[Published in: Cybernetics and Human Knowing vol 7(1), 2000, pp. 45-55.]
Organisms can be proud to have been their own designers
According to H. F. Osborn, one of the three authors of 'Baldwin effect', adaptive evolution may not require neither natural selection nor the inheritance of acquired characteristics. An adaptive evolutionary change in population without natural selection means that an identical adaptive change in genetically different organisms of a population can take place without a systematic difference in the reproductive value between them, and these changes can also become irreversible on the level of genome without the difference in the reproductive value involved. The mechanisms which allow this are known and sketched in this paper. Their description requires an approach on the level of whole genome and a look to the organism as a self-organising and communicating system. Consequently, it is possible to have a theory of adaptive evolution, for which the evolution with natural selection is a special case.
Keywords: Self-organisation, autogenesis, Baldwin effect, biosemiotics, post-Darwinism, individual adaptation, functional genome, gene duplication, gene conversion, adaptive evolution
1. Introduction: Interpreting the ‘Baldwin effect’ as a post-Darwinian mechanism of evolution
"The root-ideas of the conception of evolution are, first differentiation, and secondly the interaction of the differentiated products", wrote Lloyd Morgan (Morgan 1898: 487). This very general statement can be interpreted, for instance, as a differential reproduction, followed by competition and extinction of one of the product. However, Lloyd Morgan's emphasis was evidently the differentiation within an organism, or more precisely, the differentiation into the non-self and self (Blitz 1992). That is the problem of what is the role of distinctions made by organism in evolution.
According to a principal statement of semiotic biology, organism as a subject has activity and intentionality (cf. Uexküll 1940, Searle 1993), which are the main essentials of life. Also, if life has any way to influence its evolution, these features of the subject should play the role in this. However, it has not been easy to find an evolutionary mechanism which may correspond to this view.
The role of self-organisation is generally accepted as evident for ontogenesis. Also, a population can be seen as a self-organising system in evolution. However, the question I would like to analyse here is whether an organism can be seen as a self-organising system over the generations, as a leader and designer of its own evolution.
The view according to which the self-organisation of an organism is limited to its ontogenesis, comes from the well-known assumption that the organism cannot influence in any way the internal determinants of its offspring, i.e., its genome is inherited independently of the behaviour during its lifetime. Still, for instance, F. J. Odling-Smee (1994) has emphasised that organisms can influence the external determinants of its offspring, via choosing or changing the environment in which the offspring will develop. This is, however, not enough to make any change irreversible, that is evolutionary. Thus, he requires the natural selection in the second step to make the change irreversible. My statement here is that this is not necessary. In order to solve the problem we need to look how an organism is operating its memory, and we also need to be precise in using and defining the notion of natural selection.
One of the interesting alternative approaches to evolution was proposed a century ago by paleontologist Henry Fairfield Osborn (1857–1953) and psychologists James Mark Baldwin (1861–1934) and Conway Lloyd Morgan (1852–1936), known as the concept of organic selection or ‘Baldwin effect’ (Baldwin 1896; Morgan 1986; Osborn 1896). The Baldwin effect states that the ability of individuals to learn can guide the evolutionary process. This effect was later analysed by G. G. Simpson (1953) and C. H. Waddington (1953a, b). Waddington has used it to develop his model of genetic assimilation. Recently the Baldwin effect has gained attention again, which is expressed by the book of R. K. Belew and M. Mitchell (1996), and several other publications (Emmeche 1994; Deacon 1997; Jablonka et al. 1998; Kull 1998, 1999; Ancel 1999; Robinson, Dukas 1999; cf. also Bowler 1992, Richards 1987). With the work of Hinton and Nowlan (1987) the Baldwin effect came to the attention of computer scientists, and since then it became a tool in evolutionary computation (Turney et al. 1996, Turney 1996, Harvey 1996). Here, I propose a possible mechanism for the ‘Baldwin effect’, which uses the concept of phenotype as an interpreter of genotype, and develops the interpretation given by Simpson and Waddington in a more radical way (e.g., as emphasised by Osborn (1896) — adaptive evolution may not require neither natural selection nor the inheritance of acquired characteristics, but may use natural selection in some cases).
2. A mechanism of microevolution
The mechanism to which I would like to draw attention, can be briefly described as consisting of the following statements. Most of these steps are quite trivial, however, the consequences from the whole set will not be so trivial.
