influence at any time. We need to do that because we see, first, entities changing their traits (habits of action and thought, if you will). Second, because we see entities picking up the traits of other entities. Thus while an evolutionary theory with no distinction between entities and traits is logically coherent, for empirical reasons that is not a good theory of cultural evolution.

Is there a distinction between a cultural trait as a disposition, and characteristics of the entity that is influenced by that cultural trait? Sure, if one wants to make that distinction. However, I am not yet persuaded that the distinction serves a useful purpose. In your piece with Geoff, the argument for it is largely that it makes “habits” more gene–like. But I am not persuaded that there is good reason to do this. For empirical reasons we need to have a distinction between cultural trait and entity influenced by traits in our theory. This sounds analogous to a genotype phenotype distinction. Indeed Sid and I are on record as proposing the analogy (using the term routines rather than habits). But there are a lot of differences. I do not want to push the analogy too far.

Why call a process that involves conscious learning and imitation “Lamarckian”? Why not? For some purposes the analogy is useful. Of course there are differences. But there are differences between how Darwinian mechanisms work on culture and how they work on species. Why call the processes of cultural evolution Darwinian? Because there are interesting analogies, not because they are exactly the same

I hope you are finding this exchange of thoughts useful. I know I am.
Best regards Dick

From: Thorbjorn Knudsen
Sent: Sunday, May 30, 2004 11:54 AM
Subject: Necessity of G–P distinction?

Dear Dick

Many thanks for your inspiring and useful email. As you can see from the following, I agree with most of what you said. I also draw some implications. Apologies for the lengthy response.

According to Fisher’s fundamental theorem, the rate of change in fitness of any organism at any time is equal to its genetic variance in fitness at that time. We might wish to focus solely on the phenotypic level. At this level, the rate of change in fitness of any organism at any time is equal to its phenotypic variance in fitness at that time. Apparently there is no need for the Genotype–Phenotype distinction here (G–P distinction in the following). If we roughly know how transmission (replication) influences the distribution of phenotypic traits, we might focus solely on these traits. Selection improves the average fitness of the population, but also reduces the variance in the distribution of traits.

I have now read your 1968 AER paper with pleasure and profit. It studies the path to equilibrium after the innovation and subsequent introduction of a new technology. As the diffusion process proceeds, the rate of change in productivity growth slows down as the new (more productive) technology becomes dominant. This is a selection process under the assumption that the new and more productive technology is represented by the entry of “modern” firms and the demise of old traditional technology is represented by the exit of “craft” firms.  Along the way, Fisher’s equation is rediscovered, quite an accomplishment. As you say, this works fine without a G–P distinction. I agree.

You also implied that the G–P distinction is unnecessary in evolutionary game theory. Strategies are phenotypic traits (Maynard Smith) and the question is what phenotypes will evolve. The analysis is based on a particular (but rather broad) assumption regarding inheritance (or more precisely genetic variation). Note that the variance generated in some forms of replication may undermine the validity of evolutionary game theory models!

You said: “All one needs is a mechanism by which entities with different durable or semi–durable characteristics act in an environment in which those with certain characteristics are more fit, tend to expand more or contract less, than those with other characteristics.” I agree. This kind of model can be viewed as a study of the evolution of phenotypic traits without bothering about the details of replication (they are captured in rates of reproduction or growth rates). According to Maynard Smith, this is useful if we wish to identify the selective forces that are responsible for the evolution of particular traits. For this to be meaningful, however, we need to situate such specific analysis in a general theory of evolution that accounts for the details of replication (or transmission) and selection. It is here that the G–P comes into the picture and even seems to be a logical necessity.

Replication processes introduce new variance and thus, in conjunction with selection, changes the distribution of traits. Different selection processes and different replication processes will give rise to different population distributions. However, even in your AER model and in Fisher’s work, there is an assumption of replication, which is summarized in the concept of fitness, i.e. the rate of reproduction of a trait, or the growth rate of a new technology. In order to justify this assumption, we must have a reasonable theory about the replication process.

I assume we want to understand how replication works. It is very hard to think of an interacting entity literally making a copy of itself. It is much easier to think of an entity containing information that will produce a copy (a bauplan). Even when photocopies are made, information is first extracted and then used to produce the new copy. When we wish to understand replication, something like a G–P distinction seems a logical necessity.

In Sid Winter’s recent work with Gabriel Szulanski on replication as strategy, a similar distinction appears. There are instructions for replication (e.g. of a new McDonald outlet) contained in something called the “Arrow Core” and there is the actual outcome of these instructions (the new replicated McDonald outlet).

I hardly need to point out that something like the G–P distinction is present in Nelson & Winter (1982).

