R1.1. Developmental plasticity and the constraints on adaptation

Bridgeman gives a beautiful example of developmental plasticity that was revealed during studies made in his laboratory. He and his colleagues found that unusual visual experiences in kittens can construct new and previously unseen receptive fields in the visual cortex. This example is comparable to that of the two-legged goat described by Slijper in 1942 and discussed in detail in West-Eberhard's important book, Developmental Plasticity and Evolution (West-Eberhard Reference West-Eberhard2003). This now famous goat was born without forelegs, adopted a semi-upright posture, and moved bipedally on its hind legs. After its death in an accident, extensive postmortem examination revealed many coordinated changes in the goat's morphology, such as changed hind leg and pelvic structure, curved spine, unusually thick and large neck, and various functionally correlated changes in its skeleton and musculature. Like the examples of plasticity discussed by Bridgeman, this case shows that pre-existing genetic and developmental capacities allow physiological adaptations that would not have been foreseen and could never have been directly selected for in the past. We can safely assume that walking on their hind legs did not regularly re-occur in the evolutionary history of goats, and that being visually exposed to vertical and horizontal stripes from birth was not part of the evolutionary experiences of cats. The reactions to these conditions of life, which are beyond the normal ontogenetic and evolutionary range of experiences, highlight the way organisms can make phenotypic adjustments to a new set of inputs by reorganizing their anatomy and physiology. Another case, which is of relevance to the evolution of language and which we described in E4D (Ch. 9), is the remarkable ability of the bonobo Kanzi to comprehend basic spoken English. The unusual rearing environment in which Kanzi developed led to new communicative behaviors and, we assume, to a novel reorganization of brain activity. West-Eberhard (Reference West-Eberhard2003) gives numerous other examples of adaptive plasticity at the morphological, physiological, and behavioral levels. She believes that new plastic responses to changed conditions can be the basis of evolutionary of novelties.

The limits of plasticity, therefore, are not always as narrow as is sometimes assumed. In their commentary, Dickins & Dickins claim that the degree of freedom for plastic calibration of the phenotype is very constrained. Constraints certainly do exist, but the type of evidence just outlined suggests that plasticity is surprisingly extensive. Moreover, many plastic responses at different levels can be inherited. Although it is true, as Faulkes & Baines note, that not all the “rules” of epigenetics have been unraveled, this does not mean that we can ignore the influence that induced changes may have on evolution. Not only are plastic responses and epigenetic inheritance the basis of cellular differentiation, but plastic responses and the transmission of induced changes also have a role in lineages of both non-behaving organisms, such as plants, and behaving organisms, such as animals and humans.

We consciously decided not to discuss the evolution of the nervous system and neural plasticity in E4D, because we wanted to focus on the trans-generational transmission of behavior. However, we recognize, as Foss stresses in his commentary, that when discussing animal and human evolution, there is a need to consider the evolution of the nervous systems and neural structures that enable meaning to emerge. In E4D we merely looked at the mechanisms and implications of between-generation transmission of the phenotypic products of developmental (non-neural as well as neural) plasticity mechanisms. We return to the problems of the evolution of neural structures underlying meaning in section R3.1.

R1.2. Descriptive terms: Lamarckism, neo-Darwinism, genes, and memes

Bridgeman describes the type of plasticity that he has studied as “acquiring characteristics from the environment, guided by inheritance,” a phrase that aptly describes the necessary interplay of both inherited and environmental information in development. However, we cannot support his suggestion that this should be thought of as “one-generation Lamarckism,” although we understand the temptation to do so. “Lamarckism” has always been a broad term: Nearly a century ago Delage and Goldsmith wrote, “Lamarckism . . . is not so much a system as a point of view, an attitude towards the main biological questions” (Delage & Goldsmith Reference Delage, Goldsmith and Tridon1912, p. 244). Nevertheless, Lamarckism does describe a view of evolution that usually incorporates the inheritance of acquired (induced or learnt) characters, not just the inheritance of the mechanisms enabling characters to be acquired. For this reason we would not like to expand it in the way that Bridgeman suggests. In E4D we used it in its conventional sense to highlight the evolutionary significance of the inheritance of developmentally induced and regulated variations.

Our use of the term Lamarckism, and our claim that neo-Darwinism, as developed in the Modern Synthesis, is no longer an adequate framework for evolutionary biology, drew understandable fire from Dickins & Dickins, Faulkes & Baines, and Brace. Dickins & Dickins claim that because non-genetic inheritance systems, which they see as mechanisms for calibrating genetic responses, have been designed by natural selection, neo-Darwinism, with its focus on genetic adaptation, can retain its privileged position in evolutionary theory. We certainly agree that there is no alternative to natural selection as an explanation of complex adaptations. However, what we tried to do in E4D was not to challenge the explanation of adaptation by natural selection, but rather, to extend the scope of natural selection, and offer a more inclusive framework for thinking about the origin of selectable heritable variation. We maintain, first, that there are more foci for evolution by natural selection than gene-centered neo-Darwinism assumes, because evolution by natural selection can occur through the selection of heritable variations that are independent of variation in DNA base sequence; and, second, that adaptation and diversity, which are both central to the study of evolution, will be better understood if the processes that generate heritable variation are studied. It is true that incorporating the developmental processes that underlie induction and learning into the evolutionary framework may seem to undermine natural selection, because these factors may, in some conditions, be more powerful than natural selection and lead to the spread of neutral or even deleterious variations. However, it is much more likely that developmentally induced heritable variations will be beneficial, because they are modulations brought about by an already adaptive regulatory system. Furthermore, just as incorporating drift into evolutionary theory does not lessen the importance of natural selection, so also the biases introduced by development and learning do not generally diminish the importance of selection. In fact, as we stressed repeatedly in E4D, they tend to extend its applicability.

Brace seems to disagree with our criticism of Maynard Smith and Szathmáry (Reference Maynard Smith and Szathmáry1995) for generally ignoring the role of developmental variations as direct agents of evolution, and also with our criticism of Mayr's (Reference Mayr1982) sweeping claim that there is no place for “soft inheritance” (Lamarckism) in evolutionary thinking. We know that the stance Brace takes is still the majority position, but we do not understand why he adopts it, because he gives no arguments against our point of view other than the argument from authority. Much of our book is a detailed attempt to show that, contrary to the majority position, soft inheritance is of importance in both heredity and evolution. In Chapter 9 of E4D we gave several examples which explain why we think that incorporating epigenetic inheritance would have enriched Maynard Smith and Szathmáry's explanations of evolutionary events. We have developed these ideas further in an article commemorating the enormous contribution that Maynard Smith made to evolutionary thinking (Jablonka & Lamb Reference Jablonka and Lamb2006b).

