Edited by: Kepa Ruiz-Mirazo, University of the Basque Country, Spain
Reviewed by: Giovanni Boniolo, University of Ferrara, Italy; Johannes Jaeger, Complexity Science Hub Vienna (CSH), Austria
This article was submitted to Systems Biology, a section of the journal Frontiers in Physiology
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
We address the identity of biological organisms at play in experimental and modeling practices. We first examine the central tenets of two general conceptions, and we assess their respective strengths and weaknesses. The historical conception, on the one hand, characterizes organisms' identity by looking at their past, and specifically at their genealogical connection with a common ancestor. The relational conception, on the other hand, interprets organisms' identity by referring to a set of distinctive relations between their parts, and between the organism and its environment. While the historical and relational conceptions are understood as opposed and conflicting, we submit that they are also fundamentally complementary. Accordingly, we put forward a hybrid conception, in which historical and relational (and more specifically, organizational) aspects of organisms' identity sustain and justify each other. Moreover, we argue that organisms' identity is not only hybrid but also bounded, insofar as the compliance with specific identity criteria tends to vanish as time passes, especially across generations. We spell out the core conceptual framework of this conception, and we outline an original formal representation. We contend that the hybrid and bounded conception of organisms' identity suits the epistemological needs of biological practices, particularly with regards to the generalization and reproducibility of experimental results, and the integration of mathematical models with experiments.
Scientists often describe biological organisms as exquisitely complex objects. The adjective “complex” has various meanings, and one points to a difficulty in providing an adequate account of their identity, notably in modeling and experimental practices. What does organisms' identity refer to? As for any object, the identity of an organism designates what makes it what it is and, thereby, what makes it different from something else.
We can understand every conception of organisms' identity as spanning over a spectrum going from more stringent to more inclusive interpretations. At one end of the spectrum, the identity of an organism points to its
Each interpretation of identity in the spectrum provides criteria that generate a
The choice of the interpretation of organisms' identity depends on the aim pursued. In science, moreover, interpretations and classes are not supposed to be merely arbitrary or practical groupings of objects: to be relevant, they should stem from theoretical conceptions and frameworks (Grimaldi and Engel,
In this paper, we focus on the concept of organisms' identity that is relevant to experimental and modeling practices in Biology. Experimental practices require observing particular organisms. Yet, the knowledge that biologists usually try to obtain from their experiments is not supposed to be just about particular organisms but, instead, to hold for any other organism endowed with the
The complexity of biological organisms vis-à-vis identity is the acknowledged difficulty of treating particular organisms as instances of the same experimental object, and of subsuming them under the relevant classes (Agutter and Wheatley,
The first reason is that, in both theoretical and empirical practices, scientists can only take into account a few aspects of biological organizations, understood here as the whole set of functions and processes constituting each organism. Typically, mathematical models only focus on some target features while neglecting many others, although such neglect does not rely on a clear theoretical justification and a systematic method. The same applies to experimental quantitative measurements, which are limited to only some aspects of the organisms under study.
A second reason, related to the previous one, is the strong coupling between biological organisms and their context. The context should be understood here in a comprehensive way, so as to include abiotic elements as well as other organisms, both participating in the determination of organisms' identity (Gilbert et al.,
A third reason is that contingent features that appeared throughout historical processes contribute to determining the properties of current organisms. In evolutionary theorizing, this idea corresponds to the “contingency thesis” (Beatty,
For all these reasons, an account of organisms' identity as experimental objects is a challenging task. Specifically, the challenge consists of adopting a conception of relative identity that generates one or several classes appropriate for the generalization and the reproducibility of experimental results. Such a conception would provide an operational tool for both empirical practices and mathematical modeling.
How is organisms' identity conceived in current biological practice? It seems to us that two broad theoretical conceptions can be distinguished. The first conception is
As we will discuss, each conception is open to different interpretations of identity, going from more restrictive to more inclusive ones. For instance, evolutionary taxa also stem from a genealogical conception, but these classes are much more inclusive than the ones which are relevant for most experimental practices, where biologists deal with strains rather than species or higher ranks (see Montévil,
Both conceptions are at work in experimental practices, and each of them has strengths and weaknesses. Genealogical strategies, we argue, enable scientists to consider
We can understand the relations between these two conceptions in different ways. One could favor the genealogical conception because it matches the historicity of biological organisms that emanates from the Darwinian theory of evolution. Alternatively, one could argue that the relational conception is the most fundamental one; its limited validity would be the mere effect of our (current) lack of theory and empirical knowledge. An example of the latter attitude (although not specifically addressing experimental practices) is Goodwin and Webster's relational theory of form changes that they take as a requirement to ground phylogenetic reasoning (Webster and Goodwin,
We advocate here a different view. We argue that biology requires combining genealogical and relational conceptions, with the support of an appropriate theoretical framework. The genealogical conception provides a procedure to select whole organisms as candidates to be subsumed into relevant identity classes. In turn, the relational conception – especially in an organizational version—provides explicit guidelines to understand the stability of biological organisms and, thereby, of the domain of validity of identity classes, notably in time. The main upshot of our analysis is a
We describe in this section the two conceptions of organisms identity at work in experimental and modeling practices in biology, and we focus on their background epistemology. We aim at making explicit their respective strengths and weaknesses which, because of their complementarity, open the way to the elaboration of an integrated conception.
