Biology & Philosophy

, Volume 23, Issue 2, pp 229–242

Sophisticated selectionism as a general theory of knowledge


    • Complex Systems Group, Division for Physical Resource Theory

DOI: 10.1007/s10539-007-9085-7

Cite this article as:
Andersson, C. Biol Philos (2008) 23: 229. doi:10.1007/s10539-007-9085-7


Human knowledge is a phenomenon whose roots extend from the cultural, through the neural and the biological and finally all the way down into the Precambrian “primordial soup.” The present paper reports an attempt at understanding this Greater System of Knowledge (GSK) as a hierarchical nested set of selection processes acting concurrently on several different scales of time and space. To this end, a general selection theory extending mainly from the work of Hull and Campbell is introduced. The perhaps most drastic change from previous similar theories is that replication is revealed as a composite function consisting of what is referred to as memory and synthesis. This move is argued to drastically improve the fit between theory and human-related knowledge systems. The introduced theory is then used to interpret the subsystems of the GSK and their interrelations. This is done to the end of demonstrating some of the new perspectives offered by this view.


SelectionUniversal selectionEvolutionary epistemologyKnowledgeHierarchiesInnovation


Selection theories have cropped up in several fields of study outside of evolutionary biology. However, in only one case, that of clonal selection (Jerne 1955; Burnet 1957; Edelman 2004b), have unequivocal breakthroughs in understanding been achieved thereby. The other such field that seems closest to achieve breakthroughs through the application of selection theory is neurophysiology, e.g. (Edelman 1987, 1993, 2004a; Changeux 1985; Changeux and Connes 1989; Changeux and Ricoeur 2000). Application to social dynamics has also been attempted, e.g. to culture (Boyd and Richerson 1985, 2005), and to economics, e.g. (Nelson and Winter 1982; Boulding 1981; McKelvey 1982; Hodgson 2002; Saviotti 1996; Nelson 2007), and in the “memetic effort” beginning with Dawkins, e.g. (Dawkins 1976; Aunger 2002; Blackmore 1999; Dennett 1995; Edmonds 2005). Furthermore, the area of evolutionary epistemology applies selection to the problem of human knowledge (as well as in a general sense), e.g. (Campbell 1960, 1974; Popper 1979; Bradie 1986; Heyes and Hull 2001; Lorentz 1977). It seems clear that the degree to which such efforts have been progressive is correlated with the degree to which a clear empirical connection has been made—this is what sets the former two efforts most clearly apart from the latter ones. The latter are all concerned with cultural1 knowledge in one form or another, and here no consensus has emerged as to what it is that gets selected, what gets replicated, what fitness is and so on; which of course is not to deny that more or less fruitful candidates have been proposed. Selection theories of culture have thereby remained outside of the mainstream and far from everybody is convinced that selection has an important part to play on the social scientific stage. What seems to be needed in order for theory to reach a progressive phase is a better fit between theoretical concepts and the structures and processes constituting the systems that we seek to understand. General selection theorizing addresses this question by attempting to expose what sorts of abstract circumstances that cause selection in the first place; it is in this tradition that the present work continues.

A common discourse

The gains that stand to be made from generalizing selection lie to a great extent in the prospect of being able to put different selection systems into the perspectives of one another. With the discovery of more and more interrelated systems where selection seems to be the governing principle (Jerne 1955; Burnet 1957; Edelman 1987; Nelson and Winter 1982; Cziko 1995; Boyd and Richerson 1985; Campbell 1974), the development of this branch of selection theory has become ever more relevant. For one thing, general selection theories may serve to make it easier to use insights made in one context also in other contexts while avoiding to some extent the risks entailed in the use of analogies. That is, instead of saying that human knowledge “works something like genes” we try instead to speak of both as specific incarnations of some more abstract type. Furthermore, by providing a common discourse, also deeper questions about origins and the relation between hierarchical levels of organization (e.g. cultural and biological) can be better addressed.

