The replicator: a misnomer. Conceptual implications for genetics and memetics

 Mario Vaneechoutte
  •  Department of Clinical Chemistry, Microbiology & Immunology
  •  Blok A, University Hospital, B 9000 Ghent, Belgium
  •  E-mail:
  • Most evolutionary biologists and students of evolution in general have adopted the term 'replicator' for both genes and memes. There is indeed almost unanimous agreement [12]. However, current genes are not (self-) replicators and, as I will argue below, in contradiction to the current paradigm of the RNA-replicator world hypothesis, they have never been. Thus far, I came across only one source which defends a more correct interpretation of what genes are:

    Indeed the suffix -or refers to an active agent, a processor. However, current DNA-genes are nothing but a material instantiation of information on how to make RNA and proteins, and perform no replication. Genes are replicated informational molecules, unable to replicate on their own, no more than enzymes can. Moreover, replication relies on the interaction of a whole collection of very different genes. Accordingly, 'memes' or bits of cultural information are processed, replicated and transmitted by human minds, photocopiers, presses, computernetworks, etc. The original prebiotic genes are then better compared to plasmid genes which can be exchanged freely between different bacterial cells (processors) - OR with current memes (like printed scientific theories). They then can be regarded as elements of a language which enabled free recombination of different concepts, just like current scientific reasoning is possible because of the availability of copy true bits of information (printed matter), which can be recombined over and over again.

    Partially due to the application of the term replicator for genes and memes, it has been overlooked that only the system cell as a whole enables self replication. It is important to emphasize that such a self replicating system thus far originated only once during the existence of Earth. Although it are the genes which carry heritable information, it is clear that daughter cells also need to inherit a minimal amount of processors, i.e. at least those enzymes which can transcribe and translate the DNA. So, although it are genes which are responsible for heritable information, it takes more than genes to be inherited to enable replication of genes, to make the processing system which can do so. Moreover, not only genes and some enzymatic processors, but also membranes have to be inherited, since membranes are not synthesized de novo, but are synthesized only in close connection with the existing membrane. (This points to a very central and original role of membranes or to a 'membranes early' hypothesis of the origin of life.)

     As a matter of fact, one could even consider new nucleotide strands as being a phenotypical artefact, since they are the product of enzymatic activity just as well as other cell components. Notice how this illustrates that the genotype-phenotype distinction itself has only limited usefulness, and - certainly at the cellular level - it occurs to be a somewhat artificial human-made dualistic distinction [see also 7], which in itself obstructs useful insights. One might object that for DNA-duplicates to be considered as phenotype, there is a difference with other phenotypic traits, namely that new DNA can be made only by using existing DNA as the mould. However, this observation accounts as well for membranes (see above). The only difference is the 'unlimited' [11] informational content of nucleotide molecules, but this characteristic is unlinked to replication.

    I propose to adopt a terminology which is also used in information theory (and related fields like cybernetics)[see also 7] and to drop words like genotype, phenotype, replicator, interactor and vehicle, which have the disadvantage of poorly reflecting the dynamics and interactions which occur between biological processors and between biology and culture and which are in many cases strongly misguiding. It is more revealing to speak of digital information processors (e.g. polymerase enzymes) and encoded, digital information (genes). In case of self-replicators, that is cells, both processors and digital information are embedded in a housing which consists of structural molecules (proteins, peptidoglycan, phospholipid bilayers, polysaccharides, ...).

    From these remarks, more useful definitions then can be put forward: living cells are the only self replicators on Earth. Enzymatic polymerases, resp. human brains, copiers, presses, etc. are all processors which can function as replicators of informational molecules, resp. informational symbols. These processors are information processors. Finally, informational molecules (polynucleotides) and texts, ideas, words, behaviours are replicable elements of instantiated information. For this category, we might adopt the term 'replicanda' as it has been proposed by Ghiselin [8], although the exact translation of this Latin term is 'that what has to be replicated'. This definition then only holds for extant chromosomal genes, since indeed the cell is compelled to copy its genes in order to produce cellular offspring. However, it should be noticed that, certainly in the case of cultural information, the successful reproduction of information processors (like humans) does not depend upon the replication of cultural information, and therefore that there is no functional need to replicate this information.

    'Replicata' or 'replicates' ('that what is/can be replicated') might be a more general term for informational molecules (nucleotidic genes) or cultural bits of information (memes) which are replicated.

    In summary, the use of replicator for gene or meme is twice erroneous: they are not self-replicators (remark that 'replicator' is generally used to have the implicit meaning of self-replicator) and they even are not processor-replicators. Genes and memes are replicates or bits of replicable information. For memetics this means that we have to ask the question why some information is replicated more successfully by replicators like human minds and not why they replicate more successfully than other memes. This more scientifically correct approach is e.g. used in the paper of Heylighen [10].