Although he is much more sympathetic towards our way of seeing evolution, Mesoudi queries why we stick to problematical terms like “gene” and “Lamarckism” yet reject the useful term “meme.” The answer to the gene/meme part of his question is that there is a profound difference in conceptual clarity between memes and genes. Even though the term “gene” is very difficult to define and, like every term that has been around for a long time, carries a lot of historical baggage, we do know a few things about genes (Keller Reference Keller2000). We know how DNA, the structural basis of genes, is replicated; we know that genes are usually located on chromosomes and that they mutate, are repaired, recombine, and so on. It is useful to have a term that relates to an entity with these properties, however problematic the precise functional definition of that entity may be. What memes are meant to be is far from clear, as even Dawkins has recently acknowledged, although he seems to think this is unimportant (Dawkins Reference Dawkins2006, p. 192); and it is not obvious to us what, except a lot of metaphorical baggage, they carry. As for “Lamarckism,” we stick to the term for the same reason that we stick to the term “Darwinism”: both Darwinism and Lamarckism are historically based umbrella concepts, and their historical baggage is part of their nature and their value (see Jablonka & Lamb Reference Jablonka and Lamb2007).

Lappan & Choe criticize us for our failure to provide definitions of key terms such as evolution, inheritance, and information. In fact, we did define information (E4D, p. 54), but it is true that we did not formally define evolution and heredity. We recognize this oversight, and included formal definitions of these and other terms in the Précis.

R1.3. The generality and scope of heritable non-genetic changes

Some commentators, including Faulkes & Baines and Lappan & Choe, question the generality of the non-genetic inheritance systems, assuming or claiming that we give them all equal weight. In reply we can only repeat what we have said many times: of course the importance of the four inheritance systems is not the same in all organisms. What is true for a human being is not necessarily true for an elephant, and what is true for the elephant is not necessarily true for the bacterium. Bacteria cannot read BBS, but many human beings can and do, so surely the evolution of culture, and of the mechanisms of information transmission that underlie it, have to be part of any discussion of why this is so. We certainly do claim that all four inheritance systems need to be incorporated into evolutionary thinking, but obviously only when all four are applicable! At the very end of the “Interim Summary” chapter of E4D, in which we compare the different systems of inheritance, we wrote, “Clearly, the different dimensions of heredity and evolution have different significances in different groups, and equally clearly, they all interact” (p. 238). In Chapter 9, where we discuss the evolutionary origin of the different systems of inheritance, we recognize that they have appeared at different times in evolutionary history and so cannot have the same significance in all groups. We tried to capture this idea in Figure 9.6, the legend to which reads, “The evolution mountain: moving up the mountain, there are new types of inheritance systems underlying new types of organism” (p. 341). Obviously behavioral and cultural inheritance cannot apply to non-neural organisms, but we believe that behavioral transmission is important in the evolution of social animals, especially birds and mammals. Fully blown symbol-based evolution seems specific to humans, but surely its uniqueness does not diminish its importance in the past and future evolution of human populations? We also argued in E4D (pp. 106–107) that developmentally induced changes in germline DNA (“directed” mutations) have probably not been very significant in multicellular organisms, because the constraints imposed by multicellularity would make even highly targeted mutations have functionally “random effects,” but that does not mean that developmentally induced changes in DNA are not significant in the evolution of microorganisms.

Some commentators questioned not only the generality, but also the scope and the importance of the different types of heritable variation we describe. How common and significant is epigenetic inheritance? What is the importance of behavioral inheritance in birds and mammals? How important are developmentally regulated changes in DNA?

Scope and importance are, of course, not synonyms. Some processes may be rare but nevertheless very important (e.g., single classical mutations with large effects). However, as far as we can gather, epigenetic inheritance is ubiquitous; its extent and range are becoming increasingly evident. Because both in their origin and in their transmission the properties of epigenetic variants are very different from those of genetic variants (see tables on pp. 234 and 236 of E4D), we believe that it is inappropriate to lump them together as “cellular inheritance systems,” which is how Faulkes & Baines suggest they should be treated. It is true, as Faulkes & Baines note, that there are several different epigenetic inheritance systems, although their claim that “epigenetics” and “epigenetic inheritance” originated as catch-all terms (a view that they mistakenly attribute to Morange Reference Morange2002) is not true. The historical origins and changing uses of these terms have been well documented and discussed in, for example, Van Speybroeck et al. (Reference Van Speybroeck, Van de Vijver and De Waale2002) and Jablonka and Lamb (Reference Jablonka and Lamb2007). However, we agree with Faulkes & Baines that it is unlikely that the various types of epigenetic inheritance will all have had the same significance in evolution: Their scope and importance may be very different in different groups. What the growing knowledge about epigenetic inheritance tells us is that we have to take more care in studying heredity, and include (or rule out) the different epigenetic inheritance mechanisms on a case by case basis. This does not diminish the fundamental significance of variants transmitted through epigenetic inheritance; it just makes their study, at the present time, rather difficult and labor consuming.

The evidence that was summarized at book length by Avital and Jablonka (Reference Avital and Jablonka2000) suggests that what in E4D we called “behavioral inheritance” is very common among social animals. We therefore dispute Faulkes & Baines's claim that the scope of behavioral inheritance is very limited, although it is true that certain types of cultural inheritance, for instance, those based on sophisticated cognitive abilities, are rare. They are as rare as these cognitive abilities themselves are (Whiten & van Schaik Reference Whiten and van Schaik2007). However, we do not see this as a good reason for grouping behavioral transmission with symbolic transmission and calling them both “brain-based inheritance,” as Faulkes & Baines suggest. As our summary tables (pp. 234 and 236 of E4D) show, they have different properties. It is true, as Lappan & Choe insist, that transmitting symbolic information must involve behavior. However, different symbol-based information can be transmitted through the same behavior, just as different heritable epialleles can be carried by identical DNA sequences. This is why we can distinguish between behavioral and symbolic variation, and between genetic and epigenetic variation.