(1) We need to notice that an organism can choose, which parts of its genome to use. Accordingly, the genome includes a functional and non-functional part. In most eukaryots, only a minor part of the genome is in use. The unused part consists (a) from non-coding DNA, which, still, may include some pseudogenes, and (b) from coding DNA, which is not currently used, but can be used in some other circumstances (e.g., in other cells of the organism, other period of ontogenesis, or other environmental conditions).
(2) An organism is able to change the used (functional) part of the genome during its lifetime, as dependent, for instance, on the conditions in which it lives.
(3) This change can be adaptive (due to self-organising nature of the organism) even if the conditions are new, that means if the particular combination of environmental factors has never been experienced before by this organism during its phylogeny. This derives from the nature of cell, as the cell is an adaptive system that changes its state continuously according to the communicative activity of its functional cycles.
(4) An organism as a self-organising system is able (due to its mobility and an ability temporarily not to move; and also, due to its ability to distinguish between different environmental situations) to choose the environment for living.
(5) If a population of similar organisms will meet new environmental conditions (when moving to a new place, or when the conditions change by themselves, or when the change in conditions results from the organisms local activity), then an adaptive change in the usage of genome may take place almost simultaneously for most of the specimens of the population.
(6) If the population keeps living in the new conditions (either due to the organismal preference for it, or due to the constancy of the new conditions), then the change in the usage of genome can be kept (repeated) over a number of generations.
(7) In the case of biparentally reproducing organisms, the organisms of a population keep to be similar. This is because if the genomes of the mates are not similar enough (i.e., recognisable or complementary), they cannot reproduce. Thus, the mortality due to the occasional big genetic changes is not specific to a genotype, but depends on the difference from the population mean.
(8) Due to the great similarity of individual genomes in a population, the simultaneous adaptive change in the usage of genome (in gene expression) may concern the same or similar areas of the genome in most of the organisms of the population.
(9) Some stochastic mutations, which appear in the area of the genome which is used (expressed) in the current conditions, lead to inviability of a part of the offspring. This appears with a probability which is proportional to the used part of the genome. Considering that the size of the expressed part of the genome does not differ significantly between the individuals of the population, it means that also the probability for such mutations cannot differ significantly between the individuals. I.e., the corresponding offspring mortality is not specific to a particular parental genotype.
(10) The mutations which appear in that currently unused fraction of the genome which has been in use in the previous environment, make the return to the previous functioning of the genome impossible. There is a great number of such mutations which may lead to this irreversibility. Since they appear in non-functional part of the genome, they do not influence the viability of organisms in the current conditions specifically.
Natural selection is defined as differential reproduction of genotypes (this being used as the standard definition of natural selection according to Dobzhansky et al.). Differential reproduction means here that there is a non-random (i.e., statistically significant) difference in reproductive value between the groups in a population which distinguish by some genetic marker. This type of difference was not required to reach the statement (10), i.e., non of the statements (1–10) assume any necessary difference in reproductive value between specific genotypes.
Thus, the statements (1)–(10) described the way of adaptive evolutionary change without a need for natural selection. This way of evolution means that some genes may be dropped out from the expressible genome during an adaptive specialisation of a population without any differential reproduction of genotypes involved. It can be called an evolution via forgetting of unused.
The mechanism described here is a Baldwin effect, since it assumes the organisms' adjustment, or individual adaptation, which can be seen as a form of individual learning. Accordingly, we have described a possible particular mechanism, which allows individual learning to guide certain type of evolution without natural selection being its requirement.
The importance of this mode of evolution concerns the possibility to have an adaptive evolution without natural selection. There have been described several ways of evolution which do not require natural selection, the most well-known among them being the random walk, i.e. neutral evolution. However, these cannot create adaptations. In this sense the mechanism described here is exceptional.
3. Can a growth of functional genome be non-selectional?
If the forgetting of unused would be all of evolution, it would mean a progressive narrowing of working genetic memory. Since this is not generally the case in real evolution (one can find an evidence for this probably only in some examples of specialisation), we need also to analyse the mechanisms in which the genetic memory can grow.
It is very improbable that in many individuals of a population an identical growth of their genome occurs simultaneously. Consequently, the growth of functional genome without the work of natural selection would be much less probable than its decrease. Still, the current knowledge is not sufficient to state that the lower probability would mean impossibility.
