Shyshkin, 1987. Individual development and evolutionary theory

Shishkin M.A. Individual Development and Evolutionary Theory // Evolution and Biocenotic Crises. Moscow: Nauka, 1987. - pp. 76–124. UDC 591.3:575.8 INDIVIDUAL DEVELOPMENT AND EVOLUTIONARY THEORY M. A. Shishkin Paleontological Institute of the USSR Academy of Sciences Darwin’s theory of natural selection takes as its object whole living organisms that arise through individual development and engage in the struggle for existence throughout their life cycle. Therefore, solving the key problem of this theory—the origin of adaptation—inevitably requires elucidating the laws of ontogenetic evolution, i.e., the laws governing the formation of those mechanisms that ensure a species’ reliable implementation of its phenotypic norm. From the standpoint of current knowledge one can see that many objections raised against classical Darwinism might have seemed substantial only because the questions of evolution of individual development remained undeveloped in that doctrine. A completely different foundation underlies the currently dominant genetic (“synthetic”) theory of evolution, which, in the words of I. I. Shmal’gausen (1968a, p. 20), signifies the replacement of Darwinism by genetics. The abstraction from ontogenetic problems lies in its very premises, since the material of selection is essentially thought of not as organisms but as hereditary factors or their combinations, i.e., structures preceding development; their fate is to determine the course of evolution. The question of how phenotypic realization of these factors influences the nature of selection is set aside, except for the consideration that their expression or non‑expression is taken into account and that the genotype‑environment is theoretically recognized as influencing these processes. This situation is well illustrated by the fact that until recently one could find works presenting the genetic theory in which the concept of ontogeny was not used at all (Sheppard, 1970). On the other hand, it is characteristic that the growing critical statements directed at this theory in recent years first emphasize the oversimplification and insufficiency of its premises for understanding the evolutionary role of developmental processes. Nevertheless, the history of Darwinism records two serious attempts to understand the relationship between natural selection and individual development—the theory of embryonic plasma by A. Weismann and the doctrine of stabilizing selection by I. I. Shmal’gausen and K. Waddington. Their proposed solutions are diametrically opposite. If in Weismann’s view evolutionary changes of ontogeny constitute only a passive result of selection on elements of embryonic plasma, then in Shmal’gausen’s view, conversely, the material for selective transformation of the genotype are ontogenetic aberrations. The genetic theory inherited from Weismann only his reductionist understanding of heredity—its dissection into independent factors (typical of 19th‑century heredity theories) and the replacement of whole organisms as the substrate of selection with the latter. The main unifying principle of the Weismannian concept, which made it a truly synthetic construction, remained unnoticed. It lay in the clear understanding that the mechanism of inheritance of organismal properties is expressed in the way they are realized ontogenetically, and therefore recognizing the transmission of these properties by independent carriers can only mean their independent realization in ontogeny. In other words, the idea of discrete hereditary determination inevitably implies mosaic (pre‑formed) development. Weismann built such a developmental model (based on the hypothesis of unequal inheritance division), and its logical failure led to the collapse of his entire theory. However, the conclusions that seemed to follow for understanding the mechanism of heredity were not drawn until the emergence of the theory of stabilizing selection. Concerning this latter theory, which will be analyzed further, the attitude of contemporary breeding is ambiguous. Although in Russian literature it is often characterized as the highest current achievement of Darwinism (Schwartz, 1969; Gall, 1980), still, as correctly noted by (Kirpichnikov, 1974), its acceptance is often purely verbal. The reasons for this are largely understandable. The theory of stabilizing selection, with its focus on explaining the evolution of a coherent organization, is successive to classical Darwinism (cf. Yablokov, 1981) but not to the genetic theory. Its relationship with the latter, as will be shown below, is in fact antagonistic. Therefore, for the dominant direction of modern breeding it is impossible to adopt the doctrine of stabilizing selection without revising its own foundations. If a satisfactory evolutionary theory must essentially be a theory of ontogenetic evolution and if Darwinian doctrine contains the possibility for this, then its fundamental concepts must be expressed in the language of individual development. This concerns heredity, variability, adaptation (as a process), as well as the mechanism of natural selection itself. The same analysis must be applied to the concepts of genetics used by modern evolutionists, if the phenomena they describe relate to phenotypic realization and its adaptation. The issue is to determine whether what is called alleles, homo‑ and heterozygotes, etc., are “pure” characteristics of the hereditary structure that constitutes the substrate of selection, or whether we are dealing with features of morphogenetic systems created by the evolutionary process itself. Before turning to these questions, it is necessary to highlight some of the most significant evolutionary aspects of individual development. Evolution is first and foremost a change of generations leading to a change of adaptive norms, and the transition to each new norm inevitably entails some alteration of the previous one. Consequently, in the evolution of populations and species there must exist generations dominated by stable individual cycles (forming the norm) and generations characterized by ontogenies with more uncertain outcomes. The regularities of transition from the second type to the first should be explained by the mechanism of adaptagenesis (as interpreted by the theory of stabilizing selection). Thus, in ontogeny we will mainly be interested in the ratio of stable versus unstable, or normal versus aberrant, developmental pathways. HEREDITY, VARIABILITY, ADAPTATION AND ORGANIZATION AS CHARACTERISTICS OF THE STABILITY OF INDIVIDUAL DEVELOPMENT A characteristic property of living beings is the stability of their typical morphophysiological organization, designated as the adaptive norm (Shmal’gausen, 1940a). This, in fact, makes possible typological research methods such as comparative morphology, systematics, and phylogenetics. Across generations this property of organisms also manifests as their capacity for self‑realization, i.e., as the stability of inheritance of the adaptive norm. These two features are sometimes cited as independent characteristics of the living (Kastler, 1967), but in reality they share the same basis—the stability of norm implementation during individual development. Regarding normal organization per se, this basis is self‑evident, because the adult state of the organism in question is itself a component and result of the ontogenetic cycle. The link between norm development and its inheritance is equally obvious, but some clarification is required. In Darwinian evolutionary theory heredity is defined as the transmission of phenotypic traits across generations (Darwin, 1951). This “transmission” of course means the reproduction of a trait anew during individual development, which typically combines parental gametes with the offspring’s phenotype; the reliability of this process’s result determines inheritance. Therefore, when speaking of “hereditary” and “acquired” traits, what is actually meant is the stability or lability of their ontogenetic realization relative to developmental conditions, and these terms are preferable because they convey the essence more clearly (Woodger, 1953; Shmal’gausen, 1969, 1982; Shishkin, 1981, 1984a, b, v). In the words of A. G. Gurwitsch (Gurwitsch, 1912; Gurwitsch, 1944), heredity is the process of carrying out typical development (cf. Meister, 1934; Dubinin, 1973). The term “heredity” has no other content concerning phenotypes (for which it was introduced). It is meaningless to label genetic determination of traits with this term, because all organismal properties are products of genotype‑environment interaction and thus are simultaneously “inherited” and “acquired” (Johannsen, 1926; de Beer, 1963; Kamshilov, 1972). If the concept of heredity is applied to genotypes and their elements, not only does its original meaning change, but it simply becomes a “thing‑in‑itself” that resists any definition other than a tautology (“heredity is the transmission of hereditary factors”). [The widespread claim that the object of inheritance is the individual “norm of reaction” (Johannsen, 1926; Dobzhansky, 1947) is equivalent to recognizing inheritance of individual genotypes. Yet genotypes are continuously transformed each generation by replication errors (in general) or by recombination (in sexual processes).] This important point must be noted, because the whole history of studying heredity bears the stamp of an apparently strange dualism: on one hand it is seen as a property of a certain category of traits, and on the other as something existing independently of them and belonging to the embryonic plasma (genotype). This contradiction disappears only in one case—if one assumes that the stable reproduction (inheritance) of traits is caused by the transmission through gametes of unequivocally corresponding discrete causal factors. Early genetics held this view, and this conception objectively remains the foundation of the evolutionary theory built upon it, despite the attempts of its authorities (e.g., Mayr, 1968, 1974) to deem it obsolete. The claim that organismal stability is created by a selective process is central to the theory of stabilizing selection. But if ontogenetic stability means heritability, then obviously the latter is also a product of selection. This conclusion is clearly formulated by the cited theory in terms of converting labile changes into hereditary ones, of creating hereditary mechanisms by selection, of replacing external developmental factors with internal ones (Shmal’gausen, 1940b, 1941, 1968b, 1982, pp. 109, 110, 161, 214), and, finally, of genetic assimilation of adaptive traits (Waddington, 1953, 1957). Between the statement that hereditary changes are created by selection and the usual notion that they arise by mutation lies an insurmountable philosophical chasm that should stop anyone attempting to reconcile Shmal’gausen‑Waddington’s doctrine with genetic theory. Consider the more familiar second statement. If it merely refers to genetic changes, then explaining them by mutations is tautological. If it refers to phenotypic changes that are stably preserved across generations, then it is plainly false. Experimental genetics shows that simple small mutations do not guarantee a stable effect (their expression is always more labile than the norm) and may even be silent. The preservation of phenotypic norm under continuous recombination of the genotype in xenogamous populations illustrates this well. Thus, heredity (stability) and genetic determination are different things (cf. Kamshilov, 1967). All of the above inevitably leads to a reassessment of another familiar postulate of genetic breeding concerning “hereditary variability” as material for natural selection. In fact it is also based on the identification of traits with their hereditary factors. It implies that elementary (non‑adaptive) phenotypic variations of individuals should be inherited if they are caused by genetic differences. Yet, as already noted, the mere presence of genetic heterogeneity in individuals says nothing about the character of its phenotypic expression. It may be completely unexpressed (under the cover of a normal phenotype, see: Chetverikov, 1926), or expressed inconsistently (as typical for mutations). The theory of stabilizing selection, arising from these facts and treating the very property of inheritance as an evolutionary product, naturally must regard as raw material for the latter the unstable traits, i.e., labile elementary reactions (morphoses) of individually differing genomes (Shmal’gausen, 1982; Shishkin, 1981, 1984a, b). It is precisely the ensemble of such reactions, operating on a heterogeneous basis, that corresponds to Darwin’s concept of indefinite variability, which allows speaking of changes that are not only inadequate to environmental factors but also unordered in their reproduction in the immediate offspring. The evolutionary origin of biological stability can be considered from another aspect. The very concept has many synonyms long used precisely to denote phenomena requiring evolutionary explanation. Among them are purposiveness, adaptation (Ashby, 1959, 1962; Shmal’gausen, 1968a, p. 139) and equilibrium with the environment (Spencer, 1899). All of these denote the property of individuals to respond to external disturbances in such a way as to preserve their normal viability, including successful reproduction. Adaptive organization is an organization capable of persisting (Wake et al., 1983). Historical survival of the most adapted means that selection preserves and creates increasingly stable types of organization capable of withstanding a maximally wide spectrum of disturbances. The broader and more diverse this spectrum, the greater the number of neutralizing response reactions required from the organism so that ultimately it can realize one of the permissible isomorphic normal states (law of necessary diversity; Ashby, 1959). These reactions must be coordinated, because system stability is impossible without interaction of its parts (Bertalanffy, 1969). Thus, the increase of adaptation (stability) under selection inevitably leads to greater complexity and integration of morphophysiological organization. This Darwinian principle, constantly contested from K. Negeli to modern evolutionists (Wright, 1964), logically follows from viewing the organism as a whole system; yet it becomes inexplicable as soon as we try to replace organisms as selection objects with a mosaic of their hereditary factors. All species, because they possess an adaptive norm, are equally adapted to their habitats (i.e., to their spectrum of permissible disturbances) and therefore are equivalent in the quality that can be called their relative stability. However, they can also be compared in terms of absolute stability, i.e., the degree of diversity of external factors whose effects they can relax. This indicator, as seen above, measures their organization, i.e., also measures progress. The more chaotic and unpredictable the fluctuations of the environmental factors used by a species, the higher the demands on the organism’s complexity, and conversely, the more homogeneous the environment, the lower the demands. Simple (predictable) fluctuations are, in particular, characteristic of highly ordered internal environments used by parasites, which explains their tendency toward degeneration. Since the physiological behavior of organisms is regulated toward the norm, it is directed against irreversible changes, which constitute the essence of evolution. A stable system (truly equilibrium, or quasi‑equilibrium, as living organisms and open systems are) while it remains so, by definition “does not remember” its fluctuations (modifications) and therefore does not evolve. The cause of evolution lies in the breach of stability (Spencer, 1899), i.e., in the departure of system dynamic variables beyond the limits that allow regulation of the whole. Restoration of stability at a new level (i.e., a new equilibrium with altered environment) occurs only through natural selection, which, for this reason, is strictly speaking always stabilizing. It represents a mechanism of supra‑organismic regulation of individual stability. Ideal stability, i.e., the ability to respond to any external or internal impact with a fluctuation, is of course unattainable for organisms, but the higher its absolute value, the less vulnerable they are to direct elimination (which yields to differential reproduction) and the greater their capacity to pre‑empt any objectively necessary elementary evolutionary change with a corresponding adaptive modification, i.e., to partially model it from the “available possibilities” of their morphogenetic system. This is the sense of G. Spencer’s (1899) statement that in the course of organic evolution natural selection yields to direct adaptation; although literally inaccurate, the underlying tendency is correctly understood. The same possibility of pre‑modeling underlies Morgan‑Bouldin’s idea of “organic selection,” justifying the evolutionary role of modifications (Shishkin, 1984b). Since the growth of absolute stability, or adaptation, is coupled with increasing organizational complexity, i.e., movement toward ever less probable states, evolution drives organisms further from thermodynamic equilibrium, which is possible only through ever higher levels of external energy consumption. Thus, the increase of organization (stability) is associated with higher energy costs, and the rate of entropy production is a key indicator of it (Goodwin, 1970).These costs are offset by the self‑evident advantages conferred by high fitness in the struggle for existence. STABILITY OF THE NORM AND PRINCIPLES OF THE THEORY OF NORMAL DEVELOPMENT The stability of development of a typical organization, ensuring its self‑maintenance (fitness) and self‑realization (heritability), as well as the connection of this phenomenon with the regulatory mechanisms of ontogeny, has long been noted by embryology. Already K. Baer (Vae​r, 1828) established by comparative methods a reduction of embryonic variability during development, showing that the property of the latter is a directedness toward a certain final state. For the mechanics of development, which arose at the end of the 19th century and set as its task the experimental identification and localization of the immediate operative factors of morphogenesis, this property turned out to be an insurmountable obstacle to attempts to construct a general theory of development on the basis of a causal‑analytical method (Blyakh­er et al., 1935). The result of the process resisted interpretation as a sum of effects of certain initial causes, demonstrating a significant independence from their variations. The stability of this result with respect to the ways of achieving it (Roux, 1895), demonstrated by many experiments, made it impossible to accept V. Ru’s mosaic non‑formist concept, which reduces development to independent differentiation of an initial set of primordia. The theoretical implications of developmental self‑regulation and the principled irreducibility of the latter to independent cause‑effect chains were recognized by G. Driesch (Driesch, 1908; Drisch, 1915), who described the embryo as a “harmonious equipotential system”, i.e., a complex of parts with identical possibilities (prospective potentials), governed as a whole in its transformations by an immaterial ordering factor—entelechy. This indivisible factor determines the actual fate (prospective value) of primordia according to their position in the whole and controls the coherence of their changes throughout development, including stages of self‑differentiation, when experimentally no inter‑dependencies are found among primordia. In this concept, of course, the specific “solution” of the problem (which is simply replaced by the symbol of entelechy) is less important than the clear understanding that development is an integral process whose properties are supra‑summative and stable with respect to its constituent elements. This marked the beginning of a systems approach to development, based on the Aristotelian principle that “the whole exists before the parts”. Its legitimacy had already been shown earlier in a purely empirical generalization, Baer’s law (Vae​r, 1828), which literally states the same (the general in development arises before the specific); however, only from Driesch onward was it used as a foundation for a theory of development. Modern doctrine of embryonic determination is built on it, although, unlike Driesch’s theory, it treats the factor of wholeness as material and knowable (Gurwitsch, 1910, 1912; Svetlov, 1964, 1978; Belousov, 1963). Thus, the search for a causal explanation of the robustness of normal development led embryology to regard this process as a hierarchical system (the whole and its parts) governed by its upper level, i.e., the properties of the whole. Later (essentially on the same basis) a historical explanation of this developmental property was obtained. We again refer to the theory of stabilizing selection (Schmalhausen, 1940b, 1941, 1968b, 1982), which relies on the notion of an adaptive norm. The latter (i.e., the normal phenotype) changes historically much more slowly than the way it is ontogenetically realized, which is continuously reshaped by selection toward greater reliability. Thus the capacity to vary the processes constituting development increases without compromising the stable implementation of the norm. The norm here acts as a factor of wholeness, governing (through selection) the alteration of its morphogenetic mechanisms and defining the permissible space of individual variation of their elements (Shishkin, 1981). The fundamental identity of embryological and historical explanations of stability appears self‑evident. Nevertheless, we are first interested in the embryological explanation, i.e., a scheme of cause‑effect relations suitable for describing a single developmental cycle. But first it is necessary to pause on the general theoretical requirements that such a description must satisfy. Stability of the result of normal development implies the purposiveness of this process. Both definitions characterize the same thing—the capacity for self‑regulation of the final state. Teleonomic behavior of a stable material system manifests in that, when displaced from equilibrium, it reacts in a way that ultimately returns it to that equilibrium. Accordingly, in physics and chemistry, finalistic formulations (Le Chatelier’s principle, etc.) are used to describe such processes. For closed systems, a stable state corresponds to thermodynamic equilibrium; in open systems, including living organisms, it lies at a distance from equilibrium and is characterized as “flow equilibrium” (Bertalanffy, 1949), “stable non‑equilibrium” (Bauer, 1935), or simply as a stationary state. Movement toward any type of equilibrium, or “goal‑seeking”, proceeds through closed feedback cycles, where, for example, element A, when perturbed, influences B such that B’s change corrects A’s state toward a value that reduces further correction. The system is “governed by its error” (Ashby, 1962), and its stability rests on the interaction of its elements. In open systems, correction cycles operate continuously; in closed systems, interaction amplitudes decay as entropy rises, establishing “stability with respect to a point” (Goodwin, 1970). The notion of purposive system behavior does not, of course, imply that events depend on future conditions. It merely reflects the fact that the final outcomes of elementary changes in a system are determined by the overall properties of the system itself and cannot be reduced to direct mechanical consequences of those changes. The system as a whole either does not respond to an elementary influence or shifts into one of its alternative states (modifications). In other words, teleonomic dependence is revealed when events or properties belonging to different hierarchical levels of the system are compared—specifically, when its slowly changing parameters (characterizing its holistic behavior) are juxtaposed with the rapidly varying values of its elements (dynamic variables). A finalistic description of such relationships reflects the principled impossibility of a causal description, because the properties of the whole are not reducible unambiguously to the states of its elements. “What appears as a stable structure of a certain level is in fact maintained by the continuous exchange of components of the nearest lower level” (Bertalanffy, 1969), i.e., the same property of the whole persists under different combinations of elementary interacting causes. Consequently, inter‑level relations are characterized by a sharp asymmetry of causes and effects (Belousov, Chernavskii, 1977), unlike processes amenable to causal description. [From this viewpoint, Driesch’s vitalistic concept of development appears as a characteristic reaction of a naturalist who notes the absence of the usual clear‑cut causal dependence between linked phenomena and sees no alternative description other than invoking immaterial factors]. Therefore one cannot agree with the widespread view that finalistic and causal formulations are merely two equally valid ways of describing changes in the same cyclic causal chains. Hierarchically equal elements of such cycles lack asymmetric relations among themselves; conversely, properties of different system levels are not linked by causal dependence. All of the above directly relates to understanding the mechanism of individual development. We conclude that its causal explanation is possible only if the entire chain of causal events leading to a holistic final result (normal organization) is presented as a sequence of equal (single‑level) holistic states. The presence during development of such states with stable characteristics is not only a theoretical requirement but also an experimentally established fact. Indeed, successive periods are observed that are characterized by internal wholeness (topological isomorphism) and directed transformations, with reduced sensitivity to experimental disturbances (Svetlov, 1960; Belousov, 1979). At the same time, individual development, like any other irreversible change, must pass through phases of stability loss (Bertalanffy, 1969; Volkenshtein, 1981b). These phases also appear in individual development as “sensitive periods”, characterized by labile determination and transformation of topological patterns. Consequently, a general theory of normal ontogeny must minimally include the following premises. 1. Development is a chain of mutually conditioning structurally holistic states. 2. Each of them, for the duration of its existence, determines the course and coordination of individual morphogenetic processes (i.e., acts as an “entelechy” in Driesch’s sense). 3. The realization of these processes each time results in a specific disruption of the whole’s stability and its subsequent restoration at a new level that controls further differentiation. 