Hull’s theory of selection in terms replicators, interactors and lineages is situated on a branch of the Darwinian tree that has grown in the direction of abstractness and generality (Hull 1980, 2001b; Sober and Wilson 1994; Brandon 1988). Within evolutionary biology, the framework of Hull and the related contributions of many others (Dawkins 1976, 1983; Price 1995; Brandon 1988; Goodfrey-Smith 2000; Griffiths and Gray 1994; Sterelny et al. 1996), have provided a language that allows precisely this: discussion without the need to use misleading idiosyncratic analogy concepts (e.g. genes, organisms etc.) that properly belong elsewhere or on a less general level of discourse. For broader epistemological issues—ones that span not only biology but also across the boundaries of academic fields into immunology, neuroscience, psychology, sociology, technology and epistemology—the need to make recourse to field specific concepts threatens not only to mislead but to make discourse altogether impossible. However, for framing systems as different as these in selectionist terms Hull’s framework has not had a comparable degree of success. Furthermore, the mentioned framework of Hull represents of a class of frameworks that have failed here: selection theories where replication serves a fundamental role; which is to say nearly all recent non-phylogenetic selectionist theorizing, excepting that of Edelman and Changeux, e.g. (Edelman 1987, 2004b; Changeux 1985; Changeux and Ricoeur 2000). While the nature of replication is frequently debated, the central role of replication itself seems to be considered common sense. This can perhaps be defended to some extent within biology, but even here the replicator concept becomes cumbersome as soon as non-standard models are tried out, e.g. (Griffiths and Gray 1994; Goodfrey-Smith 2000; Sterelny et al. 1996).

Why insist on selection?

Prescient2 variation is glaringly present in any process of human knowledge growth and this would seem to substantially weaken the prospects for selection to become a key concept of general epistemology. The reason to nevertheless insist on selection as the unifying causal mechanism behind knowledge3 is that if Campbell is right, and all novelty must on an ultimate level be blind (i.e. non-prescient), then blind-variation-selective-retention (BVSR) must have something to do with knowledge on any and all levels of organization (Campbell 1960, 1974; Cziko 1995).

Campbell’s argument for the necessity of non-prescience in novelty is too strong and is situated on a too low level to be refutable solely by pointing to high-level cases of proximate prescience. In fact, the relation between Campbell’s claim and high-level knowledge systems is perhaps best described as a constraint on what can and cannot be said about knowledge. The question is how this constraint can be exploited theoretically to explain (rather than explain away) the role of proximate prescience in knowledge accumulation.4 To do this, cultural knowledge must be put into theoretical contact with interrelated knowledge systems; most importantly with neural and genetic knowledge. The present work attempts to improve the meta-theoretical conditions for accomplishing such connections, but let us first dissect Campbell’s claim and the naturalistic concept of knowledge a little bit to better see what they mean.

The naturalistic conception of knowledge adopted here boldly (and not uncontroversially) states that all knowledge is ultimately explainable in terms of non-prescient variation and a posteriori selective retention (Campbell 1960, 1974; Lorentz 1977). Basically, however, this means nothing more controversial than that novelty must ultimately be accounted for without the aid of miracles (Cziko 1995, 2001): intrinsic prescience is not reconcilable with a scientific world view.

However, regardless of whether prescience is intrinsic or just hidden within a black box it still causes theoretical problems. Theories that aim to explain the emergence of instances of fit, but have to assume the actions of black boxes from which instances of fit emanate, will for obvious reasons be less powerful than theories that make do without such assumptions. Both these problems were realized clearly by Darwin: the extraordinary explanatory force of Darwinism is due to its ability to explain purpose without assuming purpose a priori.

Selection theories of cultural evolution will always be declawed in this sense because they must be formulated on a level of organization where human prescience can only figure as a black box: they must assume many aspects of the instances of fit they explain. That is, the naturalistic explanation of cultural instances of fit exist only partly on their own level of organization. It is here important to stress that such theories are still indispensable because it is just as true that a theory of brain function could itself never go the whole mile in explaining culture. Science should probably just get used to what seems to be a fact of life: all explanations do not exist in the narrow focus afforded by single theories.