    Selection, natural selection and cultural selection

     These definitions of self-replicator, replicator and replicate lead us to considerations on the essence of natural selection as a special case of selection. Selection is rather a tautological principle: whenever there is variation on a theme, one of these themes will be best fitted to certain conditions, given by the environment, which also includes the other variations on the theme. Having different kinds of cars, one car will be best suited to reach high speeds on a straight road, one car will be best suited for riding a curly parcours, one car will be best suited for transporting many people, one car will be best suited for driving and parking easily in a crowdy city, one car will confer an image of certainty and successfulness to its owner. In case most people live in cities and do not care too much about confirming their self image by means of the car they own, a type of small car that enables easy driving in the city will be sold most and will therefore be most successful when measured by number of copies made of it.

     Specific cases of selection then can be understood by defining the theme, its possible variations on the one hand and on the other hand the environment to which the different possible variations on the theme will have to fit.

    All of these general considerations on what selection is hold as well for natural selection. However, natural selection is often erroneously considered as the only possible form of selection and it is forgotten that it is a very special case of selection, because the selection occurs among self-replicators (that is, cells), unlike cars, prebiotic metabolic cycles and ideas. Moreover, all the self-replicators stem from the first system able to duplicate. Since this duplication is not watertight copyproof, several variations on the original theme (the first cell or the theme of self-replication) are continuously produced, with as a 'welcome' side-effect that in most cases some of the variations will fit to whatever environment or whatever environmental change.

     In summary, because a self-replicating system once occurred on Earth, it was possible that the original genetic code of it - wich contained all the necessary information on how to make the next self-replicating system, was copied manifold. High copy fidelity, ensuring sufficient numbers of functional self-replicators, allowed also for a number of mutants to exist. Some mutants also could survive. Some mutants even could explore new environments or did better in the old environment than their parent wild types. A small percentage of erroneous descendants is an ensurance for adapting to previously not inhabited environments or for adapting to the changing environment.

     These considerations stem from an information centered approach which leads to more generally applicable insights than a gene centered one.

    Natural selection is about the replication of informational molecules whereby the replication of the information depends on the functionality of the organisms, i.e. the genetic information processors for which the information encodes. The fitness of the phenotype is also the fitness of the genes, and vice versa. Cells are metabolically open, but semantically closed systems. From this definition the difference with cultural information becomes obvious.

     Just like variable genetic information and variable information in general, variants of cultural information are selected as well. I referred to this as a tautology.

    For example, different scientific theories are weighed by the scientific community, whereafter finally one or still a third, or a recombination of still another five ideas appears to solve best the problems tackled and is replicated more successfully than other theories. The difference with natural selection is that the defenders of one theory - the scientists, the information processors - do not die because their theory dies or that they do not produce more offspring because they defended a successful theory. The information encoded in a meme (usually) has no influence on the survival of the processor (a human in this case). Vice versa, the ideas of a parent of many children are not necessarily - certainly no longer at present - successfully replicated ideas. Scientists with many children are not necessarily successful in spreading their theories. Christ had no children. The evolution of cultural information does not depend on the differential survival of its processors to proceed.

    By reversing the adagium that memes must be replicators in analogy to the fact that genes are replicators into the statement that genes are replicates just like memes are, we can further suggest that genes were also replicates in prebiotic life. This suggestion contradicts the current paradigm that life evoled by competition between self-replicating RNA-molecules, whereby the cell gradually evolved around the fastest self-replicating RNA-molecules by means of natural selection, i.e. as a consequence of the differential reproduction rate of different self-replicating molecules. We propose to postpone natural selection till after the origin of the first chromosome containing cell (the first self-replicator) and to consider how a much more flexible and powerful kind of memetic evolution took place, comparable to what is happening since the origin of symbolic language in the animal world.

     (It should be noticed that this does not mean that we have to dismiss the several exciting findings of RNA-research [e.g. 3,14] nor that we disagree with the 'nucleotides earlier than complex proteins' hypothesis. Only, we propose to consider these findings from a different point of view.)

    Symbolisation of information thus far has occurred twice on Earth. Prebiotic life developed a manner to encode information (symbolisation) into informational molecules (nucleotide strands). In biology, informational communication first took a nonchemical way of transmission with the advent of neuronal brains in animals which could make use of nonmaterial sound wave energy and photonic energy (vision) to transmit information (the origin of memes), whereafter strong symbolisation of this nonmaterially transmitted information has occurred in humans by the development of spoken language. This has finally led to the re-entry of material substrates which could carry information as visual symbols, in the form of written language. I will argue that the initial role of nucleotides is comparable to that of written language. While nucleotides enabled to encode chemical interaction, symbolic language enabled to encode behavioural interaction.