4. As embryonic organization becomes more complex during development, each new state of wholeness stabilizes at an increasingly greater distance from true equilibrium. Apparently, the only developmental concept compatible at its core with these premises is today the theory of the biological field, proposed by A. G. Gurwitsch (Gurwitsch, 1922; Gurwitsch, 1944). The notion of a physical field—that is, a space whose properties determine the behavior of particles within it—well matches the idea of the sought material factor of wholeness that controls the entire set of developmental processes. As P. G. Svetlov (1964) correctly noted, the field principle is already clearly expressed in Driesch’s concept, which indicated that the prospective value of an individual element in development is a function of its position in the whole. [This Driesch view entered an obvious contradiction with his definition of the wholeness factor as a “non‑extensive” quantity lacking spatial characteristics]. Gurwitsch’s theory, setting aside its later elaborations (1944) concerning the idea of a cellular field, allows one to approach the simple and general laws underlying the ontogenetic process. It is assumed that, starting from the egg cell, the embryo generates around itself an anisotropic vector field whose structure predetermines the developmental outcome at the nearest infinitesimal stage. After the field space is filled, the latter “outgrows itself” and reorganizes into a field with new parameters determined by the final state of the embryo achieved at the previous stage. This creates a setting for development on a new nearest segment, and so on. During the process, fields of individual primordia subordinate to the field of the whole also form. In this concept, the factor of wholeness (the field), continuously directing development, is at the same time a continuous function of the path traversed by the substrate of its influence. Here we have a “Driesch law in differential form” (Belousov, 1979), i.e., the directing whole is no longer viewed as a final goal but as a property of successive stages, transformed according to the laws of causality. [Partly this line of events was already recognized by Driesch (1915), who noted that the completion of each developmental stage leads to “a change of the next entelechy task”. In essence, this concerns a change in the properties of the wholeness factor itself]. Development appears as an avalanche‑like process with positive feedback between a primordium and its field, with the primordium’s purposive behavior toward each newly established field state. Although evidence for the theory mainly concerns particular morphogeneses and mostly addresses the spatial aspect of changes as the most accessible for study, it is very compelling. Numerous examples show (Gurwitsch, 1944) that the determination of an entire primordium occurs in an indeterminate state of its constituent elements. The latter are only statistically determined as a set (“normalized” according to Gurwitsch), subordinate to the whole’s field. Well‑known examples include the random spatial‑temporal distribution of individual mitoses relative to symmetry axes in primordia such as the developing onion root or the retinal tissue; the overall result of these random events remains ordered. The reality of the field is vividly demonstrated by the phenomenon of “dynamic preformation” of a primordium, when the orientation of its wall cells is determined not by the actual configuration of the wall but by a force surface outside it, corresponding to its prospective outlines acquired at the next immediate stage. Although constructing a general theory of ontogeny is a task for the future, it undoubtedly cannot proceed without the systemic principles laid by Gurwitsch. Even the simplest considerations support this. 1. A developing organism is an integral dynamic system that excludes a single unambiguous state of the processes constituting it or of the initial elements underlying them. 2. At any moment of development there is no other wholeness than that inherent to the embryo at that stage. Hence the idea of a transformable whole becomes inevitable. MULTIPLICITY OF INITIAL STATES AS THE BASIS OF NORM STABILITY In a stable system, the preservation of its parameters is based on the continuous change of its interacting parts. Thus, regulation of these parameters at each disturbance—and consequently the process of attaining them from an initial non‑equilibrium state, which constitutes the teleonomic model of individual development—occurs equally effectively for a multitude of possible initial values of the system’s elements. This independence of the final properties of a dynamic system from its initial state is called equifinality in the broad sense of the term (Bertalanffy, 1969). [Equifinality was understood by Driesch (1915) as the ability of individuals of one species to regenerate the whole organism by different pathways after the same experimental injury, i.e., a phenomenon akin to what comparative embryology calls “circumstantial development” (similarity of the start and end of typical development in different forms despite divergent intermediate paths). In the sense of independence of the process outcome from the initial state, the concept was used by L. Bertalanffy (Bertalanffy, 1949, 1969), who considered such behavior a property only of open systems. This is true insofar as, for an open system, any parameter can theoretically be maintained, whereas in a closed system external disturbance leads to definite irreversible changes. In practice, for open systems the capacity for regulation does not extend to all their properties each time; for example, development of sea‑urchins from isolated blastomeres or regeneration of ascidians from operated individuals yields organisms that are normal but reduced in size (Driesch, 1915). Conversely, equifinality of closed systems with respect to maximal entropy is always absolute; however, for our purposes it is more convenient to call it the principle of multiplicity of initial states. In organismal development this regularity has many obvious manifestations; especially striking are cases where a species can reproduce by different modes (e.g., sexual and vegetative in ascidians, Cnidaria, etc.), with initial developmental stages having nothing in common, yet the result being identical. The same includes all experimental facts on ontogenetic regulation, in particular cases of whole‑organism regeneration from fragments of specialized tissues (flatworms, nemertines, etc.) or self‑assembly of the embryo (e.g., sea‑urchin gastrulae) from isolated cells followed by restoration of normal development (Svetlov, 1964, 1972). In general, not only the realization of the whole organism but any morphogenetic act in normal development rests on the well‑known independence of its result from initial conditions. For example, normal induction proceeds despite considerable fluctuations in the mass and timing of interaction of members of the inductive system, variations in reactor sensitivity, concentration and composition of activating substances, etc. (Schmalhausen, 1964, 1982). In other words, all development is built on relatively stable acts with “multiple provision”, understood as any—not only qualitative—differences in the execution of the same morphogenesis. Within the overall developmental system these relatively stable events constitute an intermediate hierarchical level (or levels) between the unequivocally determined whole (the adult norm) and the statistically conditioned elementary processes that form the basis of the system, such as cell divisions in “regulatory” ontogeneses. This is the principle of Gurwitsch’s normalization: determination and formation of an individual primordium are not tied to a rigid fixation of the initial state of its elements. The principle of multiplicity of initial states has far‑reaching consequences for the theory of normal development, already touching the domain historically separated as the theory of inheritance. Moving deep into development up to the zygote, we must conclude that the stability of the adult norm’s implementation (i.e., the)The heritability of this trait cannot be reduced to a fixed set of states of any cellular units, including the chromosomal apparatus (genome) of the zygote or oocyte. This set must be determined only statistically (normalized), i.e., it must retain uncertainty within limits that allow normal (equifinal) completion of development. This is precisely what the theory of stabilizing selection asserts. According to Schmalhausen (1982, pp. 84, 174), the stability of organization is not a property of the elements of the hereditary substance, but an expression of the interaction of the parts involved in development. The organism is “more stable than its genotype” (if the latter is understood as a set of specific states of chromosomal units). This is one of the most important conclusions of the theory, based on the synthesis of empirical data, and at the same time a deduction from the premise that system properties are irreducible to the properties of its elements. The adaptive norm is determined only by the holistic species‑specific structure of the germ cell, which corresponds to a multitude of interchangeable genome variants capable of realizing this norm under typical developmental conditions. For this set of variants the concept of a “genotypic norm” can be introduced. What facts support the considered conclusion? First of all, the genetic heterogeneity (“mutation load”) of natural amphimictic populations established by S. S. Cheterikov (1926), hidden beneath the adaptive norm and revealed by inbreeding. Continuous redistribution of elements of individual genomes during recombination across generations does not change the outcome of their phenotypic realization in the majority of cases. However, the principle of multiplicity of initial states contains no prohibitions that would limit the variability of initial developmental factors within the resolving capacity of hybrid‑genetic (Mendelian) analysis. We may assume that even in communities that appear homogeneous at this level (in pure lines of self‑pollinators and clonal populations), hidden genetic diversity always actually exists. Numerous experiments with plants (cereals, legumes) and parthenogenetically reproducing insects (aphids) that were grown under extreme conditions after strict selection of autogamous varieties convince us of this. In these experiments a diversity of individual physiological responses is always found, and the most viable variants become fixed by selection, accompanied by the emergence of stable morphological features (Samokhvalova, 1951, 1954; Shaposhnikov, 1961, 1966; Agaev, 1978). Although these experiments are interpreted differently, it is quite obvious that they concern genetic variability that acquires phenotypic expression because environmental conditions exceed the limits that permit normal (equifinal) development. [These facts contradict the common view that selection is impossible in pure lines, based on the classic experiments of Johannsen (1933) with beans. The doctrine of genetic assimilation of unstable traits now allows a different interpretation of those experiments. For a labile reaction to become fixed by selection, hidden heterogeneity of its carriers is insufficient. It is also necessary: a) that the carriers express this reaction under the same deviating conditions, and b) that selection initially operates under those same conditions. Both requirements were not met in Johannsen’s experiments, and the individual causes of identical grain‑size modifications were not controlled. A comparable experiment would be an attempt to fix a thermomorphic trait in Drosophila by using its expressions obtained under opposite temperature deviations, and selecting among the immediate offspring raised under normal conditions. The results of these experiments do not differ fundamentally from those obtained by genetic assimilation (stabilization) of individual structural and physiological morphs in xenogamic organisms such as Drosophila (Kamschilov, 1941; Waddington, 1957). From this one can conclude that, regardless of the mode of reproduction, the adaptive norm is indeed realized on the basis of multiple permissible states of the germ‑cell genome (or vegetative embryo).] This degeneracy of correspondence between variations of the genetic basis and the result of normal development constitutes, as already noted, one of the key principles of the theory of stabilizing selection. According to this view, the stability of the norm is expressed by the creation, through selection, of a regulatory epigenetic mechanism capable of buffering (canalizing) variations of genetic factors and environmental conditions over a wide range. Thus, in forming the adaptive norm, selection inevitably must increase the permissible genetic variability underlying it, as is observed in reality. The outcome of normal development is not reduced to a fixed sum of initial causes. On the contrary, for a genetic theory of evolution that attempts to describe phenotypes in terms of specific genes, the logically expected result of selection would be the creation of maximal homogeneity in populations for all selected alleles or their combinations. The mismatch between this expectation and the actual genetic structure of the norm forces recourse to various additional hypotheses (about the balance of selection and mutation, antagonistic or frequency‑dependent selection, etc.), the abundance of which (Kojima, 1971; Solbrig, Solbrig, 1982; Ayala, 1981) testifies to the intractability of the arising difficulties. EPIGENETIC SYSTEM AS AN EXPRESSION OF THE GENOTYPE‑PHENOTYPE CONNECTION The adaptive norm does not exhaust the developmental possibilities of an individual germ cell. Around the normal phenotype lies a region of diverse unstable deviations (morphs) that arise when developmental conditions are disturbed. The set of such reactions, realized on the basis of a heterogeneous pool of zygotes (capable under normal conditions of regular morphogenesis), forms a “mobilization reserve” of the population, i.e., hidden variability that, according to Schmalhausen (1941, 1968b), constitutes the potential material for evolutionary transformations (Shishkin, 1981, 1984a, b). Yet alongside these deviations there appears, at first glance, another source—mutational variations of the germ‑cell genome that disrupt the development of the adaptive norm even under conditions normal for it. This category of aberrations is regarded by most evolutionists as the material of natural selection. The question arises: are we really dealing with two distinct categories of deviations, and how do they relate to normal development? What commonality exists in the functioning of “normal” and “mutant” genomes that allows us to view them as variants of a single species‑specific genotype? If individual mutational lesions do not push development beyond the species‑specific anomaly domain (Dubinin, 1966b, p. 240; Myr, 1968), then, with even greater certainty, the same applies to disturbances caused by external factors. It follows that the constraints are the same in both cases. Indeed, the parallelism of phenotypic expression of mutational and modification changes observed both in nature and experimentally, as well as the parallelism between hereditary (stable) and modification traits in closely related races and species, is a well‑known fact that underlies a number of historically linked yet profoundly different evolutionary theories (neo‑Lamarckian concepts, Morgan‑Bouldin’s idea of “organic selection,” and the theory of stabilizing selection). This commonality permits us to assume that the entire set of possible ontogenetic pathways inherent to a species is a manifestation of the stable properties of the holistic developmental system, and that the evolutionary process leads to changes in the structure of this whole system. If each species indeed possesses a limited repertoire of ontogenetic implementation variants, then it is evident that it constitutes a specific “space of possibilities” characterizing the behavior of that system. For a single zygote this space of definitive states can be represented on a plane divided into discrete fields, and its realization pathways form a bundle of diverging trajectories, of which, in a typical case, only one (with its terminal branches) corresponds to the adaptive norm, understood as a more or less narrowly bounded region of the field (Fig. 1a,b). For all germ cells of the species this space is identical (equifinal), and individual differences among them (principally their genomes) lie only in the relative probabilities of following different developmental trajectories under given external conditions. All genome variants for which the normal phenotype constitutes the most probable (stable) developmental outcome (Fig. 1b, n1–n3) can be identified as the genotypic norm; the remaining variants are what are usually called mutations, although in fact all genomes in a population are mutants relative to each other. Any disturbance of the developmental system that does not destroy it (i.e., allows development to be completed) can only alter the choice of a particular trajectory, but cannot produce a result that lies outside the species‑specific space of possibilities. In other words, the system’s response to a perturbation will ultimately shed its own specificity regardless of whether an external influence affected the developmental course or the structure of the germ cell itself was altered. [IMG_1] Fig. 1. Relationships between zygotes and development types a — equifinality of typical development of normal heterogeneous zygotes; b — limitation of the species‑specific space of developmental possibilities under any variations of the germ‑cell genome. Solid lines — stable developmental path; dashed lines — labile developmental paths; p1–n3 — normal zygotes, m — abnormal, N — adaptive norm.A comparable theory of developmental systems, describing the relationships between the individual structure of the zygote, developmental conditions, and the final outcome, has not yet been created. Yet its main principles have been clear for a long time. It can be argued that its foundation was laid by R. Goldschmidt’s (Goldschmidt, 1938, 1940) notion of the reducibility of all phenotypic deviations, regardless of their initial causes, to quantitative shifts within the developmental system. Another important element is C. Waddington’s (Waddington, 1957) teaching about the epigenetic landscape as the structure of this system and about genetic assimilation as a means of its remodeling. We will attempt to integrate these views into a single coherent concept and trace the most obvious conclusions from them that are consistent with experimental genetic data. The concept of a species‑specific developmental system is used by us following the cited authors, alongside synonyms such as “reactive system” or “epigenetic system.” A related commonly used notion is “reaction norm” (Johannsen, 1926), denoting the possibilities of epigenetic realization of an individual genotype. [IMG_2] Fig. 2. Dependence of allelic phenotypic changes on quantitative gradations of the factor governing developmental pathway switching (illustrated with the vestigial wing mutation in Drosophila). Intensified thermal exposure during larval development shifts the mutant phenotype toward weaker alleles, eventually matching the norm: the moment of deviation from the normal pathway is displaced to progressively later stages. N — norm, vg1–vg5 — phenotypes of an allelic series Goldschmidt’s ideas about the systemic properties of individual development stem from the already mentioned fact of parallelism between mutational and modification changes (first experimentally demonstrated with genetic control by N. V. Timofeyev‑Resovsky in 1926). As experiments with Drosophila and other organisms show, the effect of virtually any mutation, including its pleiotropic manifestations, can be produced as a morphosis (phenocopy) by shock treatments applied to one or another sensitive developmental period. From this, Goldschmidt concluded that both types of phenomena share the same physiological basis—nonspecific disturbances of normal developmental event coordination, namely mismatches in the rates of morphogenetic reactions and changes in the quantities, concentrations, and timing of interacting substances. The possibilities of such disturbances that still allow development to be completed are limited within the system, thereby limiting the set of phenotypic deviations it can realize. Mutations in this system can cause only those anomalies that can also be produced by external influences that generate analogous quantitative shifts within it. Mutational effects, including all their pleiotropic manifestations, are not properties of the gene but embryological consequences of the time, place, and type of primary developmental disruption caused by the mutation (Goldschmidt, 1933; Goldschmidt, 1938, 1940, 1955; Kamshilov, 1940; Schmalhausen, 1982). The character of the anomalies ultimately reflects the holistic properties of the system itself, not the specificity of any particular chromosomal locus. All this is supported by numerous phenogenetic analyses of mutations, for example those such as Bar or vestigial in Drosophila melanogaster. Phenotypes of the entire series of alleles (up to the norm) for each can be obtained by gradating temperature exposure of the larva. Clearly, specific allelic phenotypic changes here represent the developmental system’s response to purely quantitative alterations of a temperature‑sensitive factor. Moreover, the smaller the factor’s deviation from the norm (neutralized in the named mutations by increased temperature), the later development is disrupted (Fig. 2). Thus, at maximal vestigial effect the defect appears already at the imaginal wing disc stage, and it does not develop beyond the base; at the minimal reduction the wing forms completely, and only later its blade is partially lysed in the pupa (Goldschmidt, 1938, 1955). In some cases the nature of the quantitative factor determining the character of a qualitative phenotypic anomaly can be identified more concretely. For instance, it may be the growth and segmentation rate of the primordium, as shown for the aristopedia mutation (in Drosophila) linked to the transformation of the antennal arista into a limb‑like structure. In the mutant, the imaginal disc growth rate is elevated to the level typical of a normal limb primordium; however, when growth is delayed with colchicine, development proceeds along the normal pathway (Balkashina, 1928; Goldschmidt, 1938, 1955). In mice, the Dh mutation causing polydactyly of the hind limbs operates by slowing the death of cells in the apical ectodermal ridge, thereby prolonging its inductive action on the limb bud. Conversely, the Os mutation accelerates ridge regression and shortens the induction period, leading to oligodactyly or even reduction of the bud itself (Konyukhov, Nonchev, 1981). Such examples lead many researchers to acknowledge a nonspecific and indirect influence of genes on the choice of realized phenotype (Wolpert, 1976; Alberch, 1982). If the nature of an anomaly is indeed not directly determined by the specificity of the initial disturbance, then this should also be observable under different modes of external influence on development. Phenocopy experiments confirm this. Many types of shocks directed at the same sensitive developmental period yield the same result, and conversely, the same shock can produce qualitatively different anomalies depending on the developmental phase it affects, as well as its intensity and duration (Goldschmidt, 1955). On the other hand, we should expect that genetically distinct disturbances will also lead to identical anomalies. This indeed occurs. The entire body of experimental genetics indicates that mutations uniquely effective in their phenotypic impact apparently do not exist (Timofeyev‑Resovsky, Ivanov, 1966), which merely restates Goldschmidt’s thesis on the nonspecificity of mutational impact on the developmental system. Accordingly, researchers identify more or less extensive “heterogeneous groups of genes” that have different locations but cause similar or identical anomalies when mutated. A characteristic example is the minute group in Drosophila (shortening of thoracic bristles), encompassing about 60 loci across three chromosomes (Timofeyev‑Resovsky, Ivanov, 1966). Likewise, the bithorax phenotype (doubling of the median thorax) can be obtained by a mutation of the eponymous locus on chromosome III, by a combination of mutations on three different chromosomes, and finally by a “normal” genome subjected to shock exposure of the larva (Waddington, 1966). The same is observed in analyses of natural phenotypic deviations, where identical types are linked to mutations on different chromosomes or simply to external influences (e.g., the Abnormal abdomen phenotype in Drosophila; Hlubovsky et al., 1974). Finally, it is well known that identical normal or aberrant traits in closely related species or geographic races of a single species often arise from different genetic bases; common examples include parallel flower coloration in cotton species (Harland, 1937) or wing coloration in butterflies, such as the yellow form of the German and Italian Callimorpha dominula or the white form of the English and Canadian Biston betularia (Goldschmidt, 1940; Sheppard, 1970). Genetic analysis of these analogues yields different segregation types. In cases of different species, where such analysis is usually difficult, these examples are often linked to parallel or homologous mutations; however, we should understand this not as the presence of “the same gene” but as similarity of epigenetic systems that permit the same developmental pathway despite different genetic structures (Goldschmidt, 1945). All these facts clearly indicate an overall indeterminate correspondence between primary elementary disturbances of the developmental process and its particular outcomes. The aggregate of such outcomes preserves species‑level stability with respect to variations in developmental conditions, i.e., no phenotypic trait is determined by a fixed combination of chromosomal locus states and external factors. This contrast (asymmetry) between the multitude of possible developmental conditions and the limited spectrum of possible outcomes shows that events belonging to different hierarchical levels of a single holistic dynamic system are being compared, a system that exhibits stable behavior. Essentially, phenogenetic analysis of mutant anomalies reveals the same systemic property of development that earlier experimental embryology encountered—the fundamental impossibility of reducing the developmental result to a specific sum of initial causes. Moreover, it becomes evident that this applies not only to the normal course of ontogeny but to the entire space of its aberrations. Of course, this does not deny the specificity of the primary function of genomic loci. It merely means that changes in that function by themselves do not determine phenotypic traits, but manifest as quantitative disturbances of certain parameters of the developmental system, to which it responds qualitatively, altering the choice of phenotypic outcome. It is not implied that the primary products of different elementary disturbances causing identical aberrations are identical. Their mechanisms of action must differ. In genomic changes, these are various reaction chains leading to unregulated shifts at a particular developmental phase; external stimuli disrupt that