In David’s work on general selection theory, something like the G–P distinction also shows up. For example, in his BBS target article with Langmann and Glenn, selection is defined as “repeated cycles of replication, variation, and environmental interaction so structured that environmental interaction causes replication to be differential.” Here, the G–P distinction shows up in the requirement that interaction (phenotypic level) must cause differential replication (genotypic level). In a commentary, Pepper and I argued that Hull et al.‘s treatment of operant learning as a selection process did not work because it only considered interaction, but disregarded the need to include multiple concurrent replicators in any definition of selection. David agreed. The implication is that something like the G–P distinction is at work.

There are additional reasons to consider something like a G–P distinction. The notion of double–loop learning from information science is consistent with this two–level approach. One loop responds to changes of the environment within limits given by the second loop. The second loop is reset less frequently than the first. In changing environments, entities that are capable of double–loop learning usually beat entities that are only capable of single–loop learning.

We should be careful in using biological language. Rather than speaking of a G–P distinction in the social world, we might speak of an instruction set that contains the information necessary to replicate an entity and its observable features that are essential to environmental interaction. Note here that the distinction is between an instruction set and observable features of an entity, NOT between the entity and its features as Dick seems to imply. To study the ways in which replication can happen and the implications of particular replication processes for the ways in which selection processes work seems a very exciting subject deserving more attention. This requires something like the G–P distinction.

Whether the G–P distinction is present in Darwin’s work is a different issue. Lewontin has a nice entry on the Phenotype–Genotype distinction in a volume by Evelyn Fox Keller and Elisabeth Lloyd (1992). It was introduced in 1911 by the Danish biologist Wilhelm Johannsen. According to Maienschein in the same volume, Johannsen also coined the term gene in 1909, intending it to serve the same purpose as Darwin’s “pangen”. Let me add that it is not out of loyalty to another Dane that I focus on the G–P distinction. According to my own reading of Darwin, in particular the 7th chapter of Origin, the G–P distinction is clearly present in his work and necessary to complete his theory (as Geoff and I said in a footnote). Independent of the outcome of this discussion among specialists, I would say we need the G–P distinction in order to understand replication. 

Should we call a process that involves conscious learning and imitation “Lamarckian”? I accept your hesitation in using biological concepts and take it seriously. This is one of many reasons that I think we should drop the term “Lamarckian” altogether. As Geoff and I argued, it is a biological concept that does not travel well. I also agree with Dick that we need a concept of selection. Understanding replication processes adds the need to specify instruction sets and observable features. We also need to understand how information is preserved in replication processes and how such processes give rise to new variation. Even if these issues involve the core Darwinian principles, it might be wise also to drop the term “Darwinian”! I realize that habits of thought are very persistent, but I find it highly promising that we are both considering the terms and the underlying issues.

I do find this exchange of thoughts very useful!!!

Best wishes

From: Richard R. Nelson
Sent: Monday, May 31, 2004 8:54 AM
To: Thorbjorn Knudsen
Subject: Re: Necessity of G–P distinction?

Dear Thorbjorn,

I think we are coming to see things the same way. I too am finding the exchange helpful both in clearing and testing my thinking on this complex set of issues. Let me lay out some general points on which we now see eye to eye, or I think we do:

The purely logical constraints on a theory of cultural evolution are pretty loose

This means we can choose the details of such a theory so as to square with the empirical facts of the processes involved, as we are coming to understand these.

Much of the current detail in the theory of evolution in biology is there to square with the empirical facts as they have come to be understood, not for reasons of logical necessity

This means we should be wary of mimicking those details, unless they seem to fit the empirical facts of cultural evolution as we understand these.

More specifically, one can develop a logically coherent theory of cultural evolution without treating the characteristics that mark an entity as separate from the entity itself. Of course to understand the evolutionary forces at work, it is important to understand what characteristics advantage or disadvantage an entity, and how, and the mechanisms of expansion or contraction. But there are no logical reasons why characteristics should not be regarded as part of the entities.

There is an empirical reason for not having this tight coupling, however. In biology it is that organisms die, and give birth, and some of their characteristics seem to be “inherited”. (We now know a certain amount about the mechanisms at work). Regarding cultural evolution, individuals and organizations are not always stuck with the cultural elements they hold, and we know there is a considerable amount of teaching, learning, social pressure, and imitation going on. Thus elements of culture are not permanent properties of individuals or organizations, and we need to treat them, therefore, as a separate class of things in our theorizing.

But there is no logical reason why these cultural elements should not be regarded as characteristics of the individuals or organizations that possess them at any time, that directly affect their fitness. These cultural elements could be treated like good and bad lice, that adhere to some individuals, and not to others. That is, a “habit of action” directly influences performance through determining actions, or we could assume this if we chose.