A phenomenon which appears very interesting in this context is the ability of homologous chromosomes to stay extremely similar. This is usually explained as a result of homolog recognition — in the case of big differences meiosis does not go correctly and the sexual reproduction will be inhibited. However, there are known few additional exciting details. In the course of recombination also certain recombination repair of DNA takes place. At least in some cases, it may include a mismatch repair between the strands of DNA which are derived from different parents. In this case, gene conversion can occur, where the mismatch repair can convert one allele into the other. This phenomenon can be detected through the offspring non-Mendelian ratios, and is frequently observed, e.g., in fungal crosses (Stacey 1994).
Whether the gene conversion may account for distribution of gene duplications in a population, we do not know. However, this seems to be a reasonable hypothesis. Namely, assuming that gene conversions can lead to an appearance of a gene duplication in the DNA strand derived from the parent that did not possess it, and, at least sometimes, with higher probability than the dropping out of the duplicated gene, then this may have important consequences for the genome evolution. This phenomenon — a rapid non-Mendelian distribution of a gene duplication in population — if it occurs, should be quite rare. The speed of real evolution is not great.
Considering the hypothesis above, we may sketch a mode of progressive evolution without natural selection.
(1) The genome may grow, for instance, through duplication of genes. Generally, there is no reason to assume that the reproductive values differ significantly between the organisms which have a particular duplication and which have not. The duplication can be not an exact copy, but still a gene, which is simply not expressed and not needed when it appears. A duplication, as well as any other mutation, can spread over the population by random walk. In addition, it can be that some recombination repair mechanisms enhance the distribution of duplications.
(2) In the case of several copies of a functional gene, one may mask the existence and slight differences of the other. The changes in the masked copy are nevertheless restricted and kept similar within a sexual population, in order to allow the sexual recognition (on the level of pairing the homologous chromosomes) to take place.
(3) If most of the organisms of the population have obtained in this way a masked and slightly different copy of some genes, these can be used in a new way when the behaviour of the organisms will change in the changed conditions of the population. This can be described as we did through (1)–(8) in the previous chapter. The only difference is that here a newly appeared gene will be taken into a use, i.e., the functional genome may become larger, whereas in the previous case it could only become smaller.
Thus, also here, we can see a mode of evolution which may occur without differential reproduction (i.e., without significant differences in the number of viable offspring between the different genotypes).
Still, a small nuance may be needed to explain in relation to the notion of natural selection. If a gene is distributing in a population in result of random walk, this is not natural selection, according to its definition. But if the distribution is accompanied by gene conversion, then the distribution can be shifted from random. However, a non-equal copying itself is usually not assumed for being a constituent of selection, it rather belongs to a special sort of mutation. Therefore, to be more precise in defining natural selection, we need to add that this is a significant difference in the viability of offspring between the groups of parents with a particular genotypic difference, except the difference which is resulted from non-identical replication.
Accordingly, also in the case of the growth of functional genome, we can see (at least hypothetical) possibility for a non-selectional adaptive evolution.
4. Further comments on adaptation
Thus, we reach the conclusion that the adaptive evolution (specialising, and may be also progressive) can occur without natural selection necessarily involved. This does not mean, of course, that natural selection is not working in evolution — there are many examples documented where it does. However, this analysis demonstrates that adaptive evolution is a more general process than the adaptive evolution via natural selection. Which means that Baldwin effect indicates to a possibility for a generalisation of theory of evolution.
It requires a special experimental work to find out in what extent the adaptive evolution in the wild is non-selectional. Theoretically, it allows much higher speed of adaptive evolutionary specialisation than the evolution restricted by the mechanism of differential reproduction.
Consider the following scenario. A population of organisms will move from one place to a new one, where the environment (e.g., food) will be different from the initial one. As a result, the organisms of this population will all need to use a part of their genomes which was not used before (i.e., expressing other genes), and will not use another part of the genome which was used before. If the individuals have an ability for individual adaptation, which will include a change in gene expression, then they can do it. Assume that the population will stay in its new place, and thus the change in the usage of genome will be preserved. Then, the stochastic mutations in the non-used region of the genome fix this change on the genetic level and thus make it irreversible, i.e., evolutionary.
Thus, the appearance of a new adaptation may occur during one generation as a response of the organisms' self-organisation and communication, and simultaneously for the whole population. The genetic fixation of this change will take, of course, many generations afterwards (or, in the case of gene duplications, it needed a number of generations before it for the distribution of the duplication). However, it is still enabling much quicker evolution than the case where the new adaptation first appears in a single organism (a mutate), and then via the competitive advantage distributes over the population.