    The essential thing about encoded information is that it can be mixed, recombined, merged, over and over again, because of the fact that it is not a metabolic process but a code on how to perform such a process, which has the virtue that the information is not transformed (only transcribed or translated) during interaction. It is impossible to mix processes: One can't mix the activity of enzyme x with that of enzyme y, one can't take the bacterium E. coli and mix it with a piece of human tissue, while keeping a functional process, one can't mix the ideas of two people by mixing their brains.

     But what if these processes also are encoded somewhere? Evolution has managed many times to combine the functions expressed in different modules of enzymes with each other. This could be done because there was a genetic code for it and by recombining parts of this code, something completely new, composed of already existing components but arranged in a different manner, could emerge. Similarly we can mix the eubacterium Escherichia coli with human enzymes, by inserting into its genome the code for such a human enzyme (e.g. for the production of insuline for diabetes patients).

     We can mix different ideas since they can exist as a memory in our mind. This process is largely enhanced by the availability of written records, since this provides a more faithful, copy true back up. One can mix as many of these ideas as one can gather (e.g. by reading). Many different lineages of information can continuously come together. Sometimes (but apparently rarely) this leads to new useful insights. Furthermore, it takes only one processor and one template to make an unlimited number of copies of encoded information, something which is not achievable when dealing with processes.

    The following is just a brief exercise of how we could envisage analogies between 'memetic' evolution towards the origin of the genetically encoded cell and present day cultural evolution (which may or may not lead to a novel kind of 'memetic' self replicator, e.g. some kind of self-assembling, duplicating robot as suggested by Tipler [13]).

    Not in contradiction with some current thinking [e.g. 6] we assume the existence of a premetabolic network of interacting chemical molecules, organized in smaller networks of heterocatalytic cycles, mostly embraced by membranes. Such a membrane with its internal metabolic cycling can be called a protocell (which is devoid of genetic material), which interacted intensively with other protocells and/or free-'living' cycles and available molecules. The prebiotic system as a whole can then be compared with the current ecological network, which is composed of interacting genetically encoded organisms of different complexities.

    The basic proposition of this paper is that symbolisation of chemical interaction (communication) between these protocells and/or parts of their internal networks is suggested to have been the original function of nucleotidic information. Instead of direct metabolic interactions (comparable to direct behavioural interaction between animals), some cells also started to make use of informational molecules, which not directly took part in the interactions but which nevertheless could influence the behaviour of other protocells (or of intracellular cycling), just like symbolic sounds or written texts can influence the behaviour of other humans (or of 'intrapersonal' thinking). (Remark that information was defined by Bateson as a difference which makes a difference to 'someone' else [7].)

     Written/printed information, just like nucleotide information, can than be used without being transformed (i.e. analogously processed) during this interaction: it is digital.

    With respect to the evolution of information, the chemical cycling within protocells is best compared to the animal body, performing metabolically essential functions. The membrane is best compared with the neural system (incl. brains and perception organs) of animals, since it performs interactive, communicative functions. Like some of these protocells started interacting by means of symbolic molecules (informational molecules, stretches of nucleotides), some animals (humans) started interacting by means of symbolized gestures and sounds. This leads to new interactions, also with nonsymbol using protocells (accordingly with other animals, ..., bacteria). E.g., this symbolic information may have enabled the symbol using protocells to use metabolic energy from simpler protocells, like humans started to use animal power. This is to be regarded as some kind of symbiosis, since both symbol using and nonsymbol using protocells can have metabolic profit, like the number of horses increased correspondingly with the number of humans.

    (Remark that an inadvertent outcome of this kind of reasoning is that this story is about symbiosis and mutual profits, in contradiction to the 'natural selection early' approach whereby the selfish fastest replicators won the game.)

     While the informational molecules may have been internalized and processed at first in order to create new ones (like humans process words in an analogous manner: transformation), the interaction between protocells and informational molecules may have lead to the production of external processors (metabolic and informational enzymes). Similarly, humans aided by words have produced different external metabolic processors (wheels, ploughs, tractors, factories) and information processors (pens, type writers, presses, copiers, computers).

     Because in this model the replication of informational molecules presupposes only the ongoing of metabolic activity in general and thus not depend on the direct 'survival' of some particular protocell - unlike in e.g. models of molecular self-replicators and chemotons [11] and unlike present day chromosomal genes, both cases where the fate of the informational molecules is tightly linked to that of the metabolic efficiency of its cell - a lot of recombinatorial freedom was available to prebiotic chemistry. Similarly, cultural information can be recombined in an at random manner, whether or not this leads to silly ideas (like this one!?), because it is not linked to the fate of the replicators (human brains, computers). In fact the interaction between protocellular membranes, nucleotide encoded information and extracellular processors may have enabled to produce large extracellular heterogeneous protochromosomes, just like the interaction between humans and printed code and external information processors like presses and computers resulted in the production of scientific theories and complex machines.