However, there are good reasons for not treating all cultural elements this way. In biology, for the most part genes do not determine the operating characteristics or behavior of an organism directly (thought in some cases they do) but rather are an influence on these. .Thus the phenotype is not completely determined by the genotype. And regarding “habits of thought” at least, one would not want to assume that these map deterministically into patterns of behavior. Rather, they define proclivities.

While this has not been a topic in our conversation, I would propose as well that it is not useful to think of all elements of culture as adhering to particular individuals and organizations, whose fitness or lack of it depends on them. Thus a scientific theory is to a considerable extent a community property. It is tested and found fit or wanting through community action and decision. This is a much more important aspect of it than whether or not a particular scientist believes it at any time.

Where does this lead me to come out on “Darwinism” and “Lamarckianism”? I am reluctant to push analogies with biological evolution, generally. Sid and I said this, in no uncertain terms, in our book. I have been saying this ever since. Certain commonalities are interesting, and it may be useful to give them names. Thus I am comfortable with the proposition that cultural evolution has a number of “Darwinian” characteristics. However, there certainly are important aspects of Darwinaiansm that do not carry over from biology. Thus, the problem with the concept of “inheritance” in cultural evolution hits the Darwinian analogy as much as it hits the Lamarckian. Similarly, there are real problems with the notion of “natural selection” if this is taken to connote that the life and death of individuals is a stake. Sometimes it is but mostly it isn’t. I also am more comfortable than you and Geoff are, or David, saying that it has certain “Lamarckian” properties as well. But the analogizing here is a matter of convenience. And there certainly is a downside to such analogizing. It is important to be clear that there are important differences in the details of the mechanisms involved.

Geoff. Are you and I in tune on all this?

Best regards, Dick

From: Thorbjorn Knudsen
Sent: Monday, May 31, 2004 12:06 PM
Subject: A remnant of disagreement

Dear Dick

I here focus on a remnant of disagreement.

In my previous note, I argued that the G–P distinction seems to be a logical necessity when accounting for the details of replication. We seem to have a different understanding of the G–P distinction. The distinction I had in mind was not between entity and traits. Instead, I had in mind a distinction between traits (possibly represented by the entity itself) and the information necessary to produce a similar copy of an entity with these traits. When producing a new entity, the translation of this information is subject to information loss and environmental influences. As you say, the phenotype is not completely determined by the genotype. And regarding “habits of thought” at least, one would not want to assume that these map deterministically into patterns of behavior. Here we completely agree.

A possible disagreement is that I see the G–P as a logical necessity in accounting for replication. Accounting for replication requires that we can explain how the information required to produce a new entity is (reliably) transferred. As you say, we do not need to treat the characteristics that mark an entity as separate from the entity itself. I agree, but this ignores my argument that we need to treat the information that is required to produce the entity (Genotype) as something different from the entity itself (Phenotype). If you agree to this point, something like a G–P distinction follows as a logical necessity.

Your proposal that it is not useful to think of all elements of culture as adhering to particular individuals and organizations is indeed pressing the issue of information storage and transfer. To understand the problem, I think we will find it necessary to treat the information that is required to produce the entity in question as something different from the entity itself.

My argument was further that we can ignore the G–P distinction for specific purposes, for example, if we wish to identify the selective forces that are responsible for the evolution of particular traits. This requires that we are confident that the effects generated in replication processes do not violate the assumptions of our formal models. This is not always the case.

I added that we need to situate such specific analysis in a general theory of evolution. There are some constraints. I see variation, selection and inheritance (information transfer) as abstract and general principles providing minimal constraints. We might benefit from further refining these concepts and, in particular, the information transfer condition necessary to understand replication.

I see a possible reason for the difference in our views regarding the use of the terms Lamarckism and Darwinism. Geoff and I, among others, have argued that the principles of variation, selection and inheritance can be viewed as general and abstract. Abstraction is different from analogy. Abstraction removes the particular content from the specific domains in which a principle applies. The abstract principle is therefore sufficiently general to be applied in different domains. For example, entropy is an abstract principle that applies to the domains of earth, wind, fire and water as well as to the biological domain. Possibly, variation, selection and inheritance are abstract principles that apply in a similar manner to the natural and the social domain. I cannot see a similar justification for the use of Lamarckism. Of course, social scientists might develop their own definition of Lamarckism, but I do not see the reason for it.

Apart from this remnant of disagreement, I think we see things much the same way.

Best regards, Thorbjorn

From: Geoffrey Hodgson
Sent: Monday, May 31, 2004 2:25 PM
Subject: Re: Necessity of G–P distinction?