A genealogical (or historical) conception of identity may take different forms. For instance, genealogical identity can be understood as the
Historical identity is—unsurprisingly—at work in systematics, the discipline that elaborates the classification and taxonomy of biological organisms and whose results are used ubiquitously in biological practice. In systematics, particular organisms are considered as members of the same class if they belong to a monophyletic group, which includes only and all the descendants of a last common ancestor. How do systematics build classes? While the concept of genealogy comes from Darwin's theory of evolution, genealogies are usually not observable as such. For example, it is not possible to ascertain that a given fossil species is an ancestor of a current species. Instead, it is possible to show that a given specific fossil species is more closely related to a given current species than to another one. As a result, unlike the genealogy
The use of the genealogical conception of identity extends to day-to-day experimental practices across various biological disciplines. Biologists establish laboratory strains and usually run experiments on organisms coming from the same strain. By this practice, experimental biologists consider different individual organisms as hypothetically identical. For example, biologists assume that the properties of these organisms follow the same probabilistic distribution in statistical tests. When applying this conception, biologists do not exhibit a given set of observable properties that the organisms would share; instead, they build the identity class by referring to their shared recent origin. The “Methods” section of most experimental papers explicitly relies on this strategy.
Compared to the phylogenetic method of classification, the experimental practice is, at the same time, less conceptual and more operational. Experimental biologists do not estimate the genealogy by theoretical arguments based on similarities and hypotheses on evolutionary processes. Instead, they control genealogy empirically by letting the “ancestors” reproduce in laboratory conditions (Chia et al.,
Whatever interpretation is adopted, the genealogical strategy provides criteria that apply to both synchronic and diachronic identity of organisms. A group of organisms shares the same synchronic identity if they have a genealogical connection with a specific common ancestor. Likewise, each organism remains diachronically a member of the same class whatever difference (due to variation) appeared—or will appear—between it and the ancestor through time.
Identity classes built on genealogical conceptions (at least in the version discussed here) put no
Let us mention one last aspect concerning genealogical strategies. In principle, ascribing a relative genealogical identity to a group of organisms requires estimating their genealogy and their connection to a common ancestor. However, in systematics, the common ancestor is not directly accessible and cannot be an empirical reference. Instead, biologists anchor a name to a specific individual organism called a “name-bearing type” that is the ultimate reference for this name (CZN International,
Although biologists do not use strains universally, organisms obtained in this way are widespread in experimental practices. Yet, what justifies the fact of subsuming them under taxonomic classes, and giving them names coming from systematics? The implicit hypothesis is that strains under control are subsets of taxonomic ranks: for instance, the strain “black 6” is supposed to belong to the systematic class of mice (
The genealogical conception of biological identity has several strengths. This conception allows ascribing an identity to organisms as
In turn, genealogical identity suffers from significant weaknesses from the perspective of experimental practices, or applications such as medicine. While systematics aims at reconstructing the past and describing the present in light of the past, experimental practices investigate the relations between the parts of organisms, as well as between organisms and their surroundings. Because of these different goals, identity classes in systematics can include tetrapods that are such without having four limbs; in turn, empirical practices need classes that sustain reproducibility and generalizability of the results over a (hopefully large) group of organisms.
The source of the problem is the same that generates the strengths of historical definitions
The relational conception of identity stems from a different epistemological stance. The description (and, in science, the theoretical determination) of an object mainly appeals to the
The relational epistemology pervades most natural sciences and especially physics. For example, although the electric charge seems to be an intrinsic property of objects, it is ultimately a quantity that describes how charged objects exert forces on each other: therefore it is grounded on relations
What relations are relevant in the biological domain? After all, one may argue that genealogy is also a relation. In fact, what matters from a relational perspective is the
In biology, we distinguish two versions of the relational epistemology and the resulting conception of identity. A first version, adopted in particular by biophysics and systems biology, consists of studying biological organisms by using conceptual and mathematical tools common to other natural sciences, as physics or chemistry. While it relies on well-established and operational tools, this “biophysical” version tends to look at biological organisms as physicochemical systems and, therefore, to emphasize common aspects while neglecting specifically biological ones. The resulting conception of biological identity applies to those aspects, and their relations, which are captured by the models. Different organisms are synchronic instances of the same object insofar as they possess the same aspects and relations captured by the model, and they maintain their identity diachronically if they conserve them in time.