phase more directly (Goldschmidt, 1955). The presented ideas imply that parameters of morphogenetic processes constituting development are characterized in each epigenetic system by a specific set of threshold values, crossing which determines the selection of a particular developmental trajectory. It is necessary to clarify how this selection is made and to understand the difference between normal and aberrant developmental pathways within a single system. The answer is largely illuminated by the model of the epigenetic landscape proposed by Waddington (1947, 1960, 1970a; Waddington, 1957, 1966), which describes the general properties of the developmental system as a series of branching inclined valleys diverging from a common initial point (Figs. 3, 7). This model has a dual meaning. First, it reflects the usual interpretation of development as a hierarchy of stages of increasingly specific differentiation of primordia, leading to progressive restriction of their further formative possibilities. The system of trajectories, or valleys, along which individual processes “move,” symbolizes the limited and discrete nature of differentiation pathways available to derivatives of a given embryo or primordium. This property of development is a well‑known fact demonstrated, for example, by the behavior of embryonic tissues in explants (Svetlov, 1964). The second, more substantial meaning of Waddington’s model is that it allows the depiction of the normal developmental path of an individual primordium (and, in the limit, of the whole organism) against the backdrop of the entire field of its potential developmental possibilities within the epigenetic system characteristic of the species. In this view, the normal path corresponds to a deep valley, or “creode” (literally “necessary path”), while alternative possibilities correspond to its shallower branches. Deviation onto any of them involves overcoming a more or less high threshold (separating that valley from the creode floor) and signifies a disruption of normal development. Because branching points correspond to depressions in the creode walls (Fig. 3), they mark moments of relative instability in the choice of determination, i.e., sensitive periods whose perturbation leads to experimental aberrations. [IMG_3] Fig. 3. Relationships between the structure of the epigenetic landscape and the nature of damaging influences a — deviation of development into a side valley due to strong external impact (long arrow); b — the same deviation caused by a strong mutation disrupting the creode; c — intermediate state (interpretation modified from Waddington, 1957) Stable (canalized) development, or movement along the creode toward an adaptive norm, is ensured by its regulation. Regulation is expressed in that a process, once deviated by some influence, rolls back into the valley’s channel if the displacement does not exceed its slopes. Since, as normal development proceeds, the capacity for regulation usually declines, this means a gradual flattening of the creode. When several adaptive norms exist, the developmental system possesses multiple alternative creodes, the choice among which is controlled either by environmental factors (in modification polymorphism) or by systematic recombination of the chromosomal apparatus (e.g., sex determination). In contrast to the norm, aberrant developmental paths represented by flattened valleys have limited regulatory capacity, and their outcome is relatively labile (a general rule for mutations and non‑adaptive morphoses; Schmalhausen, 1968b; Waddington, 1944, 1970a). In other words, the discreteness of these paths relative to the norm does not imply their stability. Conversely, this discreteness itself is not absolute and becomes weaker the later an irreversible deviation occurs in development. For example, it is minimal or absent in weak expressions of Drosophila mutations such as eyeless or vestigial (Goldschmidt, 1938; Rapaport, 1943). Thus, the epigenetic landscape characterizes a species‑specific space of developmental possibilities, encompassing regions of stable process flow (creodes), regions of most probable aberrations (side valleys), and zones with minimal likelihood of realization (watersheds between valleys that developmental trajectories tend to avoid). This structure reflects the properties of a holistic dynamic system, showing that its response to perturbations depends on which point of its spatiotemporal extension was affected. The closer a point is to the creode region, the more likely that diverse influences will be buffered similarly, whereas in instability zones similar causes can have profoundly different consequences. In a more general form, the course of the entire development of a multicomponent system can be described as the movement of a point in a multidimensional (phase) space, where its coordinates at any moment correspond to measures of individual interacting elementary components (Waddington, 1957). Creodes correspond to the most stable trajectories, capable of attracting nearby points; the whole set of possible trajectories in this space constitutes the system’s “phase portrait” (Beloousov, 1979; Beloousov, Chernavsky, 1977), visually represented by a three‑dimensional model of the epigenetic landscape. The relatively simple ordering of the developmental system demonstrated by this model is a “higher‑order property” relative to the functions of elementary genomic units (Waddington, 1957, p. 34) and is based on the interaction of their entire ensemble. Now consider the character of epigenetic system reactions to elementary perturbations. Obviously, such influences cannot transform the system itself, but only change its state. They either alter the choice of trajectory within the landscape or are subject to regulation and do not change the developmental outcome at all. In both cases the final result is determined by the properties of the system itself. When the intensity of an influence exceeds the regulatory capacity of canalized development, its effect can be interpreted in two quantitative ways. Either the level of some critical factor capable of diverting development into a side valley exceeds the threshold permissible at that time point for normal process flow, or the threshold separating the creode from that valley disappears (i.e., sensitivity to the switching factor increases). Both types of changes may also occur together. Clearly, this is the essence of the quantitative shifts to which, as Goldschmidt showed, the consequences of all diverse elementary influences on the developmental system inevitably converge. Based on the foregoing, possible outcomes of a single damaging factor’s impact on individual developmental cycles in a normal heterogeneous population can be described. All zygotes of a species belong to variants of the same epigenetic system; their genetic differences under identical developmental conditions determine individual details of landscape modeling (different heights of thresholds between valleys, variations in the degree of expression of the latter). For normal zygotes these differences are minimal, i.e., the path leading to the normal phenotype is the most stable (canalized) for them and exhibits only minor variations in the form of local differences in the heights of protective creode thresholds. For elementary influences capable of causing deviations from the norm, three main situations are possible (Shishkin, 1984a, b). 1. A sharp genomic change that reduces the protective threshold on a specific segment of the creode so strongly that, regardless of variations in its previous height, development invariably deviates onto the same side path (Fig. 3, b). This is an ideal mutation, most convenient for genetic analysis, i.e., a locus‑specific alteration that, when introduced into any genome variant, most likely produces a particular developmental anomaly.{ "translated_text": "However, in practice the result still turns out to be not entirely stable, because the smoothing of aberrant valleys of the landscape itself excludes effective regulation of ontogenetic trajectories by them. And indeed, in practice the expression of even strong “raw” mutations remains variable (Waddington, 1970a). 2. The opposite situation – an extreme strong external impact on the course of development, overcoming any threshold of its stability at a given time point and leading in this developmental cycle to the same result as a strong mutation (Fig. 3a). 3. Between these two extreme cases lies a huge region of intermediate states, when the character and even the possibility of deviation depend on the specific ratio between features of the epigenetic landscape and developmental conditions (Fig. 3b). The action of the same mutation will either lie below the disruption threshold or exceed this threshold at various points of the creode – depending on individual landscape features (determined by the original genome constitution) and environmental factor fluctuations. Conversely, the same deviations must arise under different combinations of external and internal developmental conditions. In these cases we speak of mutations with unstable expression and manifestation, i.e., not showing proper Mendelian inheritance upon analysis.\nThis last, most typical situation corresponds to the real picture of indeterminate variability observed in natural populations. It is well known that the category of aberrations that can be characterized as dominant mutations with good expression is absent or rare in them (Gershenson, 1941). At the same time, even large homogeneous aberrations, when analyzed, turn out to be associated with different chromosomes or parts induced from outside (for example, Abnormal abdomen in Drosophila; Golubovsky et al., 1974). When sufficiently large samples of such phenotypes are available, they display a wide range of inheritance stability – from ratios close to Mendelian to complete loss of expression (for example, the “spotted eyes” phenotype in Drosophila; Dubinin et al., 1937). Therefore, within such groups of isoreagents the authors often do not even attempt to draw a line between hereditary (mutational) and non‑hereditary (modificational) changes (Balkashina, Romashov, 1935) or they look for it between lines with minimal inheritance of the aberration and lines with its complete absence. It is quite obvious that the boundary spoken of does not exist in nature. All phenotypes of one class are variants of the same ontogenetic trajectory, differing in stability because they are conditioned by the most diverse combinations of individual genetic constitution and environmental factors. The concepts “mutation” and “modification” are in fact incomparable, since the former refers to comparisons of individuals, and the latter – to comparisons of developmental possibilities of the same individual. All identical phenotypes (as well as any others) are always genetically non‑identical and therefore can be regarded as hidden mutants relative to each other regardless of hybrid analysis results. Conversely, any phenotype evaluated on the basis of such analysis as mutant represents only one of the developmental possibilities within the epigenetic system of a given zygote, i.e., one of the alternatives (modifications) relative to the species‑normal trajectory. The latter is especially evident in cases where return to this trajectory is practically easy by changing developmental conditions (for example, the pennant, vestigial, Abnormal abdomen mutations and others in Drosophila; Shmalhausen, 1968).\nThus, the heterogeneity of homogeneous developmental anomalies predicted by the epigenetic landscape model is confirmed by the real picture of their inheritance in natural populations. Likewise, another conclusion of this model is confirmed – that the same external disturbances will, as a rule, cause divergent deviations from normal development according to the individual constitution of the affected zygotes. The increase in variability under disrupted normal conditions, noted already by Darwin, is a well‑known fact comparable to the dispersion of a light beam passing through a prism (Lobashev, 1947). The transition of a population under extreme conditions from phenotypic uniformity to the manifestation of divergent variations was termed the unveiling of a “mobilization reserve” of variability (Shmalhausen, 1941; Gershenson, 1941). This phenomenon means that the canalizing mechanisms of development, which buffer individual genetic differences of organisms, become disrupted upon reaching a certain threshold of external influences, resulting in these differences being expressed as phenotypic aberrations. All normal zygotes in the population differ both in the width of the interval of conditions allowing canalized development and in the nature of the morphoses they undergo under the same extreme conditions (Fig. 4, 5).\nBased on the concept of the epigenetic landscape, other predictions become possible and experimentally testable. If the disruption of normal development of organ A, caused by a mutation (or any damaging influence), reduces to a change in the magnitude of some factor destabilizing trajectory A, then one can expect that an artificial redirection of the development of this primordium toward organ B will render its new trajectory independent of the specified factor, i.e., it will be “out of reach” of mutations affecting only organ A. Conversely, mutations affecting the development of organ B should also act on any other primordium determined toward B, i.e., the sensitivity of primordia to particular disturbances should be determined not so much by their normal determination as by the choice of the actual developmental path. These regularities were indeed established by I. A. Rapaport (1941), who showed that in Drosophila the effect of the dominant Met mutation, leading to the absorption of presumptive wing material by the mesothorax, is not altered by the introduction of various wing mutations into the genome. Conversely, mutations that normally affect the structure of the dorsal bristles, but do not involve the wing, have the same effect on a giant mesothorax altered by the Met mutation.\n[IMG_4]\nFig. 4. Individual intra‑population differences in the stability of normal morphogenesis\na – continuous curves of phenotypic realization of individual genomes versus external conditions; solid sections correspond to the norm (with modifications A and B), dashed sections to maladaptive morphoses; b – modification spectra controlled by these genomes within the same interval of conditions extending beyond the normal range. Spectrum sections corresponding to modifications A and B are hatched.\n[IMG_5]\nFig. 5. Dependence of morphogenesis on developmental conditions\nThe solid line corresponds to the overall adaptive norm for the population (with two modifications); dashed lines correspond to individual morphoses. Horizontal axis – environmental change; vertical axis – developmental result: a–a1 – interval of conditions permitting normal development.\nFrom the considered representations follows another conclusion – that the discreteness of phenotypic changes caused by mutations of a single chromosomal locus reflects properties of the whole developmental system, not the mutations themselves. Early works by Goldschmidt (1916–1917) on the mechanism of sex determination and by S. Rait (Wright, 1916) on the phenogenesis of albinism showed that allelic changes are based on gradations of the same factor (e.g., amount of substance or reaction rate) acting with a threshold effect. From the viewpoint of the epigenetic landscape model, this discreteness of response (decreasing in later developmental stages) is an inevitable consequence of the fact that the stability of a canalized trajectory to fluctuations in the level of various morphogenetic factors declines intermittently over time, creating jumps at “sensitive periods” corresponding to branches of aberrant valleys (Fig. 6a). The higher the level of a given factor exceeds the regulated limits under the influence of mutations, the earlier the affected critical phase, i.e., a trajectory increasingly divergent from the normal one, is realized. Fluctuations of the factor between two threshold levels (Fig. 6b, levels B and C), defining the extreme possibilities of a given deviation, do not acquire phenotypic expression, and all one‑locus mutations causing them will be regarded as the same allele. Exceeding a threshold will lead to a discrete phenotypic change, i.e., a switch of development to an earlier or later branch of the normal trajectory.\nIt needs little explanation that the concept of an epigenetic system objectively underlies the theory of stabilizing selection. The view of development as a system with a limited set of most probable trajectories (phenotypic states) allowed the breakdown of the impassable barrier between mutational and exogenous changes that existed for classical genetics (Dobzhansky, 1947, p. 209), reducing both to discrete reactions of the whole system to quantitative changes in switching epigenetic factors (Shmalhausen, 1982, pp. 82, 89, 103, 170–173). Establishing this regularity opened the way to understanding that inheritance stability is not a property of individual chromosomal genes, but the result of selection creating a new genotype organization that ensures canalized development of a previously labile (non‑heritable) trait. Heredity thus became an integrated, historically conditioned property defined as the stability of the result of epigenetic interactions (Shmalhausen, 1982; Waddington, 1957), and natural selection, from a sorter of independent hereditary units, turned into a mechanism for generating inheritable changes.\n[IMG_6]\nFig. 6. Discreteness of aberrations as expression of stepwise reduction of noise‑resistance of canalized development\na – a section of the epigenetic landscape showing jumps in the height of walls (stability thresholds) of the creode at points where aberrant valleys branch; b – dependence of aberration character on the ratio between the level of a damaging morphogenetic factor and creode stability. In the interval between two stability thresholds (hatched) fluctuations of the factor’s magnitude do not change the type of development; a, b, c, d – threshold levels of stability for successive creode segments; A1–A3 – aberrant developmental paths; N – normal course of factor level changes and typical developmental path. Vertical axis – factor level; horizontal axis – time.\nShmalhausen’s (1968a, 1982) ideas about the historical emergence of new adaptive traits are easily interpreted as a description of transformations of the species‑specific epigenetic landscape. According to these views, the whole process starts each time with the labilization of the previous norm’s development (Fig. 7a,b; N) and the selection of one of the elementary maladaptive reactions that arise. This means that under new conditions individual landscape variants realize different aberrant trajectories, one of which leads to the most viable phenotypic deviation (Fig. 7b, N1). As selection proceeds in its favor, the former normal trajectory loses stability regardless of developmental conditions, i.e., its valley becomes smoothed, erasing differences in stability between the former norm and the selected deviation (Fig. 7b, N, N1). Gradual selective stabilization of the new adaptive trajectory turns it into a creode (Fig. 7c,g, N1); simultaneously, the entire zone of the landscape around it transforms from a region of unlikely events into a region of most common deviations, i.e., it fragments into new valleys. Thus, the whole landscape pattern changes progressively (Fig. 7a–g). In other words, the aberrant space of the epigenetic system is reshaped.\n[IMG_7]\nFig. 7. Reorganization of the developmental system during the emergence of a new elementary adaptation\n1 – changes in the epigenetic landscape, 2 – typical changes in modification spectra, 3 – changes in the dispersion of the transformed adaptive trait in the population: a – canalized development of phenotype N (corresponding to the main band in spectra and the peak of the variation curve); b – destabilization of development leading to a decrease in the frequency of phenotype N and its role in spectra; c, g – two successive stages of canalization of ontogenetic trajectory N1 with the development of a new network of aberrant valleys around it. Phenotype N1 transforms from an unstable morphosis into a new adaptive norm; the former norm becomes a morphosis or disappears. Creodes are highlighted with bold lines; development types corresponding to successive norms are hatched. Vertical axis – frequency of the trait, horizontal axis – trait change." }Thus, at any stage of its evolutionary changes, the system detects a specific set of developmental possibilities that is characteristic of it only in that particular period of its history. Therefore, the claim that the material of evolution consists of random gene variations is, in Waddington’s words, “empty” (Waddington, 1957, p. 188). No matter how random nucleotide changes in chromosomal DNA may be, their possible effects on the phenotype are always limited by the historically formed structure (landscape) of the epigenetic system. A change in this structure is expressed at the population level as a change in the nature of indeterminate variability (Kamshilov, 1967). Hence Darwin’s (1952) statement that variability is not caused by selection must be understood correctly. Independent of selection is only its mere existence, not the character of its phenotypic expression. **EPIGENETIC SYSTEM AND THE CONCEPT OF THE MENDELIAN FACTOR** The presented ideas inevitably lead us to the problem of ontogenetic interpretation of Mendelian inheritance. The notion of a Mendelian factor implies the presence of two stable alternative (allelic) states of a trait, each inherited in the corresponding “pure line” unambiguously, and in hybrids (the second generation) in certain numerical ratios. From an epigenetic perspective this means that the development of the trait in the compared groups of individuals is channeled in two different directions, with the robustness of each based on the interaction of all genotype elements and expressed as insensitivity to recombination effects that arise when crossing within a given group (line). In inter‑line crossing, both directions (creodes) merge in a single epigenetic landscape as stabilized branches of one trajectory, and the developmental outcome can in principle oscillate between two extreme situations. In one of them a particular intermediate aberrant trajectory is realized, often yielding an unstable result, which in Mendelian analysis practice is described as a lack of uniformity among first‑generation hybrids (Filippenko, 1924). In the second case (Fig. 8) development is directed toward one of the two existing attraction regions (creodes), i.e., pronounced dominance is present. In any variant, the choice of developmental path depends on a nonspecific shift in the phase of unstable determination (the branching point of creodes), ultimately linked to the function of a particular pair of homologous genomic units. The latter is identified by Morgan’s chromosomal genetics with the allelic states of a Mendelian gene. As already noted, phenogenetic analysis of allelic series of phenotypes shows that their discreteness reflects not the properties of the determining factor itself, but a threshold response of the developmental system to its quantitative changes. From this viewpoint, the action of a pair of loci responsible, in a given system state, for switching the developmental path can control three gradations of the factor—two extremes (parental) AA and aa and one intermediate (hybrid) Aa, denoted in genetics as homo‑ and heterozygous states. Between them lie either two critical thresholds, or in the case of dominance only one (between aa and Aa—Fig. 8). It thus appears evident that ordered realization of each parental phenotype does not require exactly the same intensity of action of “iso‑allelic” switch loci (or, even more, identical nucleotide structure). It is sufficient that the cumulative effect of any individual pair taken from the aa line does not exceed the threshold permitting development toward the corresponding phenotype (Fig. 8). In both the general case and under dominance this is the threshold between the aa and Aa gradations. When first‑generation hybrids are crossed, according to the laws of chromosomal segregation in meiosis, zygotes with locus combinations of all three types arise in the quantitative ratio 1:2:1, which means that the three corresponding measures of the switching factor and their associated developmental variants are realized in the same proportions, i.e., Mendelian phenotypic segregation occurs. If the system possesses only one switching threshold (dominance), only two stabilized initial variants arise in a 3:1 ratio. [IMG_8] Fig. 8. Ontogenetic mechanism of Mendelian inheritance Choice of developmental path in the second generation of hybrids stabilized lines with phenotypes A and a. If at the critical developmental moment X the level of the switching morphogenic factor does not exceed the aa threshold, development deviates from creode A toward the recessive phenotype a. If the threshold is exceeded, development proceeds toward the more stable (dominant) phenotype A. Horizontal axis—time; vertical axis—measure of the switching factor and direction of development. Aa, AA—above‑threshold factor values realized in hybrids. In the first critical phase the two realized developmental variants give a 3:1 ratio depending on whether the factor I threshold (aa) is exceeded. In the second critical phase, where the choice is determined by factor II (threshold vv), each variant differentiates into two new ones—also in a 3:1 ratio. Ultimately four phenotype classes arise in the proportion 9:3:3:1. AA, Aa, aa, BB, Bv, vv—levels of switching factors determined by recombination of two pairs of loci; AV.., Av.., aV.., av..—classes of realized phenotypes. Any ordering of biological phenomena must be regarded as a product of natural selection (Mayer, 1981), and regarding Mendelian inheritance rules this connection is entirely evident. The ordering of developmental outcomes here rests on the presence in the parental epigenetic system of the corresponding creodes, which can arise only through stabilizing selection. Genomic mutations, like other singular perturbations of the system, can, as shown, only disrupt creodes, but do not themselves create robust developmental pathways. Accordingly, the expression “raw” mutational anomalies, both natural and experimental, is highly variable and generally poorly conforms to Mendelian rules (Dubinin et al., 1937; Gershenzon, 1941; Kamshilov, 1940; Schmalhausen, 19686), leading to the concepts of a gene’s genotypic environment, as well as expressivity and penetrance of a mutant trait. Only the effects of the largest genomic disruptions show relatively little dependence on genotype variations, i.e., such mutations fairly deterministically produce a particular type of anomaly; yet this correspondence is never completely stable [The latter, in particular, concerns large genetic anomalies in humans, which, even when dominant, may sometimes not manifest in the homozygote and generally show variable expression (e.g., polydactyly; Gershkovich, 1968)].