Dear Dick

I am finding the current series of exchanges between the several of us very useful. What follows is a very quick telegraphic and incomplete response to what you say below, Dick, as I am busy getting ready to go abroad early tomorrow (I return on 20 June).

I am mostly in strong accord with what you say below. For instance, I strongly agree with what you say about the importance of empirics. I strongly agree that we should be highly sensitive to the specific details of cultural evolution. And of course the phenotype is never entirely determined by the phenotype alone. There are many other points with which I agree.

I agree that “there are no logical reasons why characteristics should not be regarded as part of the entities.” But I would point out that the characteristic/entity couple is not the same thing as phenotype/genotype distinction. Actual entities and characteristics are both phenotypes.
My inclination is to suggest that some kind of replication or inheritance can be found in both biological and socio–economic systems. But the mechanisms involved are very, very different. Furthermore, the conceptual work involved in establishing such communality is as yet incomplete, and not consensus exists on what the social replicators are. Accordingly, inclinations aside, I believe that we have to be very careful, and not claim too much.

I’m sorry that this is very truncated. But I must get back to my packing.

Best wishes, Geoff

From: Richard R. Nelson
Sent: Monday, May 31, 2004 3:02 PM
Subject: Re: A remnant of disagreement

Dear Thorbjorn,

My argument with your argument here is that the case for having something like a genotype–phenotype distinction in analysis of cultural evolution is not logical but empirical. Such a distinction is not logically necessary in order to explain “replication” The entity can simply split in two.

The argument that culture influences, but does not determine, human action is an empirical one. Similarly, the argument that what is learned by one actor is, in some sense, a conditional program for affecting something, rather the thing itself, is an empirical proposition. Thus routines, habits, are gene–like in some ways. I have no trouble with that proposition. What I do not believe is:

First, that the argument is of logical necessity, rather than motivated by our attempts to develop a theory of what really is going on, as we understand that empirically.

Second, that this argument calls for picking up other parts of the baggage of biological gene theory.


From: Thorbjorn Knudsen
Sent: Monday, May 31, 2004 6:42 PM
Subject: Information transfer

Dear Dick

An entity simply splitting in two?  I find this example empirically implausible. Are there any empirical examples of entities simply splitting in two in cultural evolution or other forms of evolution? More importantly, I also find the example unconvincing as an explanation. What explains the splitting? What explains that the entities in question do not vanish if we keep on splitting? What explains any similarity after the splitting?

A straightforward explanation is that the entities can be similar if the first entity carries the information necessary to produce a similar copy. If this information is also transferred to the second in the copying process, the process can continue forever. This is a very general account of a replication process. By logical necessity, it seems that the information that is needed to produce the entity is something different from the entity that is being produced.

You are right that my argument was limited to replication processes. I agree that similar copies of an entity can be produced in other ways, at least in principle. I am not sure, but it seems to me that in order to explain the production of similar copies of an entity, we need some kind of information condition. It is not easy to think of a counter–example, but perhaps there is one.

The other way round, if information is not needed to produce an entity, there is no logical necessity. I agree. But in this case, how do we explain how similar copies of an entity are produced? The logical necessity seems to arise when we want to explain, rather than assume, the production of similar copies of an entity.

I very much agree that something like the G–P distinction is also motivated by our attempts to develop a theory of what really is going on, as we understand that empirically. For now, I remain unconvinced that it is not also a logical necessity if we want to explain replication, and perhaps more generally.

I also agree that this argument does not call for picking up other parts of the baggage of biological gene theory. As I have argued all along, the requirement, at least if we wish to explain replication, seems to be a distinction between the information that is needed to produce the entity and the entity itself. I see an information condition, but I do not see any biology in this distinction. In the “The Limits of Lamarckism” paper, Geoff and I were also very clear about a social G–P distinction having no biological content whatsoever.

A very useful outcome of our discussion is to instill a keen sense of alertness towards the transfer of strictly unnecessary components of biological thinking and concepts. I assume that we agree that we need a concept of selection and a concept of variation. Our possible remaining disagreement concerns whether we need a concept of information transfer in the explanation of replication processes. I think we do.


From: Richard R. Nelson
Sent: Tuesday, June 01, 2004 7:50 AM
Subject: Re: Information transfer


As I was reflecting on how to respond to your last note, I was looking out the window, following some raindrops running down the glass pane. As I was watching several of them split into two, each of which after a pause continued running down the pane. There is a good understanding in physics as to why a raindrop running down a pane might split (there are a number of possible causes). And there are good reasons why each of the now two drops should form a drop in most ways like its ancestor.