So, Campbell’s claim is really no more controversial than the naturalistic world view that it is an expression of. In forbidding intrinsic prescience it however neither forbids nor removes the theoretical problems caused by instances of proximate prescience. What it does, however, is to introduce a powerful constraint on proximate theories not wholly unlike a physical conservation law. For instance, the law of conservation of energy allows us to eliminate any proposed model that violates it from further consideration, but this by no means implies that it thereby analytically provides us with all theory that we need. Not violating this law is a necessary but not sufficient criterium in determining whether a theoretical model is valuable or not. This realization is expressed also by Hull et al. in stating that “the efficient production of novelty and order may not sound like an oxymoron, but we suspect that it is” (Hull et al. 2001).

Why the limited success of high-level selection theory?

Human knowledge systems are particularly important in relation to general selection theories because they have proven to be highly problematic to apply selection to. Whether the limited success of high-level selection theories stems from a lack of fit of the selectionist approach itself or from shortcomings in formulations remains to be decided. In any case, that there is a lack of fit is clear. On the one hand we have a lack of clear matches between central theoretical concepts and empirical entities and processes. The highly elusive cultural replicators would be the prime (but by no means only) example of this.

On the other hand, as mentioned in the previous section, we have the issue of the explanatory value of selection theories in various contexts. Opponents of cultural selection theories (and often thereby also a general role of selection) base much of their complaints on the evident efficacy of human intentionality which, being prescient variation, robs selection of much of the natural scientific explanatory value that it wields in biology. Selection is thereby relegated to a secondary role not wholly unlike that which was widely attributed to it also in biology for a long time, see e.g. (Gould 2002). Proponents of selection theories, unfortunately, respond to this criticism not so much by attempting to deal with intentionality theoretically, but more by playing down the importance of intentions and foresight (Hull 2001a; Hodgson and Knudsen 2004). This must, of course, and in particular in the context of economics, be seen against the backdrop of the extreme rationalist position taken by some philosophical and social scientific schools of thought (e.g. neoclassical economics) and as a reaction against the lack of realism in such positions. The strategy of countering denials of the importance of non-prescience with denials of the importance of prescience, however, seems to have done little to convince critics and it has done even less towards progressive theoretical development. Intentionality is there for all to see—we have instances of fit between present structure and future system5—the question is whether this anomaly can be turned from a seeming falsification into a “resounding victory” for a general selectionist research programme (Lakatos 1978).

So the problem with high-level selection theories can perhaps be summarized as the result of: (i) a lack of correspondence between theoretical concepts and empirical reality, and (ii) a widespread suspicion that such theories may have limited explanatory value in any case. What is attempted here is to improve the correspondence and to better specify and delimit the distribution of explanatory responsibility within hierarchies of proximate selection theories. As discussed in the previous section, high-level selection theories must often be limited in their explanatory role, but they are nevertheless indispensable because they add parts to the greater explanatory picture that cannot otherwise be achieved.

The approach

From the preceding discussion it is clear that one thing that is needed is a way to deal explicitly with proximate prescience: it is only by doing this that a unified account of knowledge can be achieved or be revealed as not worthwhile pursuing. To do this, a general theory (or meta theory) is needed that can integrate nested hierarchies of knowledge systems; systems whose dynamics are causally interlocked but that operate on different scales of time and space: phylogeny, thinking, culture and so on.

But the ambition here is not only just to argue for the fit between proximate theories (derived from the presented general theory) and systems, but also to address relations that exist between specific systems. Since this important type of questions cannot easily be addressed from within proximate theories, it is perhaps here that the greatest benefits from a general theory stand to be reaped. Problems of importance that a theory thus relating different knowledge system could address include: the relations between living and non-living, between nature and nurture, between mind and matter, between designed and evolved, whether and in which ways human intentions can be properly viewed as prescient or even a priori (in the sense of Lorentz (1982)) and in which ways social systems can and cannot be understood as being ‘Darwinian’. Some of these questions will be touched upon in this paper but space considerations preclude any effort approaching the close attention that they ultimately demand.