**EPIGENETIC SYSTEM AND UNDERSTANDING OF THE MENDELIAN FACTOR**

The necessity of proper inheritance being preceded by stabilization of the crossed phenotypes is evident already from the fact that Mendel’s laws were established precisely on rigorously selected (stabilized) lines carrying alternative traits. The specificity of the conditions under which these laws hold is especially illustrated by the second law—the law of independent assortment—stating that in inheritance each pair of allelic factors “behaves as a single unit” (Filippenko, 1924), i.e., always yields in the second hybrid generation a segregation into the two parental phenotypes regardless of any recombination variants of all other factors. From the standpoint of chromosomal genetics, acknowledging the universality of this rule would be equivalent to an unacceptable claim that gene expression does not depend on variations of the genotypic environment. Clearly, such independence is possible only if the influence of recombination on developmental trajectories has been neutralized by prior selection, so that any recombination variant leaves only a choice between two possibilities, determined at some critical developmental moment by the state of a single pair of loci. Accordingly, it is indeed recognized that a prerequisite for genetic analysis based on Mendelian inheritance rules is the prior selection of lines for the traits under study (Lobashev, 1966); and, in particular, mutant phenotype analysis begins from this point. In genetic thinking, the result of such selection is the creation of a pure line of carriers of the mutant allele, i.e., homozygous for the locus identified with the anomalous trait. From an epigenetic viewpoint, this procedure means stabilizing a previously unstable developmental trajectory of the trait, based on a reorganization of the entire genotype of that line (Waddington, 1957). The question arises: what is the “decisive” role of a single locus pair in defining such a trait, as revealed by crossing analysis? First, it characterizes only its ordered development created by selection, not the original state in which it was merely an unstable anomaly. Second, this defining role does not pertain to the unequivocal channeled development observed in parental lines, but to the hybrid variant where two different developmental possibilities coexist in one epigenetic system. The essence of the ontogenetic action of a “determining” locus pair in a hybrid system is that, during the temporal interval corresponding to the choice between two trajectories, its effect exceeds the regulation of one alternative creode while remaining within the normal range for the other, so that one of the stabilized developmental variants still occurs. The switch determines only the choice of developmental path, not its execution, which is governed by the entire genotype. Regarding homozygosity of the parental line for one member of the “determining” pair, as noted, this concept does not mean absence of locus variability, but rather that its variability is limited to those bounds within which the action of its double dose does not disrupt the stable development marking that line (Fig. 8). Such variability limits, constraining the possibilities of channeled development, exist in this system for any pair of loci; thus all of them are, in this sense, “determinants” of the trait.

**Fig. 8. Ontogenetic mechanism of Mendelian inheritance** The choice of developmental path in the second generation of hybrids is stabilized by lines with phenotypes A and a. If, by the critical development point X, the level of the switching morphogenetic factor does not exceed the threshold value aa, then development deviates from creode A towards the path corresponding to the recessive phenotype a. If this threshold is exceeded, development continues towards the more stable (dominant) phenotype A. Horizontal axis – time, vertical axis – degree of switching factor and direction of development. Aa, AA – suprathreshold factor values realized in hybrids.