    The whole community of (simple protocells and) symbol using protocells became more and more dependent on the usage of external informational molecules and external informational and metabolic processors. The same happens in our present society, which would collapse at once when language, printed information, production machines or computers were taken away. In the end, most metabolic functions and most information processing were done by digitally processing external enzymes (e.g. polymerases), so that the role of the original protocells was merely a motivating one. Accordingly, humans are less and less needed to keep the information processing going on. We put in a search string, push some buttons and the world wide web searches the relevant information. We are more and more needed only as motivators. Current co-enzymes (motivators of enzymatic activity) accordingly can be regarded as remnants of the original premetabolic protocells.

    Finally, protocells consisted of a membrane, probably flattened since at this stage it contained only few internal chemical cycles, with an externally attached DNA-protochromosome - consisting of genes encoding for very different functions - and some externally attached enzymes and RNA-molecules. At several occasions such membranes may have 'gastrulated', whereby the chromosome and the enzymes and possibly some external medium with free enzymes and other molecules were internalized (in accordance with the obcell theory [1,2]). At one such occasion the enclosed informational molecules and processors contained all the necessary information and functionality to produce a self-replicating system (remark that reverse transcriptase was not enclosed). A similar event has not yet happened in culture (self-assembling robots [13] are just one imaginable outcome).

    While explaining the origin of life as a memetic process, this approach also might provide a unifying theory for the evolution of information, whereby at two events symbolisation of information occurred - once of chemical interaction, once of behavioural interaction - and whereby at one occasion this encoding lead to a chemical self-replicator (the biological cell), a memetic evolution which seems to be happening again.

     This symbiotic approach may circumvent many of the current problems like the 'chicken or egg' problem of DNA/RNA/protein interaction, not (yet) solved by the RNA-replicators world hypothesis. It also more easily explains how a chromosome consisting of very different functions, and whereby none of these genes by theirselves contain sufficient information to cover the complex process of self-replication, can have originated. It circumvents Eigen's paradox whereby different self replicators outcompete each other when brought together in a cell, instead of merging into a chromosome (Eigen's paradox is solved (?) only by the rather artificial stochastic corrector model [11]). The much faster and flexible memetic (cultural) evolution also might resolve the time window problem (whether real or not [see 6], which lead some eminent biologists to adopt an extraterrestrial origin of life [4]. Of course, this 'memetic origin of life' hypothesis poses its own problems.


     1. Blobel, G. 1980. Intracellular membrane topogenesis. Proceedings of the National Academy of Sciences, USA 77: 1496.

     2. Cavalier-Smith, T. 1987. The origin of cells: a symbiosis between genes, catalysts, and membranes. Cold Spring Harbour Symp. Quant. Biol. 52: 805-824. BioMedNet Medline Search Abstract.

     3. Cech, T.R. 1986. A model for the RNA-catalyzed replication of RNA. PNAS, USA 83: 4360-4363.

     4. Crick, F.H.C., & L.E. Orgel. 1973. Directed panspermia. Icarus 19: 341-346.

     5. Dawkins R., 1976, 'The Selfish Gene', Oxford Univ. Press.

     6. de Duve, C. 1995. Vital dust. Life as a cosmic imperative. Basic Books. New York.

     7. Emmeche, C., & J. Hoffmeyer. 1991. From language to nature. The semiotic metaphor in biology. Semiotica 84: 1-42.

     8. Ghiselin, M.T. 1994. Chapter 9, pp. 143-148. In Metaphysics and the origin of species. SUNY Press, Albany, NY.

    9. Ghiselin, M.T. 1987. Replicators and replicanda. Bioeconomics and the metaphysics of selection. J. Soc. Biol. Structures 10: 361-369.

     10. Heylighen, F. 1998. What makes a meme successful? Selection criteria for cultural evolution. First Symposium on Memetics, Namur, Belgium.

     11. Smith, J.M., & E. Szathmary. 1995. The major transitions in evolution. W.H. Freeman & Co. Ltd. Oxford, UK.

     12. Speel, H.-C. 1998. Memes are also interactors. First Symposium on Memetics, Namur, Belgium.

     13. Tipler, F. 1995. The physics of immortality: modern cosmology, God and the resurrection of the dead. Anchor.

     14. Zhang, B., and T.R. Cech. 1997. Peptide bond formation by in vitro selected ribozymes. Nature 390: 96-99.