The presented theory can to some extent be seen as the result of a reapplication of some of Campbell’s early insights to the functional theory of Hull (1980, 1988, 2001b) (Campbell 1960, 1974), and as a re-evaluation of selection where biological selection is seen as primitive; primitive as opposed to fundamental (in the sense of being closer to an abstract selection theory). The approach taken here is not to look to the primitive, bare-bone and early for clues but to instead look more to the refined, specialized and late. This angle of attack is motivated by arguing that primitive forms (e.g. primordial replicator-interactors) are general structures, not general functions; they are general in the sense that they pack many functions into the same structure but this is not the sort of generality that we seek. Looking instead at adaptively refined and specialized systems (in particular the human cognitive apparatus and culture-level epistemic processes) general functions may be suspected to be served by dedicated structures that are better compartmentalized along functional lines: structures where the match between function and structure/process will be more clear and one-to-one. In other words: where observation of structure is more likely to reflect something clearly about general function.

A selectionist general theory of knowledge

The first central component of the work presented here is a proposed shift from a replication function to two functions: memory and synthesis (synthetic propositions; i.e. novelty). It is easier to speak about replication as a function that combines the functions of memory and synthesis than it is to speak about e.g. neural memory, books, artifacts and neural novelty in terms of replication. In addition to memory and synthesis, the third and final function is interaction, which remains very close to Hull’s usage.

The second central component is the framing of the problem in a hierarchical structure, where hierarchy is here used in a wide sense that includes also heterarchies and other related concepts; the important feature is the existence of interlocked dynamics on different scales. The GSK is thus viewed as a nested hierarchical set of selection dynamics acting on different scales of time and space. By doing this, some observations that seem problematic, or even like falsifications, from the perspectives of more dogmatic (in the sense of insisting on a strong proximate role) forms of selectionism, are internalized and interpreted within a selectionist framework.

The proposed theory employs the following three fundamental functions:
  1. 1.


  2. 2.


  3. 3.



Interaction concerns the “testing” of existing structures that have the potential to be remembered (or forgotten) by the system. That is, for knowledge it is interaction that holds the key of access to memory. The interactionmemory interplay hence forms the fitness concept in selection systems; although, as argued in footnote 5 of Sect. “Why the limited success of high-level selection theory? The hierarchical structure of the GSK”, the theoretical power of the fitness concept is often proximately much lower than it is in evolutionary biology. Synthesis, finally, is whatever sources of novelty that exist. As a minimum requirement, novelty must of course be possible to store in memory in the first place: novelty that is automatically forgotten (such as acquired traits of organisms) trivially has no fitness in a selection dynamics. Furthermore, in many knowledge systems competition will yield most sources of novelty nearly entirely unimportant. For example, the sort of variation introduced in photocopying makes negligible contributions to cultural knowledge in competition with human thinking.

It is important to stress that interaction, memory and synthesis are abstract functions. As soon as we speak of a specific knowledge system they become characterizations of things that material processes and structures are found out to be doing. That is, given a knowledge system we must ask: which structures and processes seem to serve these interaction, memory and synthesis tasks? What is contended here is that if we are indeed looking at a naturalistic knowledge system, then we will always be able to identify such structures and processes (Fig. 1).

But quite the opposite of claiming that on some bottom level we will find “interactors,” “memories” and “synthesizers” or some such, the expectation must be that the more simple and primitive our knowledge system is, the more general will the structures we find be. An RNA replicator as envisioned in the “RNA world” for example combines all three functions in the same structure (Gilbert 1986; Maynard-Smith and Szathmáry 1995), see also Fig. 2. As noted in Sect. “Discussion", it is not until we look at systems on a high-level of organization where there is a long history of BVSR adaptation that we can expect to see material bases structured along the proposed functional lines.
Fig. 1

Above is an approximate relation between function and structures/mechanisms in five knowledge systems. The internal captions implicitly refer forward and will become more clear when consulted later in the text
Fig. 2