This also indicates to a solution in the debate between punctualism and gradualism. The data which led to the formulation of punctualism came from the morphological studies of phylogeny, showing the periods of stasis and change. The evidence from the molecular data, vice versa, shows that the lineages change gradually. This is exactly what the mechanism as described above predicts — the morphological change and the genetic change may not be necessarily concurrent. E.g., the morphological change may initially be an ontogenetic adaptation which develops quickly, and it will be afterwards fixed in the genome gradually.
The main statement from which the conclusions given here follow is that for an organism, (a) there exist many different ways to behave and build itself in the case of the exactly same genome, and (b) there exist possibilities to behave in the same particular (constant) way for a quite large variety of genomes. In other words, (a) an organism has many ways to interpret its genome, and (b) there exist inheritance mechanisms in addition to the genetic one (e.g., epigenetic, or just the stability of the environment) which enable for an organism to keep some features of its structure and behaviour unchanged even if some changes in the genome take place.
Notably, the organism’s phenotype and genotype are largely uncoupled because (generally) an organism may not require its whole genome for living, and there exist potentially more functionally expressible parts in the genome than those which are currently in use. I.e., there are many ways to live using the same genome, as, for instance, there are many cell types in a multicellular organism which are using different parts of the same genome. However, this freedom appears due to behavioural activity (which includes perception and operation, i.e., the functional cycle, according to Uexküll (1928)), and particularly, due to mobility, so changing and selecting the environment. Thus, the uncoupling is a result of co-work of two levels of functional circles which cells (or phenomes) possess — one of these acting toward the genome, the other toward the environment.
A cell, having several ways to interpret its genotype (which means that the cell can change the expression of particular genes and, together with this, change the part of the genotype which is in use), also has several ways to preserve a particular interpretation over a number of generations (through epigenetic inheritance mechanisms, or due to a permanent change of environmental conditions). This gives to a stochastic genetic changes occurring in an unused part of the genotype a time to accumulate and to fix the otherwise only phenotypic changes. For instance, this means that if a particular cell type has not been formed during many generations, this may not be possible to form it later again, due to stochastic changes in the part of the genome used by (required for) this type of cells.
It is important to admit that a similar interpretational shift can take place simultaneously in many individuals of a population (e.g., as a result of invasion of the population into a new environment, for instance in the case of monophage or oligophage insects when they inhabit a new host species). The phenetic shift which this implies may be sufficient to decrease the efficiency of recognition of the source population specimens (which is needed for mating) down to the level which guarantees the sufficient isolation and provides time to the mutation processes to fix this separation also on the level of genome or cytoplasm incompatibility (Kull 1993, 1988, Paterson 1993).
Here, I would like to draw attention to an interesting paradox of natural selection, which can be called the paradox of unique child. That is, in the case of sexual reproduction, almost every descendant has the genotype which has never been present before (e.g., in the sense of a new combination of alleles of the whole genome), i.e., which survival has never been checked by natural selection. Nevertheless, the offspring usually includes (particularly in the species which have low reproduction rate) a great percentage of individuals who stay alive.
A solution of this paradox states that every organism has many ways to carry out the tasks they need to fulfil — the unique genetic memory may not destroy this ability, and if some structures happen to be corrupted, one can find several others in itself which still work, or which may be put into work. Little quantitative changes in an efficiency of one particular enzyme can often be compensated by small changes in the production of other enzymes, without real influence to the reproduction. Also, for the most part, it is not necessary for the living organism to be digitally precise.
For a new character to appear in the phylogeny of species, it is not necessary to assume that there had to be one specimen who gained this character first due to mutation, and then this mutation had to be spread over the species to all those whose grandparent that first mutant individual is. This would be required only if the digital preciseness in the determination of the character is assumed. As far as this is not the case, many different genetic changes in many individuals of the population may be simultaneously behind the same new character. And the genetic fixations of the new character (in the sense of making its appearance irreversible) could have been taken place after the first appearance of this character in the paleontological record.
Baldwin effect implies that evolution may take place without differential reproduction of genotypes. Assuming that the mutations in the expressed part of the genome cause the inviability of a certain (generally — about equal) percentage of offspring in all individuals of the population involved in the interpretational shift, we have a mechanism of evolution which works without the differential reproduction of genotypes. The neo-Darwinian mechanism is thus a special case of this mechanism, since it requires an additional assumption (e.g., that the percentage of inviable offspring is systematically different in different individuals, and this difference is correlated with a particular genetic character of parents).