In the first critical phase, two realized developmental variants yield a ratio of 3:1, depending on whether the threshold value (aa) of factor I is exceeded. In the second critical phase, where the choice is determined by the level of factor II (with a threshold value of bb), each of the variants differentiates into two new ones – also in a ratio of 3:1. As a result, four classes of phenotypes arise in a ratio of 9:3:3:1. AA, Aa, aa, BB, Bb, bb – levels of switching factors determined by recombination of two pairs of loci; AB.., Ab.., aB.., ab.. – classes of realized phenotypes.

Any ordering of biological phenomena must be regarded as a product of natural selection (Mayer, 1981), and regarding Mendelian inheritance rules this connection is entirely evident. The ordering of developmental outcomes here rests on the presence in the parental epigenetic system of the corresponding creodes, which can arise only through stabilizing selection. Genomic mutations, like other singular perturbations of the system, can, as shown, only disrupt creodes, but do not themselves create robust developmental pathways. Accordingly, the expression “raw” mutational anomalies, both natural and experimental, is highly variable and generally poorly conforms to Mendelian rules (Dubinin et al., 1937; Gershenzon, 1941; Kamshilov, 1940; Schmalhausen, 19686), leading to the concepts of a gene’s genotypic environment, as well as expressivity and penetrance of a mutant trait. Only the effects of the largest genomic disruptions show relatively little dependence on genotype variations, i.e., such mutations fairly deterministically produce a particular type of anomaly; yet this correspondence is never completely stable [The latter, in particular, concerns large genetic anomalies in humans, which, even when dominant, may sometimes not manifest in the homozygote and generally show variable expression (e.g., polydactyly; Gershkovich, 1968)].

The conditionality of correct inheritance by prior stabilization of crossed phenotypes is evident from the fact that Mendel's laws were established precisely on strictly selected (stabilized) lines carrying alternative traits. The content of the second law – the law of independent recombination – which indicates that in inheritance each pair of allelic factors "behaves as a single unit" (Filipchenko, 1924), i.e., always gives a segregation into two parental phenotypes in the second generation of hybrids against the background of any recombination variants of all other factors, particularly vividly illustrates the specificity of the conditions under which these laws are fulfilled. From the perspective of chromosomal genetics, acknowledging the universality of this rule would be equivalent to the unacceptable assertion for it that the expression of a gene is independent of genotypic environment variations. It is obvious that such independence is possible only if the influence of recombination on the developmental process is neutralized by prior selection, so that in any of its variants, only the choice between two possibilities is preserved, determined at some critical moment of development by the state of only one pair of loci. Accordingly, it is indeed recognized that a necessary prerequisite for genetic analysis based on Mendelian inheritance rules is the prior selection of lines for the traits being analyzed (Lobashev, 1966); and, in particular, the analysis of mutant phenotypes begins with this. In the concepts of genetics, the result of such selection is the creation of a pure line carrying the mutant allele, i.e., homozygous for the locus with which this anomalous trait is identified. From an epigenetic point of view, this procedure means stabilizing a previously unstable developmental trajectory of a trait, based on the reorganization of the entire genotype of this line (Waddington, 1957).

The question arises: what is the 'decisive' role of one pair of loci in determining such a trait, revealed by analyzing crossing? Firstly, it characterizes only its ordered development, created by selection, and not the initial state, in which it was merely an unstable anomaly. Secondly, this determining role concerns not the unambiguous canalized development of the trait observed in the parental lines, but the hybrid variant, where two different possibilities of such development are combined in one epigenetic system. The essence of the ontogenetic action of the 'determining' locus pair in the hybrid system is that, in the time segment corresponding to the choice between two trajectories, its effect goes beyond the regulation of one of the alternative creodes but remains within the norm for the second, so that one of the stabilized developmental variants is still realized. The switch here only determines the choice of the developmental path, not its actual realization, which is determined by the entire genotype. As for the homozygosity of the parental line for one of the members of the 'determining' pair, as already stated, this concept means not the absence of locus variability, but its limitation within the boundaries where the effect of its double dose does not disrupt the stable development of the trait marking this line (Fig. 8). Similar variability limits, restricting the possibilities of canalized development, exist in this system for any pair of loci, meaning all of them are 'determinants' of the trait in this sense.

Thus, the concept of the Mendelian gene implies a certain state of the species' epigenetic system, in which, due to selection, two stable variants of trait development become possible, chosen depending on the combination of states of a specific pair of chromosomal loci. Therefore, the laws of their recombination, determined by the behavior of chromosomes during meiosis and fertilization (separation and independent assortment of homologs), acquire unambiguous expression in the inheritance of alternative traits with such a system organization. This unambiguity constitutes the essence of Mendel's laws, meaning the Mendelian gene is an indicator of a certain orderliness of the epigenetic system, not the chromosomal locus itself. The latter acquires the properties of a switch for allelic traits only to the extent that the possibility of switching itself exists within the hybrid developmental system, manifested by the presence of two stabilized creodes. Only under these conditions do the concepts of homo- and heterozygosity make sense, as they denote those gradations of the epigenetic effect of the locus pair that determine the choice of discrete developmental pathways in this system; the position of thresholds between these gradations, as well as the presence of the activated pathways themselves, characterize the system, not the genome region. In the absence of the second creode in the system (i.e., when the trait has only one norm), any genomic change (recombination, mutation) can have only two effects – either it does not affect normal development, or it leads to unstable aberrations that do not show proper inheritance (as observed with natural indefinite variability). This property of aberrations means that for a single-creode system, the concept of an allele loses its meaning altogether, as there is simply no alternative possibility of stable development that could be activated by any locus change.

It follows from the foregoing that a Mendelian factor is not a material particle, but a relationship between two stable alternative states of an epigenetic system ('pure lines'), revealed in hybrid analysis. This relationship does not exist outside the comparison of these states. The number of Mendelian genes determining the analyzed trait in a given crossing system means nothing other than the number of binary choices between successive branches of canalized trajectories that must be made in the hybrid's developmental system to ultimately obtain one of the parental phenotypes (Fig. 9). It is easy to see that if each such choice depends in a given system on one pair of loci, giving a 3:1 ratio upon realization of the two corresponding trajectories (creodes), then the total number of possible developmental outcomes (i.e., different phenotypes) will be 2ⁿ, and their quantitative ratio (3+1)ⁿ, where n is the number of successive critical phases corresponding to bifurcations between daughter creodes, or, equivalently, the number of 'determining genes' (Fig. 9). These ratios constitute the essence of Mendelian inheritance.

**Fig. 9. Dihybrid segregation as a result of two successive choices between dichotomous creodes in the epigenetic system of second-generation hybrids**

All this makes it clear that the concepts of Mendelian allele and genomic mutation denote entirely different phenomena belonging to different hierarchical levels of the system. The Mendelian gene with its allelic states characterizes a mode of system organization in which a wide spectrum of genome variations as a whole does not lead development beyond a limited set of discrete stable pathways; the choice among them each time depends only on the total activity level of one specific pair of homologous loci. Conversely, a genomic mutation signifies an elementary disruption in the system that can only disrupt its orderliness but cannot create a new one. To equate a Mendelian gene with a chromosomal region and an allelic change with a mutation of that region means attributing the properties of the system itself to an element of the system.

This identification, adopted, albeit not without hesitation, by Morgan's chromosomal genetics (Moran, 1937), compels us to return to the question of the possibility of gene change under selection. Such a possibility, seemingly derived from a series of experiments (Castle, 1916), was vigorously rejected by orthodox Morganian doctrine, which saw the gene as a locus capable only of spontaneous mutation. But even if we consider the locus itself, in modern understanding, it is a polynucleotide matrix, whose sensitivity to selection is recognized to some extent even by the neutral theory of molecular evolution. All the more reason for no doubt regarding the selective determination of Mendelian factors (which are revealed by classical genetics methods), since, as shown above, they are simply symbols of epigenetic relationships created by selection. Simply put, the Mendelian gene is a product of selection (Shishkin, 1984b).

The idea of genetic balance underlies chromosome‑genetics concepts of the ontogenetic (“physiological”) action of a gene (Morgan, 1937d; Lobashev, 1963; Meyer, 1968, 1974; Dubinin, 1976; etc.). Accordingly, the notion of a one‑to‑one gene‑trait relationship is formally rejected by genetic theory; it is characterized as Weismann’s error (Dubinin, 1966a, p. 238; Morgan, 1937a, b, c), as “pea‑bag genetics” reflecting early Mendelian positions (Meyer, 1968, p. 216), or finally as an incompetent opinion mistakenly ascribed to genetics (Timofeyev‑Resovsky, Ivanov, 1966, p. 116). The role of individual genes is reduced to shifting developmental pathways (Münzinger, 1967, p. 73).