Above is a schematic visualization of the proposed nesting structure of one aspect of the GSK. The color convention of the previous figures is used also here. “I” stands for Interaction, “S” for Synthesis and “M” for Memory

By identifying the material bases of the fundamental functions one can also begin to formulate a proximate theory for the system in question. When there are material bases (such as replicators) serving more than one fundamental function at the same time, then the proximate theory will be ontologically different from the general theory. It will still be possible to reformulate a proximate theory in terms of the general theory but by doing this we must lose much of the theoretical power associated with the proximate theory. This simply reflects the fact that there are many things that are peculiar to specific knowledge systems and that taking such things into account allows us to formulate a more powerful theory: a theory producing a better fit, as it were. The class of knowledge systems where we have replicators combining the memory and synthesis function is an obvious example of this. There are many things that the presence of replicators (that is, unless watered down) allows us to say theoretically, see in particular (Eigen 1971, 2002).

Using the theory: the hierarchical structure of the GSK

Every time we touch down in the respective systems of the GSK we also find ourselves in qualitatively new theoretical and empirical contexts. Needless to say, having to cover so much ground in only a few journal pages precludes anything approaching a proper comprehensive treatment. But since it is nevertheless necessary to argue for the potential of the proposed theory to make such connections to empirical reality the choice has been made to omit many questions altogether and instead dedicate the little space there is to a smaller number of issues.

The general plan of what follows is to pose and discuss some questions about the emergence of coexisting knowledge systems. Which modes of coexistence seem possible? For example: coexistence with separate origins, with common origins, in competition with one another and complementing one another. Relatedly, how do knowledge systems arise out of one another to generate a hierarchical structure? Can their relation to one another be interpreted using the proposed framework?

The emergence and coexistence of knowledge systems

The Hullian replicator/interactor Bauplan is universal among biological entities employing the \(\hbox{DNA} \rightarrow \hbox{protein}\rightarrow \hbox{morphogenesis }\hbox{machinery}.\) This quite complex theme (even in in its simplest forms) is in turn generally believed to stem from an “RNA world,” see e.g. (Gilbert 1986; Maynard-Smith and Szathmáry 1995), where we are decidedly beginning to move from a biological to a chemical discourse. On the geological time scale the DNA-based solution emerged quickly and if there have been any later usurpers—i.e. other “origins of life” or fundamentally different solutions with shared ancestry—they at least never thrived to the extent that they left any traces we have been able to find. It is hard to see an adaptive path leading to a fundamental replacement of the \(\hbox{DNA}\rightarrow \hbox{protein}\rightarrow \hbox{morphogenesis machinery}\) once in place; it is also hard to see how a spontaneously (from chemistry) arisen challenger would either avoid competition (there are only so many useful chemical compounds around) or compete successfully against an already sophisticated occupier of “the niche of being alive.”

However, what has occurred is the generation of new knowledge systems out of that which is responsible for the evolution of organic life. Here, some additional issues materialize: Why and under what circumstances can such new knowledge systems arise? Second, if such a system has arisen, there will presumably be knowledge present from several sources within the same structure: how may the turf come to be divided between them?

As for the first question, Edelman sees the general task of recognition as something that recurs between different systems operating on the principles of selection (Edelman 2004b): e.g. phylogenetic, clonal and neural selection. This recognition task could well be precisely the chief adaptive reason for knowledge systems to arise on top of one another. We must however still ask: why would a successfully operating knowledge system need a new “engine of knowledge”? Why does it not simply do the job itself? The answer to this question is also important for addressing the question of the division of turf.

One general such reason would be if the original knowledge system is somehow unable to operate efficiently in all relevant domains. Most importantly, if an interactor faces threats and opportunities on scales of time and space where its present knowledge processes are structurally hampered. The genetic machinery, for example, is hampered on short time scales by being locked into the cycle of organismal generations and, not least, by the utter lack of prescience of its synthesis function. The cycle of generations is necessarily considerably longer than the time scale of interactions during the life cycle and there is obviously little room for building any meaningful amount of prescience into the cell division machinery. This means that there can be an adaptive potential in exploiting the lacunae in recognition capability that are left by the structural constraints operating on the original knowledge system.