In E4D we did not attempt to review the already vast and rapidly growing literature on non-genetic inheritance, and we cannot attempt to do so here. We recognize the importance of evidence, of course, and give some additional examples and references in the Précis of our book. One of us (Jablonka) is currently preparing a review in which she intends to summarize the evidence about epigenetic inheritance, and from the information already collated it is clear that the number of well-studied examples has reached triple figures. All we can do here is point to some more articles that give an idea of the range of non-genetic heredity phenomena to which we have referred. Information about the transmission of cellular epigenetic variations, particularly through the chromatin marking and RNA-mediated systems, is abundant: extensive reviews of cellular epigenetic inheritance in plants and its significance for understanding ecological and evolutionary processes are presented, for example, in Grant-Downton and Dickinson (Reference Grant-Downton and Dickinson2005; Reference Grant-Downton and Dickinson2006), Rapp and Wendel (Reference Rapp and Wendel2005), Takeda and Paszkowski (Reference Takeda and Paszkowski2006), and Zilberman and Henikoff (Reference Zilberman and Henikoff2005). Recent examples of epigenetic inheritance in animals are to be found in Chan et al. (Reference Chan, Yuen, Kong, Chan, Chan, Ng, Tsui, Lo, Tam, Li and Leung2006), Chong and Whitelaw (Reference Chong and Whitelaw2004), Rassoulzadegan et al. (Reference Rassoulzadegan, Grandjean, Gounon, Vincent, Gillot and Cuzin2006), Richards (Reference Richards2006), and Vastenhouw et al. (Reference Vastenhouw, Brunschwig, Okihara, Müller, Tijsterman and Plasterk2006). There is now evidence that in microorganisms such as yeast and Podospora, prion variants are transmitted through meiosis (Tuite & Cox Reference Tuite and Cox2006; Wickner et al. Reference Wickner, Edskes, Ross, Pierce, Baxa, Brachmann and Shewmaker2004), and examples of transmitted self-sustaining loops have been discovered and studied in microorganisms (Casadesús & Low Reference Casadesús and Low2006; Rando & Verstrepan Reference Rando and Verstrepen2007; Zordan et al. Reference Zordan, Galgoczy and Johnson2006).

Avital and Jablonka (Reference Avital and Jablonka2000) have given many examples showing that information is transmitted through maternal substances that effect embryonic development, as well as through the effects of milk, feces, saliva, and various forms of maternal and paternal behavior on the newborn. A fascinating case of behaviorally mediated inheritance in the rat has now been described in molecular detail (Meaney Reference Meaney2001; Weaver et al. Reference Weaver, Cervoni, Champagne, D'Alessio, Sharma, Seckl, Dymov, Szyf and Meaney2004). The importance of the effects of transmitted substances and early behavior on human embryonic and postnatal development is increasingly well recognized (Gluckman & Hanson Reference Gluckman and Hanson2005; Gluckman et al. Reference Gluckman, Hanson and Beedle2007).

Taken together, the data suggest that one would be hard pressed to think of a biological feature that cannot be influenced by some type of developmentally responsive inheritance mechanism, be it epigenetic, behavioral, or symbolic. As Faulkes & Baines rightly claim, these modes of inheritance have not yet been incorporated into mainstream evolutionary thinking, but this is not surprising. Ever since the beginning of the 20th century, when genetics began to be unraveled, evolutionary thinking seems to have lagged 15 to 20 years behind the findings from experimental studies of heredity. The same seems to be happening today, although as we describe in section R4, evolutionary biologists are beginning to incorporate information about non-genetic inheritance into their models and experimental studies.

Another issue which worries some commentators is the status of developmentally induced and regulated DNA changes that are transmitted from one generation to the next. Lappan & Choe and Dickins & Dickins raise doubts about the importance of such mutations in evolution. This type of mutation does seem to be of particular importance in unicellular organisms, and, as we have already noted here, in E4D (pp. 106–107) we argued that they are likely to be less significant in multicellular organisms. However, we may well have been too negative about this, because recent studies of stressed plants – plants subject to internal stresses resulting from hybridization or polyploidization, or to external stresses involving nutrition or infection – have shown that stress can sometimes induce targeted and reproducible changes in the genome. (See articles in Biological Journal of the Linnaean Society 82(4), 2004.) It begins to look as if there may be complex evolved mechanisms that, in severely adverse conditions, alter the gross structure of the genome in a targeted manner. As yet, these genome-altering mechanisms are poorly understood, but we believe that they are probably adaptive. Note that we think the mechanisms are adaptive – we are not saying that the mutations they generate are necessarily of adaptive value: The mechanisms increase the probability that a mutation or genomic change that is of adaptive value will occur. The situation may be comparable to the evolved mechanisms of targeted mutation in the immune-system genes; these mechanisms that produce mutations are adaptive, although most of the specific mutations they generate are non-adaptive. We follow Darlington (Reference Darlington1939) in thinking that the way the genetic systems have evolved has been an important factor in evolutionary history, and add our voice to those of McClintock (Reference McClintock1984) and others who have suggested that organisms may have evolved the capacity to make genomic changes in response to conditions of extreme stress. We return to this point in section R4.2, where we consider the predictive value of the four-dimensional approach that we advocate.

R3.1. Symbols and the problem of meaning

We defined symbols as a special type of signs, and defined signs as pieces of information transmitted from sender to receiver. The signs-that-are-symbols have a double nature: They refer to entities, processes, or relations in the world, and they also refer to and evoke other signs in the symbolic system, in a way that is bound by more or less binding rules, specific to the symbolic system in question (E4D, pp. 194–201). This definition of symbols focuses on communication between individuals and on the intersubjective aspect of meaning. People agree on a set of meanings, or definitions of a situation, and that is what makes communication among them possible. The ability to agree (implicitly or explicitly) is based on people sharing common cognitive properties and common-sense, so that fairly complex meanings (such as the meanings being conveyed in this exchange in BBS) can be reconstructed and understood by people who have never met each other. Clearly, a lot is left out of such intersubjective exchange: the idiosyncratically subjective, for example, as well the private aspects of emotion and feelings, which are very difficult to communicate even through facial and body gestures. Intersubjective communications through symbols come only through abstracting from private personal experience. The symbol, and especially the linguistic symbol, is far more discrete and digital than the analog, rich, and fuzzy reality and the private experiences to which it refers (Dor & Jablonka Reference Dor and Jablonka2000). Nevertheless, the very limitations of the symbol are also its strength: Communication through symbols allows effective communication of (partial) meanings, and sometimes, albeit rarely, as through great poetry, it even enables the communicators to transcend the inevitable loneliness of private experiencing. Of course, once symbols exist, they too assume a private aspect: we engage in internal monologues (communication with oneself); symbols become objects of deep emotions; and symbolic communication influences our most basic behaviors.