However, adopting such a position obliges one to answer the question: how should the results of Mendelian analysis, the principal cognitive tool of classical genetics, be interpreted in this case? The latter, as is known, is built precisely on the unambiguous projection of genes onto traits (Schmalhausen, 1982; Stoletov, 1967; Meyer, 1968), and this explains why the principle of defining a trait by the whole genotype is sometimes called a purely theoretical declaration of genetics that does not alter its reductionist views (Svetlov, 1964). Indeed, can the two approaches be reconciled? If, for example, differences between two constant (pure‑line) phenotypes are reduced by analysis to a difference in a single gene understood as a chromosomal locus, then translating this conclusion into the language of the balance hypothesis can be done only in one way—by recognizing that the genomes of all individuals of both lines are absolutely identical except for a single segment that shifts the genetic equilibrium toward one of the two variants. But this assumption is unacceptable, because in reality the genomes of any laboratory line or natural race are always variable at many loci (cf. Kamshilov, 1939; Dubinin, 1966a). In practice these contradictions are often not even noticed, i.e., the interpretation of Mendelian analysis does not reveal a particular problem. It is noted, however, that its specific result should not be understood as revealing the entire genetic basis of a normal trait, which is always very complex. The analysis of each simple segregation uncovers only a single element of it; a complete picture is obtained by studying all possible mutational anomalies of the trait. Clearly, such an interpretation is incompatible with the balance hypothesis, because in its framework any individual’s trait always has the same genetic basis—the whole genome. Thus two conclusions follow. 1. Chromosome genetics lacks an explanation of Mendelian analysis compatible with the balance theory. 2. Reducing a monofactorial difference between two pure lines to a change in a single chromosomal locus does not reflect the actual differences in their genotypes. Hence the question is whether it is possible at all to avoid combining in genetic theory two mutually exclusive principles: “gene (group of genes) → trait” and “genome (or genotype) → trait”. Logically there is only one possibility: to recognize that the balance‑hypothesis gene (chromosomal locus) and the Mendelian‑analysis gene are not the same. In other words, the Mendelian factor must express a difference between two genotypic variants that is fundamentally irreducible to the properties or states of their individual elements. This is the conclusion we reached above, based on the theory of the epigenetic system and its corresponding real phenomenology of inheritance. A correct Mendelian segregation expresses the relationship between two system states corresponding to two alternatives of canalized development. The choice between them in hybrids is determined by the quantitative level of activity of one homologous pair of loci, which in this case acts as a “differentiator”. This role, symbolized by the Mendelian gene, is provided exclusively by the property of hybrid genotypes to realize two stable developmental variants; loss of this property eliminates the very orderliness of inheritance that allows a controlling factor to be distinguished. This conception of the essence of a gene is not new to genetics. A similar view was held by the author of the term “gene”, Johannsen (Johannsen, 1926), who understood it not as a particle of hereditary substance but as a unit of difference between two genotypes (Baur, 1922). The achievements of chromosome theory of inheritance, in the eyes of many researchers, especially phenogeneticists, did not cancel the basis of such a gene concept. Even when assigned to a chromosome, it was thought of as the cause of developmental differences that does not exist outside the properties of the integrated genotype (Promptov, 1934; Kamshilov, 1934); the assumption of an independent gene action was equated to the possibility of an “isolated bagel‑hole” (Promptov, 1934). According to Kamshilov (1935, p. 141), a gene is not a locus but a distinction between two whole genotypes, “manifested in a characteristic type of development”. Even for Morgan, recognition of chromosomal localization of the gene was accompanied by some fluctuations, and at one time it was allowed that the term referred only to an abstract property somehow linked to a given chromosomal region (Morgan, 1937g). Today, with the development of biochemical methods of genetic analysis, where a locus functions as a unit of matrix synthesis, its non‑identity with the Mendelian gene becomes increasingly evident to geneticists (Grant, 1980, p. 290; Golubovsky, 1982). Recognition of the systemic conditioning of “genic” properties of a locus was partly expressed in Fisher’s (Fisher, 1930) hypothesis on the evolution of dominance. According to this, the ability of a locus to determine a trait in the presence of a non‑identical homolog (i.e., in heterozygotes) is the result of its interaction with a system of modifiers created by selection. Thus it is acknowledged that at least the classic Mendelian phenotypic segregation ratio (3:1) is not a property of homologous chromosomal units and arises only as a product of selection‑driven reorganization of the whole genotype. From this, only a step remains to extend this conclusion to any ordered correspondence between a locus and a trait revealed by hybrid analysis. On the other hand, the first step toward recognizing that quantitative differences underlie the choice between allelic states was already made by early Mendelians, taking shape in the well‑known “presence‑absence” hypothesis of Bateson‑Pennington formulated to explain dominance (the dominant trait is determined by the presence of a factor in gametes, the recessive by its absence). Criticism from the Morgan school, pointing to incompatibility of this view with multiple allelism and the notion of linear gene arrangement on chromosomes, led to a new formulation of the hypothesis, which more explicitly stated the quantitative nature of allelic difference (the recessive factor is something that has been reduced compared with the dominant; Pennington, 1930). Finally, Goldschmidt, independently grounding the idea of “gene quantity” as the basis of allelic changes, developed it into a concept of threshold levels of morphogenetic effect that determine phenotypic choice at critical developmental points (Goldschmidt, 1927, 1938, 1940). As shown above, the presence in a chromosomal unit of Mendelian‑factor properties means that this genotype structure permits several stable developmental pathways for a trait. In the absence of such a choice, i.e., the possibility of switching stabilized trajectories, there can be no switch locus, i.e., no gene “determining” the trait. Consequently, in the general case the role of an individual locus in determining phenotypic properties remains undefined and can be described only as an element in the complex system of the whole genotype’s functioning during development. But if an individual locus has no independent expression in the phenotype and the latter in each of its traits is determined by the entire embryonic plasma (genotype), then what is the principle of particulate (discrete) inheritance that underlies chromosome theory? This question became evident to Morgan already in 1918, before the balance hypothesis, and his answer deserves attention. He concluded that if the indicated relationships between traits and embryonic plasma indeed exist, the latter nevertheless remains, in any case, built from elements independent with respect to mutation, crossing‑over, as well as homolog divergence and pairing during gamete maturation (Morgan, 1924, pp. 232, 235). “In this, and only in this sense, we may speak of a particulate structure of the plasma and of particulate inheritance” (ibid., p. 235). These statements are striking. They imply that Morgan’s theory of inheritance describes only the regularities of chromosomal element redistribution, yet does not know how they may be linked to trait inheritance! Thus it indirectly acknowledges that Mendelian rules of trait inheritance describe something entirely different from the elementary properties of chromosomal units. The lack of a method for interpreting developmental outcomes in terms of discrete chromosomal genes, implicitly recognized by Morgan, is not a consequence of incomplete knowledge, as is often assumed. It is of a principled nature. Embryological data leave no doubt that development is an indivisible epigenetic process based on the interaction of all its elementary factors and a continuous increase in qualitative diversity. Phenomena of cytoplasmic prelocalization of primordia and associated phases of pre‑formed (mosaic) development constitute only particular episodes, conditioned by epigenetic interactions in oogenesis and later replaced during development by regulatory processes (Spemann, 1925; Davidson, 1972; Svetlov, 1978). Determination of parts is defined in development only by the determination of the whole, and adult organism traits cannot have correlates in the zygote or its genome (Gurvich, 1944; Svetlov, 1964, 1978). As we have seen, any developmental outcome—normal or aberrant—is always equifinal with respect to variations in the values of its elementary causal factors. The stability of traits is “not a property of genes but an expression of interdependencies of parts in the correlation systems of the developing organism” (Schmalhausen, 1982, p. 174). This irreducibility of development to a pre‑formation model that treats the organism as a sum of consequences of independent initial causes, long understood by experimental embryology (Blyakhter et al., 1935; Belousov, 1979), is formally acknowledged also by chromosome genetics, which sees in this one of its main differences from Weismann’s embryonic‑plasma theory (Morgan, 1937a, b; Dubinin, 1966a). As already mentioned, the absence of a specific role for individual chromosomal genes (loci) in determining phenotypic properties was most convincingly demonstrated by Goldschmidt, and he was one of the few leading geneticists who clearly saw the consequences of this for developmental theory (and thus for inheritance). “Genetic facts can, of course, be described in terms of genes, but the embryonic‑plasma theory must completely abandon the concept of genes as units” (Goldschmidt, 1938, p. 311), i.e., the genotype is an indivisible basis of phenotypic development. “The embryonic plasma as a whole controls a certain reactive system, which is not a mosaic of separate effects but a unified developmental system governed as a whole by a single factor… Many geneticists find it difficult to think in such terms because most of them are so bound by an axiomatic belief in an atomistic gene theory that they cannot think otherwise; but embryologists, physiologists, and perhaps systematists will not have trouble accepting this concept” (Goldschmidt, 1940, p. 218). Goldschmidt’s predictions regarding genetics’ acceptance of these views (to which it seemed to come close) were fully justified, and the reasons for this are clear.Systemic syntheses of genetics, in which there is no room for genes as determinants of traits, belong entirely to its theoretical conceptions of individual development, i.e., that domain where it has not achieved particularly great successes. In contrast, the reductionist approach to the study of inheritance, associated with the application of Mendelian analysis, is the basis of all those huge and obvious practical achievements that have allowed genetics to claim the status of an exact science and have earned it unprecedented authority in biology in the 20th century. It is not surprising that under these conditions, when the main everyday task remains establishing links between genomic units and organismal traits, it seems inappropriate to recall that in the theory behind genes such a deterministic property is not recognized and that they should be attributed only to the genotype. The very obviousness of the discrete connections identified and the possibility of their experimental verification usually prevent reflection on what their true nature is and under what conditions they arise. The property of a locus to switch developmental pathways when crossing two phenotypes stabilized by selection is absolutized as its constant property, and its switch function, revealed only in hybrids, is regarded as proof of its special deterministic role with respect to a given trait. One of the most eloquent testimonies to the triumph of this reductionist style of thinking is the textbook assertion that the principle of “gamete purity” itself ensures the discreteness (non‑mixing) of trait inheritance, i.e., that the laws of chromosomal homolog segregation must always be directly reflected in phenotypic properties, regardless of developmental stability. The need to tie these ideas to the systemic requirements of the balance hypothesis in practice results in a “compromise” solution, i.e., the concept of a genotypic environment. The participation of the whole genotype in trait realization is understood here as the interaction of the “principal” gene (or genes) determining the trait with other hereditary elements that modify its specific action and play the role of “noise” relative to it. This understanding of gene‑trait relations has become entrenched in the minds of many researchers as typical for chromosomal genetics, regardless of whether they reject it (Goldschmidt, 1940) or consider it acceptable (e.g., Astaurov, 1971, p. 218; Bertalanffy, 1969, p. 73). It is quite evident that this interpretation remains entirely within the pre‑formist view of discrete (mosaic) trait determination. In essence it differs little from Weismann’s concept, in which the possibility of each determinant’s expression depended in the same way on the overall outcome of the struggle among hereditary elements. This seemingly constantly rejected yet in practice deeply rooted belief in the independent determination of traits by chromosomal genes is indirectly detected even where the existence of special holistic properties of the species‑level genotype and its controlled epigenetic system, unaltered by simple mutations, is verbally emphasized. For when the comparison of such genotypes is undertaken, their differences are actually assessed not in holistic properties but only in the number of non‑identical genes (Dobzhansky, 1947, pp. 106, 110, 337, 338; Meyer, 1968, p. 432). The essence of the genotype as a whole, expressed only through individual development (Kamsilov, 1934), still reduces here to the sum of chromosomal elements. * * * Historically, the emergence of each new natural science field usually begins with probing the object of study by analytical methods and, as a consequence, with the dominance of a reductionist approach to its description; the time of systemic syntheses comes later. This was also the case in the history of traditional experimental branches of biology—developmental mechanics and genetics, the latter having passed from a typical “mechanics of inheritance” (Weismann and early Mendelians) to the systemic constructions of Goldschmidt, Schmalhausen, and Waddington. It is therefore natural that the appearance and rapid development of a new research front—molecular genetics, studying the structure and function of cellular matrix structures—again led to a revival of reductionist notions of gene‑trait linkage, now expressed in the formula “one gene—one enzyme”. This introduced a new element of dualism into the worldview of classical genetics; many of its representatives, while denying direct deterministic influence of genes on phenotypic traits in theory, simultaneously had to admit its possibility regarding primary gene products (Timofeyev‑Resovsky, Ivanov, 1966; Waddington, 1966); the latter (primarily proteins) are regarded as traits that most adequately reflect their genetic basis (Gershenzon, 1974). All this could not promote the strengthening of systemic views in classical genetics, whose influence was already limited. Yet in molecular biology new notes begin to sound. The long‑standing opinion that the “gene‑enzyme” principle represents an extreme simplification of real relationships (Haldane, 1954) now finds increasing confirmation. Phenomena of enzyme polymorphism and the similarity of the spatial structure of their homologs across different organisms already themselves suggest that this structure plays the main role in determining protein functional properties, whereas the amino‑acid sequences that compose it (at least in regions outside active sites) may vary. Thus, there is degeneracy not only between the DNA nucleotide code and the linear protein structure, but also between the latter and the nature of its function (Waddington, 1970b; Volkenshtein, 1981a, b). More specialized studies show that the entire multistep process leading from a cistron to an enzyme—transcription—processing (post‑transcriptional RNA reduction)—translation and post‑translational modifications—is polyvariant at each of its stages (Inge‑Vechomov, 1976; Mikhailova, Simarov et al., 1981). One vivid manifestation of this polyvariancy is the fact that, normally, from the same DNA segment different transcripts can be read in different tissues or at different ontogenetic stages, producing different proteins (Golubovsky, 1985). Overall, such ambiguity can be caused by influences of very different nature, creating on the one hand the possibility of deviation of the final synthesis product from the norm, and on the other—providing a basis for its regulation (“phenotypic suppression” of mutations) when coding errors occur. Such regulation, in particular, has been shown in yeast‑saccharomycetes for post‑transcriptional synthesis stages, where it can be induced by changes in temperature, pH, osmotic pressure, and the action of other external agents (streptomycin, glycerol, etc.). These processes prevent losses in the polypeptide chain structure and facilitate its functional regulation at the post‑translational level (Mikhailova, Simarov et al., 1981). The latter can be experimentally induced by the same agents as post‑transcriptional regulation, but in many cases it occurs as a natural process that preserves normal biochemical function in mutants. Often, rather than correcting errors in amino‑acid sequences, their effect is neutralized. The most well‑known mechanism of this regulation, studied in various organisms, is the interaction of a mutant locus product with normal or mutant subunits of the same protein (“inter‑allelic complementation”). It either compensates for losses in the primary structure of one polypeptide chain by others, or restores the function of the whole through conformational changes of mutant subunits, i.e., by assembling available fragments into a complex that acquires a normal spatial structure by self‑assembly. Amino‑acid changes outside active sites do not disrupt normal function (Inge‑Vechomov, Soidl, 1978). Phenomena of complementation of synthesis products that break the colinear correspondence between cistron and enzyme shed light on the evolutionary significance of post‑translational changes in general. The fact that protein active sites retain relative autonomy and often can function independently of their connections (e.g., two non‑identical subunits of tryptophan synthase, separated in *Escherichia coli* but encoded as a single unit in *Neurospora*) leads researchers to suppose that many multifunctional proteins could have arisen by the fusion of enzymes encoded by different genes, and conversely that multi‑enzyme complexes in some cases are products of epigenetic fragmentation of multifunctional precursor proteins, as directly observed in certain viruses (Inge‑Vechomov, Soidl, 1978). Here the “gene—enzyme” rule no longer applies. The necessity to consider possible effects of post‑translational modifications becomes increasingly evident, prompting avoidance of direct identification of rows of isoenzymes with products of isoallelic mutations (Solbrig, Solbrig, 1982, p. 256). All these facts show that, despite the great distance separating locus and phenotype in classical versus molecular genetics, their relationships are fundamentally similar. The functional stability of the final synthesis product cannot be reduced here to the stability of the matrices themselves (DNA and RNA). It is based on regulatory epigenetic interactions that encompass the entire synthesis system and can buffer certain genetic code errors as well as disturbances of transcription and translation processes. As in “macro‑ontogeny”, the regulatory capacities of this system are sensitive to external factors. The synthesized molecule is thus equally a product of the cistron as of all elements of the synthesis system that do not depend on it (enzymes, RNA, etc.), and it need not be colinear with the cistron. All this suggests that the evolutionary mechanism of protein formation is essentially the same as for other elements of the adaptive norm. A protein molecule also has a “phenotype”, partially determined by the environment, and its correspondence to the substrate of its action must be historically stabilized (Waddington, 1970b). Evidently, the first step in the evolutionary emergence of a new type of protein must be its appearance as one of the unstable post‑translational modifications of an existing enzyme. If the latter proves adaptively valuable under new conditions, its realization path is transformed by selection toward maximal robustness. This is expressed primarily in the gradual straightening and simplification of the entire sequence of synthesis stages, i.e., in increasing colinearity between the new molecule and its original matrix. From this viewpoint, cases of ordered post‑translational transformations of normal proteins can be interpreted as intermediate stages of stabilizing their morphogenesis, “recapitulating” the course of initially unstable transformations. Examples include, in particular, the formation of insulin in mammals by proteolysis of a giant precursor molecule and its division into two subunits, or the formation of the major protein in phage T4, where about 900 identical polypeptide chains first spontaneously aggregate and then cleave off N‑terminal segments (Stent, Keldin, 1981). Probably, the excision of inactive (intronic) sequences from messenger RNA, a usual step of protein synthesis, reflects transformations that once occurred at the post‑translational level. All this leads to the view that the DNA matrix underlying a stably synthesized protein is historically not the cause of its emergence, but rather the consequence of stabilizing its morphogenesis, originally based on modification of a product encoded by another matrix variant. If this is true, we must conclude that linking the historical appearance of new homologous protein subunits, e.g., β‑globin, to duplication of the corresponding gene (as is commonly done) amounts to swapping cause and effect; ontogenetic and historical causality are not identical. A partially similar conclusion was reached by Y. M. Olenev (1977), who noted that the need for a new product arises before the corresponding gene is duplicated, and that this problem in evolution is first solved by circumvention. The possibility that functional expansion of a gene may precede its duplication is also recognized by T. R. Soidl (1983).

The need to tie these ideas to the systemic requirements of the balance hypothesis in practice results in a “compromise” solution, i.e., the concept of a genotypic environment. The participation of the whole genotype in trait realization is understood here as the interaction of the “principal” gene (or genes) determining the trait with other hereditary elements that modify its specific action and play the role of “noise” relative to it. This understanding of gene‑trait relations has become entrenched in the minds of many researchers as typical for chromosomal genetics, regardless of whether they reject it (Goldschmidt, 1940) or consider it acceptable (e.g., Astaurov, 1971, p. 218; Bertalanffy, 1969, p. 73). It is quite evident that this interpretation remains entirely within the pre‑formist view of discrete (mosaic) trait determination. In essence it differs little from Weismann’s concept, in which the possibility of each determinant’s expression depended in the same way on the overall outcome of the struggle among hereditary elements. This seemingly constantly rejected yet in practice deeply rooted belief in the independent determination of traits by chromosomal genes is indirectly detected even where the existence of special holistic properties of the species‑level genotype and its controlled epigenetic system, unaltered by simple mutations, is verbally emphasized. For when the comparison of such genotypes is undertaken, their differences are actually assessed not in holistic properties but only in the number of non‑identical genes (Dobzhansky, 1947, pp. 106, 110, 337, 338; Meyer, 1968, p. 432). The essence of the genotype as a whole, expressed only through individual development (Kamsilov, 1934), still reduces here to the sum of chromosomal elements.

* * *

Thus, the matrix‑based synthesis mechanism can be described at its core as a system regulated toward a certain final state and, evidently, transformed in evolution “top‑down” (from the final product to the matrix), i.e., in accordance with the theory of stabilizing selection. One may agree with B. M. Mednikov (oral communication), who considers that a protein is essentially an organ, to the development of which all regularities applicable to ordinary organs apply. One more consideration should be added. The effect of biochemical mutations can be phenocopied (Goldschmidt, 1955), and if this rule is as universal as for ordinary phenotypes, it follows that any synthesis product obtained on the basis of a single mutation of a given genotype could in principle be obtained on the same basis without mutation, by selecting external influences that disrupt transcription, translation, or later epigenesis in a certain way. Such cases of induced synthesis are usually explained by derepression of an operon mechanism containing the necessary matrix, according to the scheme of Jacob and Monod (Waddington, 1964; Volkenshtein, 1981 6); but in light of the considerations above, it is permissible to assume that this phenomenon has a more general basis—the limitation of the spectrum of potential primary‑synthesis possibilities allowed for a given genotype as a whole species system under all its elementary mutational changes. EPIGENETIC THEORY OF EVOLUTION Above it was emphasized that the question of the origin of normal (adaptive) organization, which constitutes the main problem of evolutionary theory, reduces to explaining how new organismal properties acquire stability and become irreversible. In general form this explanation is provided by the concept of an epigenetic system, within which a simple evolutionary change means a transition of individual development onto one of the aberrant trajectories with subsequent transformation into a canalized developmental pathway (creod). The theory of stabilizing (canalizing) selection, which treats such a course of events as the basis of the entire evolutionary process, may be called the epigenetic theory of evolution (Shishkin, 1984a). This name is justified, first, because the initiating factor of evolution here is recognized as disturbances in the course of ontogeny. Second, it avoids the erroneous notion that stabilizing selection is merely one of the particular evolutionary mechanisms alongside “directional”, “disruptive”, and other forms of selection, and that the Schmalhausen‑Waddington doctrine simply constitutes a section of synthetic theory dealing with changes in a stable or fluctuating environment. Although the differences of the epigenetic theory from traditional views were highlighted by its authors, they were never systematized, and its exposition remained largely cluttered with foreign concepts. Therefore, it is necessary first to briefly characterize its content. The epigenetic theory is based on the notion of an adaptive norm, or typical organization, as the object of evolutionary change. Adaptiveness, or the purposiveness of organization, is understood as its capacity for self‑maintenance and self‑reproduction (inheritance), i.e., its stability. The latter, in turn, is expressed in the ability of normal individual development to relax within wide limits against external and internal perturbations on the way to achieving the given adult organization. Explaining this property of development must constitute a key task of evolutionary theory. The epigenetic concept solves this problem by assuming that species‑specific individual development is a holistic dynamic system with a limited and structured space of possible final states, among which the normal (adaptive) outcome corresponds to system equilibrium, and the entire domain of potential phenotypic deviations represents its more or less unstable fluctuations. Any deviation of the developmental outcome from equilibrium in response to a damaging influence always represents a systemic reaction, the choice of which is determined not by the specificity of the damaging factor (e.g., type of mutation) but only by the magnitude, location, and timing of the disturbance it introduces into the developmental trajectory. The totality of possible system states (Fig. 1, 7), or its phase space (the epigenetic landscape), constitutes its holistic characteristic, defined by the overall organization (genotype) of the embryonic cell and indivisible into independent effects of any elementary factors acting within this cell or its ontogenetic derivatives (Gurvich, 1944; Goldschmidt, 1940; Waddington, 1957). In other words, possible alternative states of the species phenotype are parameters of the developmental system, while the states of its individual interacting elements (starting from genome structural elements) are dynamic variables characterizing the lower hierarchical level of the system. The enormous variety of combinations of elementary component values (varying in each individual developmental cycle) corresponds at the higher level to a limited space of phenotypic parameter variations, i.e., the latter form an invariant (equifinal) set relative to this diversity.Moreover, the higher the relative persistence (probability of occurrence) of a given phenotypic aberration, the broader the spectrum of possible lower‑level states within which it can be realized. The normal phenotype, corresponding to the system’s equilibrium, by degree of equipotentiality sharply exceeds all others, i.e., under normal conditions it is realized by the overwhelming majority of individual zygote variants existing in natural populations.

The phenomena of complementation of synthesis products, which violate the collinear correspondence between a cistron and an enzyme, shed light on the evolutionary significance of post-translational modifications in general. The fact that the active centers of proteins retain relative autonomy and can often function independently of their connection (e.g., two non-identical subunits of tryptophan synthetase, separated in E. coli but encoded as a whole in Neurospora) leads researchers to believe that many multifunctional proteins could arise by combining enzymes encoded by different genes, and conversely, that multi-enzyme complexes in some cases are products of epigenetic splitting of multifunctional precursor proteins, as directly observed in some viruses (Inge-Vechtomov, Soydla, 1978). Here, the 'gene-enzyme' rule is no longer needed. The necessity of considering the possible effect of post-translational modifications is becoming increasingly evident to researchers, forcing them to refrain from directly equating rows of isoenzymes with products of isoallelic mutations (Solbrig, Solbrig, 1982, p. 256).