This also fits the way in which areas of operation of different knowledge systems are distributed in higher life forms such as ourselves. On the long time scale there is phylogenetic selection operating in the memory medium of genes; this knowledge system is capable of maintaining structures ranging very widely in size but it is unable to achieve instances of fit on the short time scale of events unfolding during an organism’s lifetime. On the intra-generational time scale, the turf is divided between neural selection on the longer scales and clonal selection on the shorter scales. We can not perceive and operate sensorily/neurally/motorically on time and length scales on which microscopic threats beset us; in particular we can not operate in the massively distributed and parallel fashion that is required. Likewise the cellular interactors of the immune system cannot cope with large scale threats and opportunities around our own length scale. Both, however, maintain a memory that stretches over the same time scale: that determined by the organism life cycle.

We now have a picture where it can be seen how interactors in an original knowledge system can come to evolve an ad hoc patchwork of knowledge systems tasked with maintaining vigilance over the whole relevant spectrum of scales. The adaptive rationale is that the interactor in its totality thereby fares better in the original knowledge system.

But there is yet at least two more important issues surrounding this hierarchical mode of coexistence: (i) There must be a credible evolutionary path leading up to the point where the new knowledge production begins and can confer fitness. (ii) In some cases the autonomy of a new system becomes so great that we no longer can speak of it just as an appendage to its “inventor.” These two issues will be addressed in order.

One logical route from a precursor to an operating knowledge system would be if the precursor did roughly what we get if the recognition ability is removed or replaced with something static. If so, dynamic recognition can enter smoothly as a boost on something already functional. In addition, we may expect vestiges of such original function to remain and there may even be extant cases of it in other corners of the biological world. This is also what a look at these systems seems to suggest:

Neural systems range in complexity right from one single neuron (the worm Myxicola (Donald 2001)) up to that of the human brain. So we know what primitive neural systems are good for. They exist all around us and much of our own neural system remains in such roles, e.g. reflexes. This static sort of function preceded advanced cognitive function and the latter can largely be seen as an elaboration or deepening of the former; in Lane’s terms expanded as a new set of “sandwiched” levels of organization between sensory and motor capabilities (Lane 2005).

The genealogy of the adaptive vertebrate immune system6 can also be traced back to instructional predecessors, see (Hoffmann 2004; Schulenburg et al. 2004). Furthermore, present immune systems of invertebrates, plants and fungi are innate and instructional. Elements of this previous instructional solution can still be seen in operation also in vertebrates: what is called the innate immune system is not recombinatorially adaptive and it operates alongside the full adaptive immune system as a first line of defense along with physical barriers (such as skin) in the role of keeping pathogens out (ibid.); this clearly invites analogy with neural responses without higher cortical involvement such as reflexes.

For what concerns cultural knowledge systems the original type of system would be innate features of animal life such as behaviors and nests as well as knowledge stemming from lifetime experience but that perishes with the knower. There is of course no shortage of clear examples of genetically defined animal behavior. For learned behavior that begins to border on, and even cross over into, being culture it is in particular the Great Apes that offer examples. Recent field studies of chimpanzee groups in Africa shows that an unexpectedly large part of their behavioral repertoire is of a character that would qualify as cultural on many reasonable definitions (Boesch 2003; Sanz and Morgan 2007). Indeed, to my knowledge, the chimpanzee is the only species except for humans (and our closer extinct relatives) that has been studied archaeologically: it has been established that an observed nut cracking behavior (which is not universal among present chimpanzees) has been practiced for at least 900 years in an area (Boesch 2003; Mercader et al. 2002). Even if an immensely complex machinery has been sandwiched between our sensory and our motor capabilities it is still behavior from neural activity that culture elaborates.