Our notion of symbols and symbolic language is very different from that of Sloman & Chappell. For us there are no “internal symbolic structures,” although naturally there are neurological processes and structures that underlie the ability to communicate through symbols. We regard planning, imagining, predicting, identifying informational gaps, and so on, all of which are characteristics of intelligent mammals, as complex cognitive abilities that are necessary but not sufficient conditions for symbolic communication and representation. We believe that conventionalization, which requires a degree of abstraction from the private, is a necessary condition too. When thinking about natural human language, we incorporate into its conception a set of constraints and affordances at both the neural and the cultural levels. We strongly agree with Sloman & Chappell's view that it is necessary to focus on development, and on the likely involvement of a complex hierarchy of developmental stages. However their “g-language” is, for us, a precondition for language, which is, as they argue, apparent in nonhuman species and young children. The complex cognitive processes and behaviors that animals and young children display are not, therefore, “symbolic” according to our definition of the term. In addition, although we agree with Sloman & Chappell that these cognitive capacities had to evolve before human language emerged, we think that symbols and symbolic human language evolved in the context of communication, because the external aspect of the linguistic symbol, the intersubjective communication that requires private experiences to be pared down, is a precondition for symbolic language. It was this experiential poverty that made the symbol into such a phenomenally powerful tool for communication (of certain things), and at the same time, such an inadequate tool for the expression of the rich and holistic states of many private experiences.

This leads us to Foss's comments about the evasive nature of meaning, and his claim that “there cannot be a science of symbolic evolution” because symbols have meanings, and meanings are analog rather than digital. There are no discrete meanings but, rather, shades of meaning, Foss claims; meanings cannot be measured, and there cannot be units of meaning that can undergo evolution. Foss is highlighting the deeply subjective aspect of communication, and he criticizes us for not addressing this in E4D. This is indeed something that we avoided in E4D, and the reason we did so was that our focus was on heredity and evolution. However, Foss is right in his assertion that meaning is crucial for understanding cultural, symbol-based human evolution, because symbols, with their double, private–public faces, are the strong attractors of culture. They construct shared public values which are imbued with strong private meanings, and are crucial drivers of cultural evolution.

Can meaning be defined and measured? We think that meanings can be defined and are amenable to evolutionary study. We agree with Foss that there is no precise measure for meaning. However, there is also currently no measure for function: The function of a DNA sequence is difficult to measure because it is both context and content dependent. (For an important first step in defining the fitness value of information, see Bergstrom & Lachmann [Reference Bergstrom and Lachmann2004].) The difficulty in measuring function does not, however, preclude the scientific discussion of the function and function-dependent evolution of genes, proteins, and other biological entities. From this point of view, the cultural act is no different. However, because the cultural act has not only function but also meaning, and meanings can drive cultural evolution, a discussion of meaning is certainly desirable.

We believe that an evolutionary account of meaning is possible, but that it requires a discussion of the ability of animals to experience the world rather than merely to process sensory information and respond to stimuli as plants do. In other words, it requires an evolutionary explanation of the nature of feelings and the reward systems underlying them. Ginsburg and Jablonka (in press) propose that something can be said to have “meaning” when the animal experiences (feels/thinks) it and has a system that assigns a value to it. Value systems are organismal neural structures and processes that “reward” (positively reinforce) any state that enhances a fitness-promoting action, and “punish” (negatively reinforce) any state that leads to actions that decrease fitness. Sensory-cognitive states, such as balance (versus imbalance) in primitive metazoans and pleasure (versus pain) in more evolved animals, are the attractor-states towards which the animal strives, and are usually associated with behaviors that promote fitness. In humans with symbolic communication, ideas of the good (versus the bad), truth (versus untruth), beauty (versus ugliness), justice (versus injustice), and so on, are the culturally constructed attractors towards which we strive. Because these symbolically encoded attractors are based on publicly shared and constructed culture, they transcend the individual and can be (and often are) decoupled from inclusive fitness considerations. In other words, the symbolic system can “subvert” biological fitness more readily than the goal-directed ontogenetic reward systems of animals, because symbolic values may have little to do with the here and now. This means that cultural, symbol-based evolution has to be studied in ways that incorporate this potential decoupling. However, we do not see why this complication precludes an evolutionary approach to culture.

R3.2. How should we approach cultural evolution?

The similarities and differences between genetic and cultural transmission, and the role of genetic assimilation in evolutionary processes, are topics addressed by several commentators. Sometimes we felt that their comparisons were based on a very sophisticated notion of culture and a rather unsophisticated view of genetics. When, as we now know, a gene can code for several different protein products, some genes code only for RNA, two genes can use overlapping DNA sequences, selectable phenotypes are the products of complex developmental networks, and we have no idea what role, if any, most DNA plays in the cell, we think it is no longer appropriate to reduce the genetic dimension of heredity and evolution to the replication and selection of DNA sequences called “genes”.

Gabora calls attention to the differences between gene replication and the type of transmission involved in cultural evolution. She has pointed out (Gabora Reference Gabora2004) that behavioral and cultural evolution involve uncoded, interpreted information, rather than the coded, uninterpreted information of the genetic system of inheritance. In her commentary, Gabora claims that, “The mathematical framework of natural selection is not transferable to evolutionary processes that are not code-driven,” and therefore that behavioral and cultural evolution, just like the origin and early evolution of life, is primarily non-Darwinian. They should be seen as Lamarckian processes. According to Gabora, for a process to be Darwinian, “inheritance of acquired characteristics must be negligible compared to change resulting from differential replication of individuals with heritable variation competing for scarce resources.”