Are you sure you want to characterize those reasons in terms of “information” encoded in each drop? A bauplan is somewhat less awkward, but still pretty awkward.

You may say that cultural transmission, in the sense of “copying”, is not like this, and I would agree. I would agree that in cultural copying it makes sense to propose that something like information transmitted. But that is part of a theory of cultural replication motivated by what we know about it empirically. It is not a logical requirement of any process of copying.


From: Richard R. Nelson
Sent: Tuesday, June 01, 2004 7:59 AM
Subject: Re: Necessity of G–P distinction?


One wonderful example is Mandeville’s discussion of the “modern man of war” contained in his Fable (around 1720 I believe). He doesn’t present a variation–selection model there explicitly, but like Darwin long after him he was concerned with arguing that very sophisticated structures can come to exist without being the result of a plan, by man or God, but rather as the result of an accumulation of small advances. Adam Smith’s account of changes over time in pin making technology and organization of production is similarly “evolutionary” in spirit. There are other examples.

I have used the term “evolutionary in spirit” to describe these pre Darwinian theories of cultural change. The sharpening up of specification of the process to involve variation and selection explicitly comes with Darwin and Wallace.

From: John van Wyhe
Subject: Re: Necessity of G–P distinction?

Dear Dick,

This bears on my research on the history of cultural evolutionism:
“There is an extensive body of writing on cultural evolution, that
goes back well before Darwin, on aspects of technological and social
evolution, where it is clear physical survival is not at stake.”

Can you name some of the sources you have in mind?

Thank you.

Regards, John

From: Thorbjorn Knudsen
Sent: Tuesday, June 01, 2004 9:12 AM
Subject:  Re: Information transfer


This is getting interesting! The raindrop splitting is a beautiful example!!!

My friend, who is a physicist, and co–author for another line of my work, tells me the following. First, if you keep splitting, the drops will vanish. In order to avoid that the drops vanish you need what the physicists call translation invariance. Translation invariance can be viewed as a minimal information requirement, correcting for diminishing size. It is useful to think of translation invariance as something different from the entities being produced.

Second, as the drops are splitting there is a change in entropy. In turn, entropy can be viewed as a minimal measure of information. Thus, some kind of information condition is involved. According to my physicist friend, it is useful to think of the entropy (information) as something different from the entities that are being produced.

Third, unless the splitting happens in weightless space, the two new drops will not be identical after the split. There will be some “error” in the splitting process. More generally, as the information content of an entity increases, errors will cumulate and the copies will be less similar. Therefore, as the information content of the entities we are splitting is increasing, the presence of an information transfer condition becomes critical at some point in order to preserve a minimal level of similarity. This is a logical necessity for any kind of splitting process!

I am very pleased that you agree that in cultural copying it makes sense to propose that something like information transmitted. Great!

I am aware of your argument that this is by empirical requirement and not logical necessity. However, according to my argument here, this is by logical necessity if we are copying entities with a high information content, which is exactly what we do in cultural (and biological) copying!

I am very pleased that we are beginning to converge in viewpoints, also on this remaining issue.


PS: I am told that an example from physics that does not require translation invariance (in contrast to the raindrop example) would be the stretching of cosmic strings and a subsequent splitting (after some rotation). Thus, in spin–models we have an example of an entity where we can keep splitting without the entities vanishing.

From: Marion Blute
Sent: Tuesday, June 01, 2004 10:08 PM
Subject: Re: The Limits of Lamarckism