Once a new knowledge system has been set into motion—bringing adaptive value to its host through the feats of recognition that it is capable of providing—it is easy to see how it rapidly could make itself indispensable. Since it arose because there was an evolutionary need for its fruits and because the original knowledge system could not perform that sort of recognition, it will be subject to selection for increasing versatility; indeed just like any physiological capability. This must often mean that it step by step becomes more autonomous in its operation7. Most of the biological knowledge subsystems—belonging to either the neural or the immunological categories—have remained under sufficiently tight reins that we experience little discomfort seeing them simply as organ functions of biological entities. In some cases, however, and most importantly in that of human culture, the connection between what happens in the system and the fitness contribution it confers (as well as the identity of the recipient) becomes so abstract and stretched out that many relevant questions about its operation are best posed entirely in terms of the system itself. For example, if we are interested in the technological evolution of automobiles it becomes quite unsatisfying to only refer to the biological adaptive value of the cognitive capabilities that made automobiles and their evolution possible. We can then speak of a considerable amount of autonomy but at the same time it is also clear that we should not fall for the temptation to think that this means that biology is universally irrelevant in the study of culture.8

Cross-access from parent- to child systems of parent synthesis processes

So we have that the original immune system attacked and eliminated microscopic threats, only it did so in an instructional and less effective way. We have that the original neural system matched sensory and motor neural activity, but in a static and mechanistic way. Finally, there seems to be cases that interpolate between neurally hardwired behavior and culture, and we can clearly see what neural systems did (and very much still do) when not concerned with culture.

What we see is a recurring theme of access of the original system’s synthesis processes to the new system’s memory. As the new synthesis process becomes refined the original synthesis bit by bit cedes power and it does so mainly in a particular way: it becomes more and more abstract. For example, the adaptive immune system is genetically controlled but on a quite abstract level: rather than coding for specific antibodies, a mechanism capable of producing an immensely wide array of antibody configurations is coded for (Alberts et al. 2002). This is also readily exemplified in Mayr’s distinction between closed and open programs (Mayr 1989), where an open program is a genetic neural strategy on a more abstract level where “the blanks” are to be filled in by life-time experience—both in the shape of learning and in the shape of composing adequate responses to external stimuli from more strongly genetically determined behavior fragments (Mayr 1989; Lorentz 1977).

As discussed in Sect. “The emergence and coexistence of knowledge systems", what happens is that the new knowledge system begins to explore a niche where it complements the original system: one concerning interaction on scales where the original system is unable to perform efficiently. But since there must be an evolutionary path to the new knowledge system, such systems must typically stem from pre-selection strategies devised by the original knowledge system to the end of dealing with the same tasks. These strategies will naturally be stored in the memory of the original knowledge system, but their manifestation as interaction-serving structures and processes will also be likely to form the foundation of the memory of the emerging new system. As knowledge accumulated with the long BVSR period of the original system becomes redundant with knowledge accumulated with the shorter BVSR period of the new system the old system must be expected to adaptively specialize on creating preconditions for the operation of the new system rather than directly competing with it. This means that the original knowledge system will specialize on writing things that benefit from the long period of its operation while increasingly refraining from specifying things that are better figured out by the fast knowledge process.


Interestingly enough, as developed in Sect. “Using the theory: the hierarchical structure of the GSK", all known instances of knowledge systems are related to one another in a hierarchical fashion forming together what was here referred to as the GSK. Furthermore, none of them would be possible (that is, not be astronomically unlikely) without the spontaneous (in the sense of arising for reasons other than BVSR) emergence of biological systems. Knowledge in all forms are thus anchored in Nature in a particular way and, if the argument above was successful, it seems that for each new autonomous knowledge system, new means for serving the functions of interaction, memory and synthesis have been “re-invented.” One might say that the adaptive opportunities offered by “systems of recognition” (Edelman 2004b) have many times over favored their emergence. Looking from the perspective of the general theory proposed here, it was then argued that certain patterns came to light regarding how new knowledge systems tend to emerge from previous ones. This disclosing effect—not unlike Heidegger’s aletheia concept (Heidegger 1962)—summarizes the chief roles of a general theory: to provide fresh perspectives on already studied specific problems and to bring forth new angles of attack that were previously invisible.