There are elements of Gabora's thesis with which we are in total agreement. Like her, we believe that the nature of the information – in our terms, whether it is modular or holistic – influences how evolutionary change occurs, and that both Lamarckian and Darwinian processes occur in evolution. We also agree that the fidelity of reconstruction (a word that we prefer to replication) is important. Where we disagree is with Gabora's fundamental assumptions that natural selection “is intimately tied to the genetic code,” and that when induced (or “acquired”) variations are transmitted, evolutionary change has to be regarded as non-Darwinian. We see no justification for these assumptions, and know of no reason why natural selection – a Darwinian process – and induced heritable variation – a Lamarckian process – should not both be involved in an evolutionary change. We think that by using the gene-based replicator as a conceptual framework, Gabora has fallen into the trap of assuming that if an evolutionary change cannot be explained in terms similar to classical gene-based explanations of adaptation, Darwinian selection is not involved. One of the reasons why we find Griesemer's (Reference Griesemer2000) reproducer concept so valuable is that it avoids this pitfall. As we suggested in E4D, the significance of induced variation and Lamarckian processes probably varies widely; they are obviously important in cultural evolution, but that does not mean that Darwinian selection is not occurring.

While Foss maintains that culture cannot be said to evolve, and Gabora claims that cultural evolution is purely Lamarckian, Mesoudi sees merit not only in the concept of cultural evolution, but also in present-day models of it. He takes us to task for dismissing an impressive body of theoretical work on cultural evolution, and he is right to do so. We should have discussed the models of cultural evolution in greater detail, although in self-defense we have to point out that we did direct interested readers to the relevant sources (see notes to p. 205 on p. 402 of the book). Nevertheless, we still think that there is a need for rather different models of cultural evolution – models in which induction/learning and selection are coupled, and in which there are strong cultural “attractors.” However, this is a challenge for the future. Models in population genetics have generally ignored development, because developmental induction is less central to the genetic inheritance system than it is to cultural transmission. We think that classical population genetics, which was very important for the development of evolutionary biology because it demonstrated theoretically that small selection coefficients and chance mutations could bring about “descent with modification,” is of limited predictive value now that genes are seen as part of developmental networks. The very simplistic assumptions of classical population genetics are no longer thought to be appropriate, and different mathematical models are being constructed to accommodate the growing recognition that the effects of gene interactions complicate selection theory. Similarly, we think that a new family of models is necessary for dealing with cultural evolution. They will need to incorporate multilevel exploration, learning, and selective stabilization processes based on genetically and culturally evolved mechanisms and attractors. These models will recognize that although the selective retention of some cultural entities may be a result of a benefit accrued to an individual or a group, it may also result from the strength of symbolic, publicly shared conventions, which serve as attractors and can override individual or group benefits.

Unlike Faulkes & Baines, Mesoudi finds the analogy between cultural and genetic evolution useful, and suggests that cultural creativity and genetic mechanisms of evolvability may be analogous processes. We think that this is a very interesting suggestion. Evolvability has recently become a hot topic in evolutionary biology, and the heuristics of cultural creativity may inspire interesting new directions in the study of the poorly understood genetic “heuristics.” We completely agree with Mesoudi that “mental travel time” is a crucial and unique feature of human culture with no parallel in gene-based evolution, and we believe it to be a result of the symbolic nature of human language (see, e.g., E4D, p. 203).

Whereas most commentators have focused on comparisons of gene-based and cultural evolution, or the interactions of the genetic and cultural systems, Warlaumont & Dale stress the importance of the relationships between behavioral and symbolic inheritance, and they criticize our neglect of this topic. All that we can say about Warlaumont & Dale's comments is that they were right to criticize us, and we appreciate the excellent examples they have provided to show how the development of symbolic linguistic communication and representation is deeply interrelated with behavioral transmission. Although we were aware that behavioral transmission underlies symbol-based information transmission, and that symbolic communication and thought affect behavior, in our effort to show the autonomy of symbol-based transmission we dealt with the interactions between the behavioral and the symbolic in a non-systematic manner. We think that this is a regrettable omission for exactly the reasons stated by Warlaumont & Dale. Moreover, had we discussed the interrelations of the two brain-based inheritance systems more systematically, some of the apparent weaknesses in our position, to which the commentaries of Faulkes & Baines and Lappan & Choe point, would not have arisen.

Lappan & Choe argue that, “If symbolic evolution is completely decoupled from the symbol-producing organisms' fitness and behavior, and transmission is independent of reproduction, then the analogy to biological evolution is an interesting thought exercise, but it would be difficult to support the argument that symbolic variation represents a fourth distinct dimension of biological evolution.” A similar point is made by Faulkes & Baines, who regard brain-based and cell-based inheritance as being so disparate that a comparison between them has no value. Our view is that it is the dynamics of historical, intergenerational change that unites the different systems of transmission. Obviously the systems are very different, and we stressed these differences in Parts I and II of E4D; in the Interim Discussion we also pointed to what they had in common. We wrote the book believing that a focus on information – on its storage, its transmission, and its development – could unify and illuminate aspects of evolution and lead to a useful research program. In the last part of the book, we showed that the systems are intimately related, and that evolution on one (non-genetic) axis can profoundly affect the genetic axis. The case of language evolution discussed in Chapter 8 illustrates how the genetic, behavioral, and symbolic systems are intertwined: language evolves on a symbolic-cultural axis; the dynamically changing, yet consistent, linguistic niche affects the selection of behavioral and hence genetic variants; and the selected behavioral and genetic changes then affect the evolution of the capacity for language. We suggested how genetic assimilation, and its more sophisticated manifestations such as “the assimilate-stretch principle,” can link the constructed and inherited cultural (linguistic) niche with the genetic underpinnings of the language capacity and lead to a sophistication of linguistic behavior over generations. Symbolic evolution in this case (and in many others) is related to fitness, although on certain time-scales the dynamics of symbol-based evolution have to be considered autonomously. This very autonomy, however, contributes to the “stretch” phase in the process, which leads to genetic assimilation.

The role of genetic assimilation, which seems to Gabora to be over-emphasized in our account of language evolution, and which Steels implies is under-emphasized, is an issue about which we were very careful. We do not know how much genetic assimilation was involved in the evolution of language, although for the reasons we gave in E4D (pp. 308–10, 316) we have little doubt that it was involved.

R3.3. Memes and replicators: Why object?