All these facts show that, despite the great distance separating locus and phenotype in classical versus molecular genetics, their relationships are fundamentally similar. The functional stability of the final synthesis product cannot be reduced here to the stability of the matrices themselves (DNA and RNA). It is based on regulatory epigenetic interactions that encompass the entire synthesis system and can buffer certain genetic code errors as well as disturbances of transcription and translation processes. As in “macro‑ontogeny”, the regulatory capacities of this system are sensitive to external factors. The synthesized molecule is thus equally a product of the cistron as of all elements of the synthesis system that do not depend on it (enzymes, RNA, etc.), and it need not be colinear with the cistron. All this suggests that the evolutionary mechanism of protein formation is essentially the same as for other elements of the adaptive norm. A protein molecule also has a “phenotype”, partially determined by the environment, and its correspondence to the substrate of its action must be historically stabilized (Waddington, 1970b). Evidently, the first step in the evolutionary emergence of a new type of protein must be its appearance as one of the unstable post‑translational modifications of an existing enzyme. If the latter proves adaptively valuable under new conditions, its realization path is transformed by selection toward maximal robustness. This is expressed primarily in the gradual straightening and simplification of the entire sequence of synthesis stages, i.e., in increasing colinearity between the new molecule and its original matrix. From this viewpoint, cases of ordered post‑translational transformations of normal proteins can be interpreted as intermediate stages of stabilizing their morphogenesis, “recapitulating” the course of initially unstable transformations. Examples include, in particular, the formation of insulin in mammals by proteolysis of a giant precursor molecule and its division into two subunits, or the formation of the major protein in phage T4, where about 900 identical polypeptide chains first spontaneously aggregate and then cleave off N‑terminal segments (Stent, Keldin, 1981). Probably, the excision of inactive (intronic) sequences from messenger RNA, a usual step of protein synthesis, reflects transformations that once occurred at the post‑translational level. All this leads to the view that the DNA matrix underlying a stably synthesized protein is historically not the cause of its emergence, but rather the consequence of stabilizing its morphogenesis, originally based on modification of a product encoded by another matrix variant. If this is true, we must conclude that linking the historical appearance of new homologous protein subunits, e.g., β‑globin, to duplication of the corresponding gene (as is commonly done) amounts to swapping cause and effect; ontogenetic and historical causality are not identical. A partially similar conclusion was reached by Y. M. Olenev (1977), who noted that the need for a new product arises before the corresponding gene is duplicated, and that this problem in evolution is first solved by circumvention. The possibility that functional expansion of a gene may precede its duplication is also recognized by T. R. Soidl (1983).

Thus, the mechanism of template synthesis can be described at its core as a system regulated to a certain final state and, apparently, transformed in evolution 'from top to bottom' (from the final product to the template), i.e., in accordance with the theory of stabilizing selection. One can agree with B. M. Mednikov (personal communication), who believes that a protein is essentially an organ, to the development of which all regularities applicable to ordinary organs apply.

One more consideration should be added. The effect of biochemical mutations can be phenocopied (Goldschmidt, 1955), and if this rule is as universal as for ordinary phenotypes, then it must be concluded that any synthesis product obtained based on a single mutation of a given genotype can in principle be obtained on the same basis and without mutation, by selecting external influences that somehow disrupt transcription, translation, or late epigenesis. Usually, such cases of induced synthesis are explained by the depression of the operon mechanism containing the necessary template, according to the Jacob and Monod scheme (Waddington, 1964; Volkenstein, 1981); but in light of the considerations set forth above, it is permissible to assume that this phenomenon has a more general basis – the limitation of the spectrum of potential possibilities of primary synthesis permissible for a given genotype as a whole species system under all its elementary mutational changes.

EPIGENETIC THEORY OF EVOLUTION It was emphasized above that the question of the origin of normal-adaptive organization, which constitutes the main problem of evolutionary theory, boils down to explaining how new properties of organisms acquire stability, becoming irreversible. This explanation is given in a general form by the concept of the epigenetic system, within which an elementary evolutionary change means the transition of individual development to one of the aberrant trajectories with its subsequent transformation into a canalized developmental pathway (creode). The theory of stabilizing (canalizing) selection, which considers such a course of events as the basis of the entire evolutionary process, can be called the epigenetic theory of evolution (Shishkin, 1984a). This name is justified, firstly, because violations of the course of ontogenesis are recognized as the initiating factor of evolution. Secondly, it allows avoiding the erroneous notion that stabilizing selection is merely one of the partial evolutionary mechanisms alongside 'leading,' 'disruptive,' and other forms of selection, and that the teachings of Shmalhausen-Waddington constitute simply a section of the synthetic theory dealing with changes in a stable or fluctuating environment. Although the authors of the epigenetic theory emphasized its differences from traditional views, they never generalized it, and its exposition remained largely contaminated with concepts alien to it. Therefore, it is necessary to briefly characterize its content first.

The epigenetic theory is based on the idea of the adaptive norm, or typical organization, as the object of evolutionary changes. By adaptiveness, or expediency of organization, is understood its ability for self-maintenance and self-reproduction (inheritance), i.e., its stability. The latter, in turn, is expressed in the ability of normal individual development to relax external and internal disturbances within a wide range on the path to achieving a given mature organization. The explanation of this property of development should be the key task of evolutionary theory.

The epigenetic concept resolves this problem by assuming that species-specific individual development is a holistic dynamic system with a limited and structured space of possible final states, among which the normal (adaptive) outcome corresponds to the system's equilibrium, and the entire area of potential phenotypic deviations corresponds to its more or less unstable fluctuations. Any deviation of the developmental outcome from equilibrium in response to a damaging influence is always a systemic reaction, the choice of which is determined not by the specificity of the damaging factor (e.g., type of mutation), but only by the degree, place, and time of the disturbance introduced by it into the course of development.

This transformation of the norm by selection essentially represents the expression of the general ability of systemic objects to relax disturbances induced in them, i.e., to change purposefully. Restoration of equilibrium, or “goal seeking” (Ashby, 1962), is carried out by the system through successive correction of its state, leading to the attenuation of the initial disturbance. It is precisely such a situation, but only linked to a qualitative change of the system itself, that arises during an evolutionary shift of the norm. When a population moves into extreme conditions, its developmental system destabilizes and, instead of realizing the previous norm, shifts to random and unstable individual fluctuations. The further survival of the system in the new conditions depends on whether it can stabilize in one of these variable states. This search for a new equilibrium is carried out by the system through preferential preservation (selection) of individual developmental variants that realize the most viable fluctuation. The selection process here is nothing other than a chain of damped correction cycles with feedback, leading to the stabilization of a new norm. Each act of selection preserving carriers of an adaptively valuable aberration shifts the system’s state toward a future equilibrium; “noise” in reproducing this phenotype in the next generation signifies a new deviation from equilibrium; the next sieving act again shifts the population’s appearance toward the future norm, and so on, until each newly reproduced generation becomes phenotypically homogeneous and similar to the parent. The adaptively valuable change becomes a stable characteristic of the system. All this shows that the creative role of selection, like that of any creative process, ultimately lies in “recording a random choice” (Kastler, 1967), which in this case is expressed in the choice of one of the relatively equally probable fluctuations of the developmental system and its conversion into a stably realized new norm. The latter, throughout its formation, plays the role of a “goal” that determines the direction of correction of the system’s properties during its transformation by selection. These ideas differ fundamentally from the traditional interpretation of selection as a mechanism of sieving and combining “hereditary changes,” understood as specific effects of particular genes and their combinations. Under such an approach, explaining the inheritance of evolutionary novelties becomes superfluous, because they are seen simply as an immanent property of the corresponding genes, independent of selection. The whole procedure of “creating” an elementary change by selection is essentially equated here to a single‑act choice of its causal factor, which should be followed by automatic reproduction of the new trait in subsequent generations. In contrast, for the epigenetic theory, the stability of reproduction (heritability) is precisely what requires explanation on the basis of the principle of natural selection. The stages of elementary shift of the adaptive norm, reflecting the restructuring of the species‑specific developmental system (blurring of the old and stabilization of the new equilibrium trajectory and the associated transformation of the system’s phase space) were specifically considered above (Fig. 7). But this process can also be described somewhat differently—from the viewpoint of changes undergone by the set of individual ontogenetic cycles belonging to the system (Fig. 10). [IMG_10] Fig. 10. Transformation of individual modification spectra during an elementary shift of the adaptive norm 1 — relationship between zygotes and realizable phenotypes; 2 — modification spectra characterizing a series of zygotes within the same interval of conditions. I — stable (equipotential) realization by a group of zygotes of the original adaptive norm A in the moderately permissible conditions N; II — destabilized development of the same zygotes in the interval of extreme conditions N₁, where among the realized deviations (b), c, d, k, t, z the morph (b) has an adaptive advantage; III — stabilization of morph (b) and its conversion into adaptive modification B within the new polymorphic norm AB; IV — further stabilization of phenotype B against the loss of the previous norm or its retention as an unstable morph (a). F₁ — F₂ — Fₙ — Fₓ — generational succession; hatched areas denote realized development types; their eliminable variants are shown by descending arrows. F₁ — F₂ — exposure of variability; F₂ — Fₙ, Fₙ — Fₓ — stages of stabilizing selection.Since the probabilities of realizing the same developmental pathways for different zygotes of a species are always unequal due to their genetic differences, each zygote, within any specific range of conditions that go beyond the usual, exhibits its own spectrum of phenotypic deviations (morphs). From this perspective, a population of normal zygotes can be represented as a series of heterogeneous modification spectra, in which the adaptive phenotype always constitutes the main part, while the marginal segments consist of various combinations of morphs (Fig. 4; 10, I). Under normal conditions all zygotes develop equifinally, realizing the norm (Fig. 10, I). As the environment shifts toward a critical threshold, canalized development is replaced by divergent development, i.e., an increasing number of morphs arise in accordance with the specifics of individual spectra (Fig. 10, II). When such conditions persist over successive generations, selection in favor of the most viable morph begins, leading to a gradual increase in its stability and to the destabilization of the former norm (Fig. 10, III). Elimination of the other types of aberrant reactions is initially inefficient, because they reappear in the offspring of the selected variant due to its weak ontogenetic stability. However, as the selected phenotype stabilizes, its inheritance becomes increasingly unambiguous, and it (in the case of xenogamous reproduction) progressively absorbs the remaining aberrations in crosses, which stay unstable. This fixation of an adaptive chain reaction, turning it into a new norm, causes the modification spectra of successive zygote generations to allocate it an ever larger share at the expense of the old norm. Thus, both norms coexist at a certain stage within individual spectra as two adaptive modifications, expressed depending on fluctuations in conditions (Fig. 10, III), until finally the new one completely prevails. The possibility of expressing the former normal phenotype in the newly established conditions does not disappear entirely; it is reduced to the level of an aberrant atavistic variation (Fig. 10, IV). Consequently, each elementary step in the selective transformation of normal organization reduces to an extreme divergent modification (destabilization) of the existing norm followed by the fixation of the most optimal of the emerging individual response variants. Modification and stabilization “continuously cooperate” (Shmälgauen, 1968 6, p. 315). The first phase of this step signifies the individualization of phenotypic expression of separate genomes (Fig. 10, I–II), the second – its unification, i.e., the creation of a mechanism of canalized development of a new adaptation that levels private genetic differences (Fig. 10, III–IV). The first of these phases is a period of instability separating each two successive stable states of the norm. All this allows us to understand that, when speaking of the driving function of selection, we refer to the final result of this process, not to its mechanism. The primary disruption of the former norm, as we have seen, is not caused by selection at all but by a disturbance of developmental conditions leading to the expression of hidden variability; the creation of a new norm does not reduce to merely choosing the optimal variant of change. Selection cannot directly preserve unstable reactions that constitute a hidden reservoir of variability. The shift of the norm in favor of one of them is achieved not by the direct elimination of the others (since they initially reappear generation after generation), but by the stabilization of the optimal reaction, which consequently becomes increasingly universal for the whole set of developing individuals. Thus, a “driving selection” as a distinct process relative to stabilizing selection does not exist. This concept describes the overall outcome of a long series of alternating phases of reservoir exposure and stabilization of particular change variants (Shishkin, 1984a, b). If, according to Darwin (1952, p. 139), natural selection is “the preservation of useful individual variations,” this simply means their acquisition of ontogenetic stability. Shmälgauen’s (1968 b) note on the inseparable link between the driving and stabilizing forms of selection (unfortunately left unspecified) requires clarification. It concerns not only two sides but two qualitatively and scale‑different dimensions of a single process. The conviction that driving selection is indeed based on the stabilization of private aberrations of the norm is supported by a wide range of observations on its transformation under various natural and experimental conditions. The most stringent confirmation comes from the genetic assimilation experiments of structural morphs discussed above (Waddington, 1957), because there is no doubt about the extremely unstable nature of the original class of reactions that form the new norm. Another important category of facts consists of numerous experiments on the assimilation of physiological morphs, i.e., on forced selective adaptation of various organism groups (usually insects) to new environmental factors, for example, the experiments of M. M. Kamschilov (1941) on breeding cold resistance in Drosophila, I. V. Kozhanchikov (1941) on rearing leaf‑eating beetles on an unfamiliar diet, similar experiments by G. V. Samokhvalova (1951, 1954) and especially G. X. Shaposhnikov (1961, 1965) on host switching in aphids, etc. [The observed development of new stable adaptations is often attributed to “long‑term modifications” when parthenogenetic clones are used (e.g., in several aphid experiments), because they are considered genetically homogeneous. Yet the richness of their individual responses in extreme conditions and the subsequent course of transformation clearly indicate hidden heterogeneity. This heterogeneity arises from genome replication errors, which are inevitable in any mode of germ cell formation.] The first reaction to a sharp change in conditions is usually high mortality or reduced fecundity and, undoubtedly, an extreme physiological state of the surviving individuals. During selection for the stability of this state, it transforms from a sublethal condition into a normal or even optimal one. It is evident that this tolerance to a previously harmful factor emerges here as a new quality created by selection and did not exist within the former norm as a ready‑made hereditary variation. For example, in cooling experiments with Drosophila larvae the original population contained no individuals that were stably resistant to cold or that preferred moderately reduced developmental temperature (Kamschilov, 1941, 1979), and in aphid Dysaphis anthrisci majkopica experiments, even after 8–11 generations of rearing on a new host, no individuals preferred the old host (Shaposhnikov, 1961, 1965). The sequence of changes observed in such cases (particularly in the Dysaphis experiments) aligns well with the theoretically expected course of events. Initially, a phenotypically homogeneous population placed in unusual conditions shows a sharp increase in variability, including differential fecundity and viability. This is the phase of exposing the mobilizable reservoir of variability (Fig. 10, I–II). Then a growing unification of phenotypes in favor of a new adaptive norm occurs, with the first several generations showing no preference for the old or the new set of conditions. This is the transitional phase of coexistence of both norms within the same individual spectra as two adaptive modifications (Fig. 10, III). Finally, the new norm becomes fully stabilized as the sole adaptation (cf. Fig. 10, IV). Its separation from carriers of the former norm can reach the level of reproductive isolation (Shaposhnikov, 1978). A huge number of natural analogues of these transformations are known, most of them relating to cases of “acclimation” of various insect forms to pesticides. The genetic nature of the resistance that arises can differ in different cases (Dubinin, 1966a). As a result, new races appear that, when crossed with the original ones, show different types of segregation (Dobzhansky, 1947). There is no reason to think that their formation process differs from the one described above or that the new trait existed in a stable form in single mutants before selection began. Meanwhile, as noted earlier, this very assumption underlies the dominant views on the mechanism of driving selection. It is seen merely as a process of accumulating mutations with inheritable phenotypic effects, i.e., the latter is considered not created by selection but only captured by it. In particular, the emergence of a pesticide‑resistant insect race, if it differs from the original by a single Mendelian factor, is taken as an obvious result of the spread of a single resistance mutation (Dobzhansky, 1947, p. 190). Without addressing the theoretical side of these views, we will consider their correspondence with observed facts. First, they contradict experimental genetics, which shows that the effects of “raw” mutations, especially small ones (constituting the potential material of evolution), are unstable compared with the adaptive norm and are, to varying degrees, absorbed by it. Second, the nature of the evidence used deserves attention. Such evidence usually reduces to demonstrating changes in the frequency of a particular trait in different populations, races, or other groups, tracked either spatially (variations in body coloration, wing venation, etc.; Zimmermann, 1933; Timofeyev‑Resovsky et al., 1965) or even over geological time (e.g., changes in tooth morphotypes in mammalian evolution; Simpson, 1953). It is assumed that minimal frequencies of a trait correspond to its existence as mutations, while higher frequencies indicate its incorporation into the norm. But if no difference in ontogenetic stability is presumed between the mutant phenotype and its adaptive analog, we are pre‑emptively equating the act of mutation with the creation of a basic adaptation rather than proving it. In reality, the task is precisely to determine what the actual heritability of the “mutant” aberration was before its expansion and what its origin was. Was it always stable, or did it initially represent a labile reaction that was gradually stabilized and amplified by selection? In most cases we cannot verify this, and when researchers speak of the spread of “mutations,” they usually refer to established adaptive ecotypes (as, for example, in the case of melanism in hamsters; Gershenson, 1946), within which hereditary variability is typically observed (Dubinin, 1966a, p. 280). Where the prehistory of such forms is at least partly known, we find no evidence of their original stability. An example is the evolution of industrial melanism in the peppered moth Biston betularia, which seems to be a textbook case of “incorporating a useful mutation into the norm.” The modern melanic morph of this butterfly in England, dominant in most populations, yields monofactorial segregation with the light morph. Yet neither of these properties is historically primary. Earlier melanists (presumed hybrids) caught a century and a half ago were lighter than modern ones, i.e., dominance was incomplete, even though dark coloration already contributed to a polymorphic norm, conferring an adaptive advantage on dark‑barked trees (Kettlewell, 1956). Neither dominance nor proper segregation is observed when crossing the English dark Biston morph with the light Canadian morph (where melanists are absent); instead, intermediate inheritance occurs (Sheppard, 1970). Dominance is an expression of phenotypic stability (Shmälgauen, 1982, 1968b), and it was undoubtedly reinforced during the formation of the dark Biston morph. Extrapolating from the two known successive states (in Europe) to the initial moment of its adaptive history, we can reasonably assume that the original material consisted of weak deviations from the light coloration, sensitive to external fluctuations and the genetic constitution of individuals—i.e., morphs of the norm that selection stabilized and amplified step by step. The potential for such melanic reactions is widespread among butterflies (Dubinin, 1966a) and can be induced artificially, for example, by cooling caterpillars (Standfuss, 1902). Clearly, there is no basis for attributing the properties of modern normal melanists to elementary mutations with stable effects that arose in their ancestors.