The stochastic (entropic) changes in genotype preferentially lead to the forgetting of “unused” and the keeping of “used”. Due to the large size of genomes in terms of the number of genes, there are always many mutations which simultaneously distribute between the individuals via sexual reproduction and thus enable the ontogenetic change to become fixed for the whole population (making thus the phenotypic change irreversible); theoretically, this is much quicker than according to the classical mechanism which requires the distribution of new mutations across the population through competitive advantage.
The proposed ‘Baldwin effect’ mechanism is supported by recent studies in epigenetic inheritance mechanisms (Jablonka, Lamb 1995), and the stability of morphogenetic mechanisms (Webster, Goodwin 1996).
As a consequence of this mechanism, the activity of an organism as a subject may play a role as an evolutionary factor. This is exactly what the Baldwin effect originally claims. This also directs our attention toward the functioning and evolution of the mechanisms of interpretation (i.e., semiosis).
5. Conclusion: On the role of selection
It should be interesting to compare the behaviour of the described model with the classical neo-Darwinian or synthetic theory of evolution, and to investigate its applicability to the evolution of linguistic systems.
According to the neo-Darwinian explanation (which is accepted in the synthetic theory of evolution), the ability for learning (or plasticity) as a characteristic of a phenotype may appear and develop as a result of selective advantage of the corresponding genotype.
According to the semiotic view, and according to the explanation of Baldwin effect as described above, the results of ontogenetic learning themselves provide a factor and direction of evolution.
Jablonka et al. (1998) also see in the Baldwin effect an important aspect of interpretation of evolution. They demonstrate the existence of at least four different inheritance systems, all, according to them, having “the properties necessary for Darwinian evolution” (Jablonka et al. 1998: 209). However, they see it still in the framework of the Darwinian mechanism of natural selection, as a Lamarckism within the Darwinism.
In order to explain the relationship between the natural selection and the mechanism described above, we need a very clear definition of natural selection. Assuming that natural selection necessarily requires a statistically significant difference in the survival between the offspring of two genotypically distinguishable sets of organisms, we may conclude that, according to the Baldwin mechanism (as described above), the natural selection is not necessary for evolution (whereas the notion of evolution is still kept in its traditional definition as the irreversible change in the genetic structure of population). Because if the offspring of two genotypically distinguishable sets of organisms has the same percentage of non-viable organisms (e.g., if these sets are indistinguishable according to their mortality, or some other accepted measure of survival), there is no natural selection in this case, according to the definition accepted.
The synthetic theory of evolution shows that evolution may occur without natural selection, but in this case it is non-adaptive. We tried to show here that this can also be adaptive.
Natural selection may certainly be effective in the case when the selective factor (e.g., an antibiotic) is very strong, so that population number decreases down to few individuals who only survive. But if the population is permanently large, then natural selection cannot be very effective; however, the Baldwinian mechanism may still effectively work and consequently the evolution may go on.
Since the mechanism of Baldwin effect as described here is more general than the Darwinian mechanism of natural selection (as differential reproduction), and since the latter can be seen as a special case of the former, it seems to be accurate to use the term post-Darwinism as a name for this view (cf. Amundson 1998; Sermonti, Sibatani 1998; Kull 1999).
There certainly is selection in evolution. Most generally, the subject of selection is organism (cf. Weingarten 1993).
Amundson Ron (1998). Typology reconsidered: two doctrines on the history of evolutionary biology. Biology and Philosophy 13(2): 153–177.
Ancel Lauren W. (1999). A quantitative model of the Simpson-Baldwin effect. Journal of Theoretical Biology 196(2): 197–209.
Baldwin James Mark (1896). A new factor in evolution. American Naturalist 30: 441–451, 536–553.
Belew R. K., Mitchell M. (eds.) (1996). Adaptive Individuals in Evolving Populations: Models and Algorithms. Reading: Addison-Wesley.
Blitz David (1992). Emergent Evolution: Qualitative Novelty and the Levels of Reality. Dordrecht: Kluwer Academic Publishers.
Bowler Peter J. (1992). The Eclipse of Darwinism: Anti-Darwinian Evolution Theories in the Decades around 1900. Baltimore: The John Hopkins University Press.
Deacon Terrence (1997). The Symbolic Species. New York: Norton.
Emmeche Claus (1994). The computational notion of life. Theoria — Segunda Epoca 9(21): 1–30.
Harvey Inman (1996). Is there another new factor in evolution? Evolutionary Computation 4(3): 313–329.