Blackmore advocates a meme-oriented approach to the evolution of culture, and takes issue with us over our less-than-enthusiastic treatment of memes. For her, a meme is “that which is imitated,” and she suggests that it is imitation, rather than symbolic representation and communication, that is the basis of human uniqueness. She finds our notion of symbols incoherent, because it is not clear to her what an abstract picture or a new dress symbolizes, and she is also unhappy with the way we use the term “memeplex.” Blackmore sees an increase in the fidelity of copying as an important part of both genetic and cultural evolution.

As we pointed out in section R3.1, our notion of “symbol” is not dependent on the iconicity of the symbol, but rather, depends on the duality of its reference – to things in the world and to other symbols – and on it being a component of a rule-governed symbolic system. We agree with Blackmore that songs and dances are not inherently symbolic. Indeed, in E4D (p. 202), we described how the same type of behavioral entity (e.g., a song) can be transmitted in different ways: A song's development can depend mainly on genetic resources (as with the courtship song of male Drosophila), or on socially learnt resources (as in some song birds), or, as in humans, on symbolic resources. An abstract painting is a symbolic object par excellence, because it can be properly interpreted only within the context of modern (symbolic) art. A new dress fashion has a symbolic aspect because it implies and refers to status, or to social class, or to a public icon, and because it points to and evokes a host of related and interacting behaviors and artifacts. Tunes, cars, hairstyles, and buildings are also part of a symbolic system, although they are not “pure” symbols like words or numbers.

We agree with Blackmore that symbolic thought is a product of cultural (and genetic) evolution rather that its starting point, and in Chapter 5 of E4D we described non-symbolic cultural evolution in animals. Cultural evolution does not depend on imitation, although when imitation exists the range of cultural variants may be greatly expanded. We also agree with Blackmore that the development of high-fidelity copying is very important both in genomic and in cultural-symbolic evolution. However, we do not see how her focus on imitation avoids the fundamental problems to which we pointed in E4D. Most human (non-machine) imitation involves active exploration and learning, governed by understanding and by perceived goals (E4D, pp. 210–12; Jablonka Reference Jablonka2004b). Steels rightly points out that models of cultural evolution based on imitation assume what they try to explain, because a person has to know what counts as information and what the intention of the producer is before they imitate. Perhaps his explanation will help Faulkes & Baines to understand why we distance ourselves from memes. Both imitative and non-imitative socially learnt behaviors are developmentally reconstructed, not copied in a machine-like or gene-like way. One of the examples of memes that Dawkins gave in the Extended Phenotype (Dawkins Reference Dawkins1982, p. 109) was the socially learnt skill of opening milk bottles by tits, so we used it in E4D to show how, in this case, the meme (a habit) is transmitted through a process of developmental reconstruction, not imitation. But, even with cases of copied behavior that do involve imitation (e.g., engagement in a religious ritual), it can still be readily shown that developmental reconstruction and active, environment-sensitive learning are involved (E4D, pp. 210–12; see also Avital & Jablonka Reference Avital and Jablonka2000, pp. 25–29).

As we argued in E4D, we think it is the notion of the meme as a replicator that is fundamentally flawed. In his latest book, The God Delusion, Dawkins describes memes as replicators, and defines a replicator as “a piece of coded information that makes exact copies of itself, along with occasional inexact copies or ‘mutations’” (Dawkins Reference Dawkins2006, p. 191). Blackmore claims that the distinction between replicator and vehicle is not crucial for her notion of the meme, yet she still finds the meme concept that emerged from this distinction productive. The difficulty we have is that we cannot see how what is referred to as a “meme” can be anything other than a heritably varying phenotypic trait, which is developmentally constructed and reconstructed through social learning (be it by imitation, by emulation, or by any other type of socially mediated process). If “meme” had been used in this way – as a term for socially transmissible behavioral phenotypes, preferences, or products of behavior – we would have seen it as a useful piece of shorthand. However, the wish to see the meme as a replicator akin to a gene, rather than as a character, leads to conceptual complications which render the concept incoherent. The term “memeplex” also seems to us to be full of problems. It is defined by Dawkins (Reference Dawkins2006, p. 198) as “a set of memes which, while not necessarily being good survivors on their own, are good survivors in the presence of other members of the memeplex.” His memeplex therefore refers to the functioning of a set of memes, not to their transmission; it is an analog of an adapted gene complex. It is not a unit that is copied, which is the meaning that Blackmore gives to the term in her commentary. However, it is the conceptual problems with the meme, rather than those with the related terms used in memetics, that in our view prevent memes and replicators from being fruitful tools for understanding cultural evolution.

The problems associated with the replicator concept are also reflected in Gabora's and Steels's commentaries. Gabora identifies Darwinism with the replicator-centered interpretation of it that Dawkins and others have constructed, and is therefore driven to a position that denies that natural selection is of relevance to models of cultural evolution, a position that we think is untenable. Steels's point of view is different. He recognizes the value of the reproducer concept (Griesemer Reference Griesemer2000) and suggests that, when considering language evolution, a useful distinction can be made between replicators and reproducers, with utterances being replicators and the language system as a whole being seen as a reproducer – an entity that develops to reproduce and is a target of selection. This is an interesting suggestion, but we think that it is better to give up the replicator concept altogether, and think in terms of reproducers that have heritably varying traits, which are more or less sensitive to developmental regulation through induction and learning. The specification of the spectrum of developmental sensitivities, the genetic and/or cultural evolution of such a spectrum, and selection for increased or decreased modularity of heritably varying traits would then become well-defined research questions.

R4.1. Practical applications

Although much remains to be done, the roles of the additional inheritance systems we discussed in E4D are already being investigated and are affecting research programs. The most dramatic impact is that of the epigenetic dimension in medicine. It is becoming clear, for example, that in addition to the genetic and environmental factors that influence the incidence of disease, the probability that an individual will develop certain diseases is affected by the environments to which their ancestors were exposed, which have effects that carry over to their descendants (Jablonka Reference Jablonka2004a; Jirtle & Skinner Reference Jirtle and Skinner2007). In cancer research and developmental medicine, epigenetic inheritance is being intensely studied, and it is considered by leading authorities in the field to be of central significance for diagnosis and potential treatment. (For recent work and reviews, see Baylin & Ohm Reference Baylin and Ohm2006; Burdge et al. Reference Burdge, Slater-Jefferies, Torrens, Phillips, Hanson and Lillycrop2007; Chan et al. Reference Chan, Yuen, Kong, Chan, Chan, Ng, Tsui, Lo, Tam, Li and Leung2006; Gluckman et al. Reference Gluckman, Hanson and Beedle2007; Pogribny et al. Reference Pogribny, Tryndyak, Muskhelishvili, Rusyn and Ross2007; Rodenhiser & Mann Reference Rodenhiser and Mann2006; Yoo & Jones Reference Yoo and Jones2006.)