As I understand the paper, you argue that Lamarckian–Darwinian distinction is incoherent in the absence of a genotype–phenotype distinction or its equivalent, that with such a distinction included a Lamarckian theory in any realm is fraught with difficulties (particularly reverse translation), and hence evolutionary social science should be Darwinian in its theoretical orientation.
I have traditionally maintained, on both logical and empirical grounds, that all selection processes are Darwinian rather than Lamarckian.  I argued that briefly about cultural evolution in my 1979 Beh. Sc. paper and at greater length about individual learning in the long paper posted on CogPrints.  In that sense I am in sympathy with your general view but still have a lot of reservations about the paper.
One of the problems with this issue as with many others is that working within the variation, replication and selection or similar formula is becoming increasingly untenable. However, we won’t have anything better until we have a general theory of development, and one unified with the existing unification of evolution and heredity.  For my take on a small slice of this see the attached (If the genome isn’t a God–like ghost in the machine, then what is it? answer the brain of simple cells and the government of compound ones) forthcoming in Biology and Philosophy.
If Lamarckianism is defined as the inheritance of acquired characters as per your introduction then biological evolution is MASSIVELY Lamarckian (Jablonka & Lamb’s book 95) and this is no longer a minority view. They were however widely criticized for suggesting, somewhat disingenuously, that this introduces a Lamarckian element in evolution because the latter is most commonly defined as the inheritance of acquired adaptations.
Re unpacking “replication”. Your unpacking is causation, similarity and information transfer.  One I like better is causation; similarity and material overlap (latter from Griesmier).  Even there are problems however. In multigenerational life cycles no single event at least involves similarity although they do causation and material overlap.  On the other hand, when a prion imposes its form on another molecule there is causation and similarity but no material overlap yet prions can evolve by natural selection.  Some multicellular invertebrates reproduce with multi–celled propagules. Such having been said, it is also the case that ultimately unlimited (many possible states – MS & Sz) inheritance systems capable of sustaining cumulative evolution utilize digitally encoded information (Dawkins River out of Eden, Blute BBS 2001 24:2, p. 327) or more accurately knowledge i.e. it is the digital encoding that makes possible fidelity.
Hence rather than viewing social scientific Lamarckians as incoherent based on an insistence that information transfer is a necessary feature of replication and so too then is reverse translation, I would be more inclined to try and clarify their intuition.  They seem to think, reasonably on the surface, that what individuals learn (e.g. by trial and error) may then be transmitted culturally by social learning, that routines developed in one organization can spread to others becoming an industry standard of best practice, that attitudes characteristic of one institution (e.g. calculating attitude prevalent in capitalist economies) can spread to other institutions etc.  And they are right up to a point in the sense that there can be Lamarckian inheritance of acquired characteristics. You make the point that acquired characteristics are more commonly mutilations than adaptations.  But in multi–level selection processes the point may be more subtle. They ARE adaptations, but on a less inclusive level. What is favoured on the less inclusive is not necessarily the more inclusive.  Hence a genetic sequence able to engage in multiplicative transposition does not necessarily increase the fitness of the cell of which it is a part.  As a scientist I may learn something in my own research, but still have difficulties persuading my colleagues.  Delius I think it was once pointed out that masturbation and nose picking may be rewarding but they are not cultural norms.  If one firm is successful with a particular routine, it does not mean that others will be too – it is often the case that others should do something different to avoid competition.
M.S. & Sz. claim that the only inheritance systems, beyond those based on nucleic acid molecules, capable of expressing an unlimited number of states and hence supporting cumulative adaptive evolution, are human language and music.  I discovered one other – some traditional Indian dances are digital with discrete moves which can appear in different orders in different dances etc. Children chant words or phrases with each move as they learn them as a memory device.  Assuming that MS & Sz are basically right and I think they are, then it would appear that social learning by observation involves a primitive form of evolution without encoded information (although sometimes there is some reverse translation i.e. the observer describes the actions they are observing in words in their heads or even out loud and then uses this reversely translated information to guide their actions as they perform them). What then is the genome of an organization in particular?  Keeping in mind it must encode the programme of the entire organization not simply of individual practices or roles, the one thing that comes to mind is the written business plan. I have noticed that “business plan” these days is used not only for private, for profit organizations but also for public and non–profit ones. Of course there is a lot of loose horizontal transmission in cultural evolution i.e. elements of a plan spreading independently of the rest, but we know there is a lot more of that in biological evolution too than we used to think, particularly but not exclusively early in the history of life.
Re habits of action and thought, if you are drawing the distinction between social learning by observation and instruction which at times you seem to be, it could be done more simply.  If not, I don’t get it starting right with the notion that DNA molecules replicate directly (rather than are replicated by enzymes because of the complex indirect route of having a positive effect on a phenotype).
  Some of the social science could be sharpened up.  For example, there is some looseness re what kinds of entities you are dealing with – in the abstract “social organizations” but elsewhere “socioeconomic evolution”, “socio–economic and cultural evolution in human society”, “economies and societies” etc.  If you want to be inclusive perhaps evolution in the social sciences might do.  If you want to be narrow, you might want to stick with organizations.
I remain disappointed with the little use you make of psychology and social psychology beyond late 19th and early 20th centuries.  You ignore not just behaviourism as you state but most 20th century cognitive and social psychology.  I agree that a cognitive language to convey gene–like concepts and not just behavioural (phenotype–like) concepts are important.  But the list available is almost endless (e.g.representations, plans, algorithms, programmes, rules, norms, values, cultures) almost all of which convey a sense of digitally, symbolically encoded knowledge much more so than do propensities, habits and perhaps (although I am not sure about
this) even routines.  The statement on p. 10 about much loose discussion of transmission etc. but little dissection of the precise processes is just plain wrong.  There is a huge psychological and social psychological literature, both theoretical and empirical, on social learning.  It has been of great interest in recent years, precisely because of its potential role as a second inheritance system making socio–cultural evolution possible.