Some important implications on cultural evolution worth mentioning: (i) Proximately, cultural knowledge changes only partly according to BVSR principles. (ii) Selection theory is essential for understanding cultural dynamics, but it is not sufficient proximately by itself. Human prescience and its phylogenetic history hides the bulk of the wastefulness needed to account for human knowledge as the fruit of synthetic a posteriori accumulation (Campbell 1960, 1974); which seems to be the only naturalistically tenable Kantian category (Lorentz 1982). (iii) With a neural selection system (Edelman 1987, 1993, 2004a; Changeux 1985; Changeux and Connes 1989; Changeux and Ricoeur 2000). on which to blame instances of fit that seem impossible to account for through cultural BVSR, this non-BVSR aspect of cultural evolution poses no metaphysical problems to the sworn naturalist; other categories of knowledge may exist proximately but only as an effect of an encapsulated hierarchy of BVSR dynamics.


Culture is here used very broadly to denote everything that potentially accumulates as a result of human cognitive activity; not unlike Popper’s “3rd world” or Simon’s “the artificial’ ’(Popper 1979; Simon 1969).


What is it to be prescient? It is important to briefly discuss this question since it has such a strong bearing on the explanatory values of selection theories. The answer depends on what we mean by the question; in this case on what we are willing to qualify as being prescient. If we mean that man can know what the future holds in an absolute sense, then the scientific answer must be “no.” But, then again, in that sense it seems that she does not know anything at all regardless of whether the proposed object of knowledge lies in the future, past or present. If we instead mean that she is capable of making statements about the future that are sharply biased toward correspondence with future states, then the answer must be “yes.” Prescience, in this sense, is nevertheless a remarkable ability: it is decidedly something more than an ordinary fit between system and environment; it is more than just inductive achievements gained through past BVSR dynamics. It is, however, by all means an empirical rather than a metaphysical problem.


The wide naturalist notion of knowledge used by Campbell is employed here (Campbell 1960).


In Dennett’s terminology, cranes are ok but skyhooks are not (Dennett 1995).


Campbell characterized knowledge as a “fit between system and environment” (Campbell 1974). Although I agree with what Campbell means (on my reading at least), this characterization is still fraught with problems and one needs to be clear on what is intended. The terms “fit” and “fitness” have particular theoretical uses in evolutionary biology that hinge on the soundness of assuming environments to be sufficiently stable over time. For example, we may ask what happens in a population if a trait appears that has such and such fitness. Now, regardless of the constancy of this trait, its reproductive potential of course depends on its interactions with this environment. Hence, fitness may change at any time if the environment to which “it fits” changes. This happens seldom enough (or in a regular enough fashion; e.g. as an arms race between members of two species) in biology that fitness remains theoretically useful. In cultural systems, however, one must seriously wonder if it does not happen frequently enough—almost as a rule—that many of the traditional theoretical roles of fitness disappear. The usage here of “instances of fit,” “fitness” and other related concepts must be viewed with these complications (which need to be worked out) in mind.


Excluding some of the closer extant relatives of jawless fish, e.g. hagfish and lampreys.


If there is something to win autonomy from, that is. Biological knowledge systems, since they arose from physicochemical systems that were not knowledge system themselves, have no parent knowledge process.


There is of course also an existential dimension to the particular case of culture and social life since the sense in which we are our biological selves in the first place is not straightforward, see (Sartre 1993)



I wish to thank in particular David Lane, Camille Roth and other researchers and students at the University of Modena and Reggio Emilia, Italy, for enlightening discussions on these topics. I also want to thank Martin Nilsson Jacobi, Kristian Lindgren and several others in the center for complex system at Chalmers, Göteborg, Sweden. Peter Nylén has also provided his valuable insights and given me the opportunity to discuss these question in depth. Finally, I wish to thank Kim Sterelny and a number of anonymous referees for their efforts with this manuscript. The work was funded in part by the University of Modena and Reggio Emilia and in part by Chalmers. Also, the European Center for Living Technology, University of Venice Ca’Foscari, Italy, has kindly provided space and other resources.

Copyright information

© Springer Science+Business Media B.V. 2007