Any ordering of biological phenomena must be regarded as a product of natural selection (Mayer, 1981), and regarding Mendelian inheritance rules this connection is entirely evident. The ordering of developmental outcomes here rests on the presence in the parental epigenetic system of the corresponding creodes, which can arise only through stabilizing selection. Genomic mutations, like other singular perturbations of the system, can, as shown, only disrupt creodes, but do not themselves create robust developmental pathways. Accordingly, the expression “raw” mutational anomalies, both natural and experimental, is highly variable and generally poorly conforms to Mendelian rules (Dubinin et al., 1937; Gershenzon, 1941; Kamshilov, 1940; Schmalhausen, 19686), leading to the concepts of a gene’s genotypic environment, as well as expressivity and penetrance of a mutant trait. Only the effects of the largest genomic disruptions show relatively little dependence on genotype variations, i.e., such mutations fairly deterministically produce a particular type of anomaly; yet this correspondence is never completely stable [The latter, in particular, concerns large genetic anomalies in humans, which, even when dominant, may sometimes not manifest in the homozygote and generally show variable expression (e.g., polydactyly; Gershkovich, 1968)].

All this allows us to realize that when we speak of the moving function of selection, we mean the final result of this process, not its mechanism. The initial disruption of the previous norm, as we have seen, is not related to selection at all, but to a disruption of developmental conditions leading to the manifestation of hidden variability; the creation of a new norm is not reduced solely to the selection of the optimal variant of change. Selection cannot directly preserve unstable reactions that constitute a hidden reserve of variability. The shift of the norm in favor of one of them is achieved not by direct elimination of others (because initially they reappear in generations again and again), but by stabilizing the optimal reaction, which consequently becomes increasingly universal for the entire population of developing individuals. Thus, moving selection as a separate process does not exist relative to stabilizing selection. This concept describes the overall outcome of a long series of alternating phases of revealing the reserve of variability and stabilizing certain variants of changes (Shishkin, 1984a, b). If, according to Darwin (1952, p. 139), natural selection is 'the preservation of useful individual variations,' then this means nothing other than their acquisition of ontogenetic stability. Shmalkhausen's (1968b) remark about the inseparable connection between moving and stabilizing forms of selection (unfortunately, left unspecified) needs clarification. It is not just about two sides, but about two qualitatively and quantitatively different dimensions of the same process.

The fact that directional selection is indeed based on the stabilization of partial aberrations of the norm is confirmed by a wide range of observations of its transformation in various natural and experimental conditions. The most rigorous confirmation comes from the experiments with genetic assimilation of structural morphoses (Waddington, 1957) discussed above, as there is no doubt about the extremely unstable nature of the initial reaction range that forms the new norm. Another important category of facts includes numerous experiments on the assimilation of physiological morphoses, i.e., the forced selective adaptation of various groups of organisms (usually insects) to new environmental factors, for example, M. M. Kamshilov's (1941) experiments on developing cold resistance in Drosophila, I. V. Kozhanchikov's (1941) experiments on raising leaf beetles on unusual food, similar experiments by G. V. Samokhvalov (1951, 1954), and especially G. Kh. Shaposhnikov's (1961, 1965) experiments on host change in aphids, etc. [The development of new stable adaptations observed in this process is often attributed to "long-term modifications" when parthenogenetic clones are used (e.g., in several aphid experiments), as they are considered genetically homogeneous. However, the diversity of their individual reactions under extreme conditions and the entire subsequent course of transformation clearly indicate hidden heterogeneity. The latter arises due to errors in genome replication, which are inevitable in any way of forming generative cells].

The first reaction to a sharp change in conditions is usually high mortality or reduced fertility, and undoubtedly, an extreme physiological state of the surviving individuals. Through selection for resistance to this state, it transforms from sublethal to normal or even optimal. It is quite obvious that this tolerance to a previously harmful factor arises here as a new quality created by selection and not existing within the previous norm as a ready-made hereditary variation. For example, in experiments with cooling Drosophila larvae, the initial population had no individuals resistant to cold or preferring moderately low temperatures for development (Kamshilov, 1941, 1979), and in experiments with the aphid Dysaphis anthrisci majkopica, after 8–11 generations on a new host, there were no individuals that preferred the old one (Shaposhnikov, 1961, 1965). The sequence of changes observed in similar cases (particularly in experiments with Dysaphis) corresponds well to the theoretically expected course of events. Initially, a phenotypically homogeneous population exposed to unusual conditions shows a sharp increase in variability, including differential fertility and viability. This is the phase of revealing the mobilization reserve of variability (Fig. 10, I—II). Next, increasing unification of phenotypes in favor of the new adaptive norm is observed, with a number of generations initially not showing a preference for the old or new environmental conditions. This is the transitional phase of coexistence of both norms within the same individual spectra as two adaptive modifications (Fig. 10, III). Finally, the new norm is finally stabilized as a single adaptation (see Fig. 10, IV). Its separation from carriers of the previous norm can reach the level of reproductive isolation (Shaposhnikov, 1978).

There are a huge number of natural analogies for the described transformations, mainly concerning cases of 'adaptation' of various insect forms to pesticides. The genetic nature of the resulting resistance to a specific factor can vary in different cases (Dubinin, 1966a). As a result, new races arise, which, when crossed with the original ones, exhibit different types of segregation (Dobzhansky, 1947). There is no reason to believe that the process of their formation differs from that described above and that the new trait existed in a stable form in individual mutants before the start of selection.

* * * The presented notions of an epigenetic mechanism of the evolutionary process allow an assessment of the essence of the concepts underlying the synthetic theory of evolution. Analysis of the latter is hampered by the growing gap over time between its factual representations and those that are accepted declaratively, without being employed in its constructions. Evidently, the object of consideration must be the former category of concepts. The starting point here is the premise that the substrate of selection consists of mosaic Mendelian allelic factors identified with chromosomal loci. Evolution is a change in the genetic composition of a population, and its rate is determined by the rate of allele replacement (Dobzhansky, 1947; Grant, 1980). For chromosomal genes, an ordered phenotypic expression is implied, quantitatively evaluated as the relative contribution to fitness. Phenotypic or trait stability is considered purely pre‑formist as a property of the genes that control them, and therefore its explanation does not belong to the tasks of the theory. The notion of stability here has only a supra‑individual meaning and expresses either the fixation of certain allele frequencies (genetic equilibrium) or, in a stricter sense, homozygosity for one or another allele, i.e., achieving a 100 % allele frequency in the population. The latter occurs through the elimination of alternative alleles—by selection or genetic drift (when population size is limited), when random deviations of frequencies from equilibrium may become irreversible. Accordingly, the subject of the theory is not the organization of the genotype and the individual development it governs, but the organization of the population gene pool.

The concepts of genetic determinism and inheritance (stability) are identified by the theory, i.e., every genetic change is understood as phenotypically inheritable, although it is acknowledged that its effect may be distorted by genotypic and environmental influences. Hence hereditary changes are regarded as independent of natural selection; the latter merely acts upon them but does not create them. Non‑heritable, or modificatory, deviations are understood as noise that obscures gene action and retards the selection of genetic changes (Dubinin, 1966a; Meyer, 1968; Grant, 1980). Since, by definition, no differences in the degree of stability of their effects are recognized among the latter, the notions of phenotypic norm and aberration lose any functional role in the theory. They possess no characteristics other than the frequencies of the corresponding phenotypes and genes in the population. For the same reason, the idea of evolutionary “absorption” of some phenotypes by others through a system of crosses is completely empty here; only replacement by displacement, i.e., the substitution of the corresponding alleles, can be spoken of. This replacement reduces the genetic dispersion of the population with respect to fitness and consequently diminishes the genetic load. The greater the number of genes simultaneously affected by selection, the larger the load and the slower the changes must proceed (provided population size remains sufficient). Since every mutation is treated by the theory as an allelic change controlling a stable trait, this amounts to an explicit acknowledgment of punctuated evolution, even though it is often denied verbally (Meyer, 1968; Ruse, 1977; et al.). The constraints imposed on saltationism are technical rather than principled. Where a simple allelic difference is established between morphs or races, it is taken as evidence of the emergence of one form from another by a single mutation (Dobzhansky, 1947, pp. 50, 190; Grant, 1980, pp. 88, 174); where many such differences are presumed (e.g., between successive species), evolutionary transformation is viewed as a sequence of mutations (Dobzhansky, 1947, p. 52). Mutation thus functions as the creative factor, whereas the creative role of selection is reduced to homozygosity of recessive genes that reveal new phenotypes, to the combination of gene effects, or finally to the creation of an optimal genotypic environment for gene expression. Evolution is therefore portrayed as a dialogue between the environment and genes (Novinsky, 1978). These reductionist principles of the theory—especially gene selection and the assignment of fitness coefficients—are increasingly unsatisfactory to its proponents and have long been characterized as “over‑simplifications” (Dobzhansky, 1947, p. 106). Yet this does not bring clarity to the question of what truly constitutes the substrate of selection. Various authors point to genotypes and gene combinations (Sheppard, 1970), or solely to phenotypes (Meyer, 1968), or to both together (Grant, 1980). This reflects more an intention to move away from old ideas than a genuine departure from them. Consequently, the mechanism of the driving selection is still reduced to the replacement of one allele by another (Grant, 1980, p. 144). According to R. Lewontin (Lewontin, 1970), modern population genetics, on which the synthetic theory rests, still assumes a purely mosaic combination of genes, without even considering their organization into linkage groups.

Another general source of presumed evidence for the emergence of stable adaptations through single genetic changes is biochemical mutations in lower organisms, particularly in bacteria. When a particular culture is transferred to an inadequate medium (e.g., a strain of E. coli that ferments galactose to a medium with lactose), only individual mutant cells survive, and by the replica method, it can be shown that the adaptively valuable change is not induced by the new conditions but existed in the original culture (Dubinin, 1966a, 1976). This appears as the spontaneous emergence of a new adaptation. However, it is easy to see that what is selected here is not a stable adaptation, but simply one of the elementary modifications. For example, in our case with lactose fermentation, the ability is manifested in mutant cells (to a greater or lesser extent) only against the corresponding provocative background, while on the original galactose medium, they perform their previous normal function; otherwise, such cells would have died before transfer. In the modificational spectra of mutants of this type, both reactions are present simultaneously, and the choice between them is determined solely by external conditions. Stable fixation of the new reaction, apparently, can only occur during the reproduction of the altered strain in the new environment as selection proceeds among many individual cells. The latter possibility is admitted in many cases where the gradual adaptation of bacteria to increasing doses of a harmful factor is clearly visible (Dubinin, 1976). However, it seems that it is not always realized that this is the general pattern of the formation of new adaptations. It should be taken into account that with the high reproduction rate of bacteria, the stabilization of selected morphoses in a new environment occurs extremely rapidly, obscuring their initial instability.

Thus, the idea of evolution as selection of phenotypically stable gene variations is built on facts that cannot prove it. These facts characterize not the process of adaptogenesis itself, but either its initial basis (manifestations of biochemical mutations) or its final result, i.e., changes in the frequencies of already formed components of polymorphic systems. In practice, however, the formation of new adaptations is impossible without the stabilization of individual reactions that initiate them.

But if the material of evolution consists of labile reactions (modifications) carried out by norm-forming genotypes only under deviating conditions, then it is clear that they, by definition, do not obey Mendelian rules. Crossing two elementary morphoses of the norm cannot yield a stable result, despite the genetic differences of their carriers. In the case of crossing such a morphosis with a stable normal phenotype, the former is naturally absorbed by the latter (when the offspring develop under normal conditions). The only natural situation where correct Mendelian segregation occurs (excluding the most severe mutational disturbances) is the crossing of variants of the norm, i.e., stabilized phenotypes that form a polymorphic system. Only in this case do alternative pathways of stable development (creodes) arise in the epigenetic systems of hybrids, between which an ordered selection takes place. In short, only the products of canalized development (adaptive morphs) mendelize, not the products of its destabilization (morphoses). Therefore, for the epigenetic theory of evolution, the world of natural variability as a whole is not described in Mendelian terms. It consists not of alleles (in which the theory sees only relations between certain types of development), but of phenotypes themselves, i.e., variants of the norm and their aberrations. The behavior of individual phenotypes in crosses depends exclusively on their belonging to these two classes, which characterize the type of individual development, as well as on the conditions of development themselves. Darwin's indefinite variability is the totality of aberrations of the norm and, therefore, the realm of phenotypes that do not have ordered inheritance. The difference described in Mendelian alleles is the result of selection for alternative phenotypes, not just any manifestation of genetic difference.

From these positions, the explanation for why beneficial elementary changes do not dissolve in crosses and can be preserved by selection fundamentally changes. In the history of the criticism of Darwinism, the assumption of such dissolution is known as "Jenkins' nightmare," and throughout our century, evolutionists have tirelessly repeated that this argument has been eliminated by the discovery of Mendel's laws with their principle of "purity of gametes." But these laws are universal for chromosomes, not for traits, and are not applicable to indeterminate phenotypic variability. The true reason for the effectiveness of selection lies not in the high heritability and strict discreteness of its elementary changes, but in their belonging to a limited space of aberrations inherent to a given phenotypic norm. As a result, these variations, which are inherently unstable, are cumulatively repeated across generations, ensuring the stability of the potential substrate for selection without the stability of individual inheritance. Selection in favor of one of them leads not to the preservation of its "unmixed factors," but only to an increase in the numerical role of gametes left by its carriers, regardless of their individual genetic constitution, which changes in each generation.

The true assimilation of genetics achievements by Darwinian teaching cannot be based on the absolutization of the regularities it reveals (as well as any forms of biological orderliness in general) as universal factors acting outside of natural selection. On the contrary, they should be interpreted as results of selection that do not exist outside of it. The task of evolutionary theory is not only to uncover biological laws but also to determine the conditions for their fulfillment, i.e., to realize their relativity. Therefore, when speaking of an evolutionary explanation of genetic data, we must consider the entire real spectrum of facts accumulated by it, not just the sum of rules used for hybridological analysis of pure lines. The only acceptable basis for such an explanation appears to be the epigenetic theory of Shmalhausen-Waddington, which considers heredity as an expression of the stability of individual development created by natural selection. It can be argued that it is with this theory, which places the properties of the holistic system of organism development at the center and strives to describe the facts of embryology and genetics in a single language, that the future development of Darwinism is linked.

Agayev M. G. Experimental Evolution. Leningrad: LSU Publishing House, 1978.