From: Thorbjorn Knudsen []
Sent: Wednesday, June 02, 2004 7:40 AM
Subject: Information and overlap

As Dick’s raindrops have showed, there is not something like “simply” splitting an entity in two. A lot of constraints must be in place and additional requirements show up. For example, something must be added along the way, if the raindrops are not going to vanish as they keep splitting.

According to my argument, as the information content of the entities we are splitting is increasing, the presence of an information transfer condition becomes critical at some point in order to preserve a minimal level of similarity. I have not been persuaded otherwise. What if the raindrops were complex organizations?

Even so, an information transfer condition could merely be an empirical requirement as Dick insists. An empirical requirement according to what criteria? Merely empirical criteria? According to David, there are not any general rules that will help decide what are the strictly necessary components and which are “merely” empirical.

I am very pleased that we agree, even for different reasons, that information transfer is a component of cultural evolution. Even if it were not, in the case of replication processes, we would need some mechanism that produces a degree of similarity in copying (as in the raindrop example). I see this as an abstract principle of “inheritance”, and I think some such abstract principle is necessary to complete any evolutionary theory (involving replication) in addition to variation and selection. Our discussion has not persuaded my otherwise. A lot of examples have been raised, but all of the examples of replication processes involve a degree of similarity in copying.

Marion likes material overlap better than information transfer. It partly depends on how we define information and matter. It also depends on what causal role the overlapping matter has. Marion adds that ultimately, unlimited inheritance systems capable of sustaining cumulative evolution utilize digitally encoded information. Fine. Perhaps an information transfer condition becomes critical in systems of potentially unlimited evolutionary potential as implied by Maynard Smith and Szathmary.

Good to hear from you, Marion! I am delighted by your challenging, very detailed and useful comments. I have much more to say later in response.

It has occurred to me that we might all benefit from a meeting to discuss face–to–face. Hope this could happen.


From: John Gowdy
Sent: Wednesday, June 02, 2004 10:37 AM
Subject: Re: A remnant of disagreement

Hello everyone: I’ve been following this discussion with great interest and I thought I would jump in with a few comments. First of all I agree that biological analogies can be pushed too far. But there is a great opportunity for economists to gain insights from a field with a similar subject of study (living, evolving, complex systems) that has made greater theoretical strides than economics. Recent evolutionary concepts like punctuated equilibrium, group selection, and altruism are beginning to be taken seriously by economists.

On the phenotype–genotype distinction, the following points seem relevant.

1.  The distinction is not as clear in biology as it once was. Phenotypical changes are not exclusively triggered by changes in genes, but can also result from changes in the environment. Not to be Lamarckian, but phenotypical changes can apparently be passed on to subsequent generations. For example, when Paramecium cilia are surgically removed and placed back in reversed order this reversal will be inherited by subsequent generations. According to E.O. Wilson the accepted explanation of causality from genes to culture, as from genes to any other product of life, is not heredity alone. It is not environment alone. It is the interaction between the two. (Consilience 1998, 137)
2. This genotype–phenotype point is relevant to the parallel discussion in economics on inventions versus innovations. Both the generation of inventions and the adoption of them through innovation depend on historical contingency and the cultural environment. Dick Nelson has written about this I think in his EJ paper on the Korean automobile industry. Joseph Henrich has incorporated some of these ideas in his formal models of “biased cultural transmission”.
3. Can homeobox inventions be identified before they are innovated? For example, could the economic, environmental, social impacts of the automobile or the PC have been identified ahead of time? 
4. Getting a handle on the concept of “hierarchies of selection” could reconcile (or at least organize) the discussion on the scale issue raised by Peter Corning and commented on by Dick Nelson. Recognizing the implication of group selection is critical here.
5. One last polemical comment. Evolutionary psychologically (and neurobiology) is having a tremendous impact not only on economists understanding of consumer behavior but also firm behavior. The decisions of firms (including whether or not to innovate) are made by living, breathing humans and recognizing this will change our understanding of the innovation and diffusion process. We have to throw out our basic model of the economic system that makes every living person “irrational” and every existing market and firm “imperfect.”.

John Gowdy

From: David L. Hull
Dear Folks,

More on raindrops.  Each time a raindrop splits in two, it loses part of its matter.  After enough splits, there is not enough matter left to split anymore.  But in mitosis (sorry for using the old term), each split is accompanied by the addition of matter.  These are empirical distinctions but sufficient to explain why the splitting of raindrops does not count as replication, while the splitting of molecules of DNA does.