Hinton G. E., Nowlan S. J. (1987). How learning can guide evolution. Complex Systems 1: 495–502.
Jablonka Eva, Lamb Marion J. (1995). Epigenetic inheritance and evolution: the Lamarckian dimension. Oxford: Oxford University Press.
Jablonka Eva, Lamb Marion, Eytan Avital (1998). ‘Lamarckian’ mechanisms in darwinian evolution. Trends in Ecology and Evolution 13(5): 206–210.
Kull Kalevi (1988). The origin of species: a new view. In: Kull K., Tiivel T. (eds.). Lectures in Theoretical Biology. Tallinn: Valgus: 73–77.
Kull Kalevi (1993). Recognition concept of species and a mechanism of speciation. Folia Baeriana 6: 133–140.
Kull Kalevi (1998). Baerian biology: evolution by means of organisms' interpretation. In: Farré George L., Oksala Tarkko (eds.). Emergence, Complexity, Hierarchy, Organization. Espoo: 197–200.
Kull Kalevi (1999). Outlines for a post-Darwinian biology. Folia Baeriana 7: 129–142.
Morgan Conwy Lloyd (1896). On modification and variation. Science 4: 733–740.
Morgan Conwy Lloyd (1898). The philosophy of evolution. Monist 8: 481-501.
Odling-Smee F. J. (1994). Niche construction, evolution and culture. In: Ingold Tim (ed.). Companion Encyclopedia of Anthropology. London: Routledge, 162–196.
Osborn H. F. (1896). A mode of evolution requiring neither natural selection nor the inheritance of acquired characteristics. Trans. New York Acad. Sci. 15: 141–142, 148.
Paterson Hugh E. H. (1993). Evolution and the Recognition Concept of Species. Baltimore: John Hopkins University Press.
Richards Robert J. (1987). Darwin and the Emergence of Evolutionary Theories of Mind and Behavior. University of Chicago Press.
Robinson Beren W., Dukas Reuven (1999). The influence of phenotypic modifications on evolution: The Baldwin effect and modern perspectives. Oikos 85(3): 582–589.
Searle John R. (1993). Intentionality: An Essay in the Philosophy of Mind. Cambridge: Cambridge University Press.
Sermonti G., Sibatani A. (eds.) (1998). Osaka Group for the study of dynamic structures. Rivista di Biologia (Biology Forum) 91: 125–158.
Simpson George Gaylord (1953). The Baldwin effect. Evolution 7: 110–117.
Stacey K. A. (1994). Recombination. In: Kendrew John, Lawrence Eleanor (eds.). The Encyclopedia of Molecular Biology. Oxford: Blackwell Science, 945–950.
Turney Peter (1996). How to shift bias: Lessons from the Baldwin effect. Evolutionary Computation 4(3): 271–295.
Turney Peter, Whitley Darrell, Anderson Russell (1996). Evolution, learning, and instinct: 100 years of the Baldwin effect. Evolutionary Computation 4(3): iv–viii.
Uexküll Jakob von (1928). Theoretische Biologie. 2te gänzlich neu bearbeitete Auflage. Berlin: Verlag von Julius Springer.
Uexküll Jakob von (1940). Bedeutungslehre. Leipzig: Verlag von J.A.Barth.
Waddington C. H. (1953a). Genetic assimilation of an acquired character. Evolution 7: 118–126.
Waddington C. H. (1953b). The “Baldwin effect”, “genetic assimilation” and “homeostasis”. Evolution 7: 386–387.
Webster Gerry, Goodwin Brian (1996). Form and Transformation: Generative and Relational Principles in Biology. Cambridge: Cambridge Univ. Press.
Weingarten Michael (1993). Organismen — Objekte oder Subjekte der Evolution? Philosophische Studien zum Paradigmawechsel in der Evolutionsbiologie. Darmstadt: Wissenschaftliche Buchgesellschaft.
A longer version of this paper was presented at the Bennington International Conference on Mind and Brain: An Integration of Evolutionary Origins and Emerging Developmental Principles, in November 10, 1999, Bennington, Vermont, USA. Correspondence: K. Kull, Department of Semiotics, University of Tartu, Tiigi Str. 78, Tartu, Estonia; or Institute of Zoology and Botany, Riia Str. 181, Tartu, Estonia. E-mail email@example.com.
I thank Mart Viikmaa, Sören Brier, and Terrence Deacon for helpful comments to an earlier version of this paper.