Understanding epigenetic inheritance is crucial for further progress in cloning (Jaenisch et al. Reference Jaenisch, Hochedlinger, Blelloch, Yamada, Baldwin and Eggan2004). Epigenetic effects in bacteria are also of medical importance, because cells that in normal growth conditions are killed by antibiotics can become “persistent.” Persistence, an epigenetic phenomenon, occurs when some of the bacterial cells in a population switch into an antibiotic-tolerant state, usually by decreasing their growth rate (or by abolishing growth altogether). Tolerant cells are genetically identical to sensitive cells, and cells can switch from sensitive to tolerant states and vice versa. These switches occur stochastically, but the rate of switching may be affected by environmental conditions (Kussell & Leibler Reference Kussell and Leibler2005; Lewis Reference Lewis2007). Because this system has been partially unraveled, drugs that can counter the persistence of the pathological microorganisms are being developed. There is also increasing theoretical work being done on how such epigenetic systems may operate, and how they have evolved. This is reflected in the modeling that has been carried out (Jablonka et al. Reference Jablonka, Oborny, Molnár, Kisdi, Hofbauer and Czárán1995; Kussell et al. Reference Kussell, Kishony, Balaban and Leibler2005; Lachmann & Jablonka Reference Lachmann and Jablonka1996; Pál Reference Pál1998; Walczak et al. Reference Walczak, Onuchic and Wolynes2005).

It is not only in medicine that the recognition of epigenetic inheritance is having an effect. In agriculture epigenetic inheritance is known to be fundamental to constructing strains of genetically modified plants and animals that do not silence newly introduced, potentially beneficial genes, and is also important for carrying out risk assessments of such new strains. Ecologists, especially plant ecologists, are now beginning to study the frequency of epigenetic variants (e.g., DNA methylation marks) within plant populations (Ashikawa Reference Ashikawa2001; Cervera et al. Reference Cervera, Ruiz-García and Martínez-Zapater2002; Knox & Ellis Reference Knox and Ellis2001; Liu & Wendel Reference Liu and Wendel2003; Riddle & Richards Reference Riddle and Richards2002; Wang et al. Reference Wang, Lin, Dong, Wang and Liu2004). Although simple methods for detecting epiallelic differences other than DNA methylation have yet to be developed, many ecologists are clearly aware of the importance of epigenetic variations, as well as genetic variations, for understanding biological diversity, conservation, and adaptation.

Epigenetic inheritance is a very young field of study, so there are still many basic questions that have not yet been addressed, although they are amenable to experimental investigation. In E4D and elsewhere, we have pointed to several of them. For example, does the number of generations organisms spend in a particular inducing environment have cumulative effect on the stability of marks (Jablonka & Lamb Reference Jablonka and Lamb1995)? What is the epigenetic basis of Lansing (cumulative parental age) effects (Jablonka & Lamb Reference Jablonka and Lamb1990, Reference Jablonka and Lamb1995; Lamb Reference Lamb1994)? Are there cumulative effects on epigenetic marks when chromosomes are transmitted through one sex for several generations (Jablonka Reference Jablonka2004c)?

There is also much basic work that needs to be done on the transmission of information by behavioral means in nonhuman animals. In their book Animal Traditions, which discusses the formation and evolution of cultural traditions in nonhuman animals, Avital and Jablonka (Reference Avital and Jablonka2000) described many simple and interesting experiments that would shed light on the evolution of culture. Their suggestions include: First, studying how coprophagy (feces-eating) affects food preferences in mammals by cross-feeding young with feces from adults kept on different diets. (Such studies might also reveal whether feces are a route for the transmission of disease-causing prions. It seems to us possible that the prion diseases in wild deer populations in the United States could be transmitted by feces-eating.) Second, studying the transmission through social learning of self-medication in animals by experimentally altering medicinal practices or their objects. Third, investigating the possibility that not only material and energy, but also information (e.g., about food and other ecological preferences) is transmitted during courtship feeding. This could be done by studying the behavior of females with courtship-feeding males that have different ecological preferences. Fourth, in bird species with learnt duets, investigate by modeling and through observation the possible role of nontransferable information in the long-term stabilization of the pair bond. Fifth, in species with helpers, where the helper's feeding or caring habits might be transmitted to the cared-for young, find out whether, and to what extent, future adult behavior is affected by looking at individuals that are cared for by helpers with different (natural or experimentally manipulated) habits.

R4.2. Evolutionary implications

The view that non-genetic inheritance is important has many implications for evolutionary theory, which we have discussed in E4D and later publications (Jablonka & Lamb Reference Jablonka, Lamb, Fox and Wolf2006a; Reference Jablonka and Lamb2007). They include:

1. 1. Adaptation can occur through the selection of heritable epialleles. This may be particularly important if populations are small and lack variability. Because epigenetic variants may be induced when environmental conditions change, and many individuals in the population may acquire similar modifications at the same time, adaptation through the inheritance of newly induced epigenetic variants may be very rapid (Kussell & Leibler Reference Kussell and Leibler2005; Lachmann & Jablonka Reference Lachmann and Jablonka1996). Although epialleles may not be as stable as genetic alleles, adaptations based on such variation may enable a population to survive long enough for a process of genetic accommodation to occur (E4D, Ch. 7; Pál Reference Pál1998; Sangster et al. Reference Sangster, Lindquist and Queitsch2004; Siegal & Bergman Reference Siegal, Bergman, Fox and Wolf2006).

2. 2. Chromosomal evolution may be affected by epigenetic variations. Jablonka (Reference Jablonka2004c) suggested how the selection of inherited epialleles can influence the evolution of sex chromosomes in mammals, and Rodin et al. (Reference Rodin, Parkhomchuk and Riggs2005) discussed how heritable epigenetic silencing can affect the fate of duplications.

3. 3. Epigenetic variations can affect the generation of genetic changes by biasing the rates of point mutation, transposition, recombination, and other genomic reorganization processes (E4D, Ch. 7; Jablonka & Lamb Reference Jablonka and Lamb1995).