A related issue.  Biologists are working on the mechanisms that bring about splitting, whether at the molecular, cellular or species levels.  That they are studying splitting is conceptual; what these mechanisms happen to be is empirical.  Bur how about merger?  Those pesky raindrops split, but they also merge on occasion.  Why have biologists, especially systematists, been so slow to accepting the prevalence of merger in biological evolution?  Because it poses serious problems for the comparative method.  But still, merger does take place at the molecular, cellular and even species level.  The usual examples are horizontal transmission of viruses and bacteria, but there are lots more.  For example, every time fertilization occurs, a sperm merges with an ovum to produce a zygote.

Does the contrast between splitting and merger play any role in other areas, or is it just a peculiarity at the biological level?

Gowdy’s email made lots of sense to me.  He is absolutely right that the distinction between genotype and phenotype in biology is not as clear as it once was.  This is the fate of distinctions in biology! But we have to learn to survive fuzzy distinctions.

I was taken aback by the cilia example.  I invented it –– at least I thought that I invented it –– way back in 1980 in my “Individuality and Selection” paper.  Is this a conceptual homology or a conceptual homoplasy (analogy in old speak)?  I would be pleased if the former.

Our opponents claim that our “analogies” have yet proved to be of any use, but they have for me.  Dick Nelson has written on the role of contingency in cultural evolution.  Contingency also plays a big role in biological evolution.  This similarity makes me think that contingency plays a larger role in science in general than lots of people are willing to admit.

One final point.  In gene–based selection in biological evolution, genes, chromosomes and even entire genomes can function as replicators –– rarely higher.  That’s why I term it “gene–based.” Interactors exist at a wider range from genes through organisms and possibly up to “groups.”  One and the same entity can function both in replication and in environmental interaction, but in general these processes are distinct.  Organisms are the paradigm interactors. Their traits are what count in environmental interaction.  They are the phenotypes.  The following distinctions need to be made with respect to phenotypic characteristics in selection:

(a) The phenotypic traits are unproblematically biological, e.g., having four split telsa. This is “selection” in the most literal sense.

(b) But many organisms also exhibit behavioral traits, e.g., chicks screeching when parents approach the nest with food.  Such behavior can be construed as part of an organism’s phenotype, but this sort of selection is not all that “literal.”  However, this is how most sociobiologists treat selection.

(c) Now the big move comes.  The genetic material no longer functions as replicators.  Memes take over this role.  Most disturbingly, entities that used to be treated as replicators now become interactors –– part of the phenotype!  I suspect that this boggles most people’s minds.


From: Thorbjorn Knudsen
Sent: Wednesday, June 02, 2004 5:46 PM
Dear David

Even if matter was added as a raindrop was split, the new matter must not be added in a random way. If it were, with each split, additional random features would be added. Quickly, any similarity would go away. Adding matter after splitting an entity in a way that maintains similarity apparently requires some kind of ordering principle.

A paradigm case of split and merger comes from the study of diffusion in physical chemistry. Energy is added to a reactant, which splits in two. There is some likelihood that the two components will merge again, or merge with other reactants that are split in a similar manner. The merger only happens if a minimal ordering principle is fulfilled, i.e., the spin of the components must be aligned in a particular manner. If we think of merging entities that are more complex, one would expect that the ordering principle included more features (information). An example would be the case of a merger of business organisations. 

Entities that used to be treated as replicators can become interactors. In yet another article co–authored with Geoff, “The Firm as an Interactor: Firms as Vehicles for Habits and Routines” (forthcoming in JEE), a similar point is argued in the case of business organisations:

“There is a hierarchy of interactors, including firms at one level and individuals at another. There is also a hierarchy of replicators, namely routines, habits and genes. How do these two hierarchies relate? The distinction between ‘selection of’ and ‘selection for’ is relevant here (Sober, 1984). Just as the selection of individual organisms in genetic evolution results in selection for the corresponding genes, selection of firms in a competitive environment results in the selection for some of the replicators associated with the firms, such as their constituent routines. That is, the current properties of the firm determine whether its routines, and possibly the habits of its individual members, will be more common or more rare in the next time period.
Further descending the hierarchy, the selection of firms can also have a slight effect in the selection for human genes, given that employment opportunities in the firm can have an effect of the survival opportunities for human individuals. The selection of firms has effects that cascade down to the selection of individuals, and in turn to selection for genes. But selection for these lower–level, biological replicators can be ignored for purposes of analysing economic evolution. It is too slight to be of significance, given the much slower evolutionary processes involved.”


back to top

Permanent URL for referencing this page:

Existing members, please login to post comments.

Become a member to post comments.