4. 4. The epigenetic and genetic responses to stress may lead to very rapid evolution by reorganizing the genome. Genomic stresses such as hybridization and polyploidization are known to induce massive genetic and epigenetic reorganization in plants (Rapp & Wendel Reference Rapp and Wendel2005), and possibly also in animals, and ecological stresses involving nutrition have been shown to induce significant changes in repeated sequences in plants (Cullis Reference Cullis2005), probably via epigenetic mechanisms.

5. 5. Heritable epigenetic and behavioral variations may initiate the population divergence that leads to speciation. Reproductive isolation may begin when non-genetic behavioral differences between individuals from different populations prevent mating taking place (E4D, Ch. 5; Avital & Jablonka Reference Avital and Jablonka2000), or because their chromosomes carry incompatible chromatin marks and, as a consequence, the viability or fertility of hybrid offspring is reduced (E4D, Ch. 7; Jablonka & Lamb Reference Jablonka and Lamb1995).

6. 6. When learning from and through others is important, as it is for many social birds and mammals, group selection may play a greater role in evolution than is generally assumed (Avital & Jablonka Reference Avital and Jablonka2000).

7. 7. In birds and mammals, the necessity of exchanging information suggests alternative or complementary interpretations of the evolution of relationships between mates, and between parents and offspring. Avital and Jablonka (Reference Avital and Jablonka2000) believe cooperation, rather than conflict, may have driven or helped to drive the evolution of some of the observed behaviors. They also suggest that the preferential expression of maternal (imprinted) genes in the cortex of mammals during postnatal development may reflect the outcome of selection for compatibility between mother and offspring, rather than being the result of mother–offspring conflict, because the young have to be maximally attuned to the mother during the critical period of early learning. Hager and Wolf (Reference Hager and Wolf2006) believe that the need for compatibility between mother and offspring can explain many patterns of genomic imprinting. They suggest that during mammalian development in utero, selection for co-adapted maternal and offspring traits may drive the evolution of genomic imprinting, and they have supported this suggestion with a simple model.

8. 8. Because cellular epigenetic inheritance acts as a powerful constraint on the evolution of development, discussions of problems such as the origins of multicellularity, irreversible cellular differentiation, the segregation of germlines, and so on need to include the epigenetic dimension (E4D, Ch. 7; Jablonka & Lamb Reference Jablonka and Lamb1995).

9. 9. Zuckerkandl and Cavalli (Reference Zuckerkandl and Cavalli2007) have suggested that so-called “junk DNA,” which consists largely of repeated sequences, might be a carrier of epigenetic marks that can be communicated to and established in other regions in the genome. An altered mark could therefore result in coordinated hereditary changes in the expression of several different genes simultaneously, and thus accelerate adaptive evolution.

10. 10. Non-genetic mechanisms of inheritance have played a central role in all the major evolutionary transitions identified by Maynard Smith and Szathmáry (Reference Maynard Smith and Szathmáry1995). We outlined how they did so in E4D (Ch. 9) and later discussed this subject in greater detail (Jablonka & Lamb Reference Jablonka and Lamb2006b). In the latter article, we also discuss one of the transitions that Maynard Smith and Szathmáry did not identify, although it certainly must rank as “major,” namely, the transition from non-neural animals to animals with a nervous system.

11. 11. Because non-genetic mechanisms of information transfer play such key roles in evolution, the evolution of the non-genetic inheritance systems themselves is of fundamental interest.

R4.3. The case of the silver foxes

We end this section by discussing the study by Belyaev and his colleagues of the domestication of silver foxes, which Faulkes & Baines think we have misinterpreted. Belyaev (Reference Belyaev1979) called the domestication of animals over the past thousands of years one of the greatest experiments in biology, stressing the amount of very rapid evolutionary change that had taken place. He used his studies of the domestication of foxes as a framework for discussing the nature of these changes, and came to some interesting conclusions.

Selection for tame behavior in silver foxes began at the end of the 1950s and is still going on (Trut et al. Reference Trut, Plyusnina and Oskina2004). It resulted in a very interesting complex of heritable changes: In addition to becoming doglike in their behavior, the foxes showed various other modifications such as changes in the skeleton, decreased sexual dimorphism, hormonal changes, spotting, altered tail and ear posture, and an increase in the incidence of small, supernumerary chromosomes. Belyaev and his group studied the pattern of inheritance of spotting, an easy character to assess, and found that it was strange: Spotting behaved like a dominant or semi-dominant trait, but it had a high rate of reversion. The rate of appearance and disappearance of the character was far too high for mutation to be likely, and it could not be explained as an effect of inbreeding, because the coefficient of inbreeding was only 0.03 (Trut et al. Reference Trut, Plyusnina and Oskina2004).

Overall, the high rate of morphological and hormonal change that took place as the foxes were selected for tameness, and the parallels with the changes that occurred during the evolution of the dog, suggest that the genes selected during domestication probably had big effects on rates of development, particularly in the nervous system and associated developmental pathways. But what exactly was selected? Was it alleles, epialleles, or both?

Belyaev's suggestion was that the conditions of domestication led to neuro-hormonal changes that reactivated dormant genes in the soma and germline, thereby revealing previously hidden genetic variation. Not all genes can be reactivated in a way that is transmitted transgenerationally, but the selection for tameness picked out those allelic variants that could. This interpretation in terms of certain alleles being able to switch from dormant to active states explains the high rate of appearance and disappearance of the phenotypes. In the terms used today, we would say that Belyaev was probably studying epiallelic variants. Support for this interpretation comes from the fact that at least two of the genes (Agouti and C-kit) that Trut et al. (Reference Trut, Plyusnina and Oskina2004) think may be involved in the pathways leading to the change, are known to have heritable epigenetic variations in mice. Thus, if Belyaev is correct, selection for tameness involved changes in the frequency of both alleles and epialleles: Stably reactivated epialleles of reactivatable alleles were picked out during selection. We think that this interpretation is consistent with the observed patterns of inheritance. Now that the dog genome project is yielding more and more information, and the genes underlying morphological development as well as aggressiveness in foxes are being studied (Popova Reference Popova2006; Trut et al. Reference Trut, Plyusnina and Oskina2004), looking for the epiallelic variants of these genes is becoming a possibility. Belyaev's bold conjecture can now be tested directly.