The main strength of the biophysical conception of identity is that, in contrast with the genealogical one, it makes explicit the conditions of validity of experimental results. Generalizability and reproducibility of results hold for all organisms belonging to the same identity class, insofar as they possess the aspects and relations made explicit by the model or description. At the same time, this definition carries a crucial weakness: it considers
The problem with this abstraction move is two-fold. On the one hand, it implies dealing with organisms not as wholes, but as circumscribed sub-systems. In fact, biophysical models used in biology often apply also to abiotic phenomena (Douady and Couder,
The second version of the relational epistemology, which we label “organizational,” places a heavier emphasis on the distinct features of biological organisms. Its central assumption is that organisms are natural systems endowed with a distinctive organization. In particular, biologists can analyze organisms (be them unicellular or multicellular) as constituted of parts that depend on each other for their continued existence: biological “organization” refers specifically to such a mutual dependence among parts. Initially advocated by theoretical biologists like Nicolas Rashevsky (
Let us describe in some detail the central tenets of this organizational framework, by relying on some recent theoretical developments (see also Montévil and Mossio,
In this diagram,
The set of constitutive constraints that are both generative and dependent realize mutual dependence, which is usually referred to as
Organizational closure provides a specific interpretation of the circularity at work in biological organisms (Mossio and Bich,
With this brief characterization in hand, let us examine how the organizational framework deals with organisms' identity. As for any conception of identity, different interpretations of the organizational one can be adopted. The most restrictive relative interpretation seems to be that different organisms are instances of the same object insofar as they share the very same functional organization, i.e., if they realize (at some given stage of their lifetime) the closure of the
As a matter of fact, Difrisco and Mossio (
Let us now consider the most inclusive interpretation, according to which organisms are identical if they realize the closure of the very same constraints. We consider here
To the extent that organizational closure is a distinctive feature of biological organisms, this relational conception of organism identity seems to be more suitable because it avoids the first possible drawback of biophysical ones, i.e., the fact of leaving aside specifically biological aspects. Indeed, identity grounded on closure naturally considers organisms as whole entities. As for the biophysical conception, the organizational one makes explicit its domain of experimental validity. To be the same, different organisms must share the same organization. In contrast to the biophysical definition, however, an explicit description or model of the
A possible solution to the problem would be to establish descriptions and models of
One may object that such a solution would also face the problem of abstracting most of an organism's organization, just as the biophysical one. With no theoretical guidance, partial models would neglect aspects that might actually make a difference and induce variability between supposedly identical organisms. The objection is undoubtedly correct. Yet, we submit that the organizational framework has better prospects than the biophysical one for selecting relevant aspects of an organism within an adequate theoretical framework. The reason is that even partial organizational models are nevertheless models of closure (while biophysical ones are not) and therefore designed to account for the reciprocal stabilization of functional constraints within whole organisms. As a result, they can better determine the occurrence and impact of variations affecting organisms and the extent to which such variations could alter their identity.
The upshot of the previous section is that genealogical and relational conceptions of organisms' identity have complementary strengths and weaknesses. In what follows, we advocate their integration into a hybrid conception that, we hold, is better suited for taking up the challenge of organisms' complexity.
The connection with a fixed past allows the genealogical conception to define organisms' identity in a way that accommodates biological variations. However, genealogical identity does not refer to any observable property of organisms, which leaves unspecified to what extent experimental generalizations are legitimate. In sharp contrast, relational identity refers to the observable properties of organisms, which provide specific conditions for scientific generalization and reproducibility. Yet, relational identity faces the problem of abstraction with regards to most of an organism's organization, with the result that it seldom proves valid.
The reason why relational identity fails to apply to organisms easily is not only that a complete description of their organization is not accessible. Even if a complete description of an organization were available, we submit that the corresponding biological organisms would undergo
In this situation, we submit that an adequate conception of organisms' identity requires integrating genealogical and relational (organizational) strategies, as
Integration of genealogical and relational descriptions (reproduced from Montévil,
Physicists understand the changes taking place in a given phenomenon by variables connected by invariant relations, expressed as equations. By contrast, following the principle of variation, we submit that there is no invariant mathematical structure (as equations) underlying the behavior and dynamics of organisms.
A central epistemological implication is that we have to understand the relative stability of biological phenomena without overarching invariants. As mentioned in the previous section, organizational closure plays precisely this epistemological role at the individual scale, by contributing to explain how functional constraints stabilize each other through their reciprocal relations and interactions. As recently argued (Mossio and Pontarotti,
How should organism identity be characterized in this theoretical framework? We propose six main tenets. First, organism identity requires elaborating a generic description of organizational closure, which is supposed to apply to a group of individual organisms. Such a description aims to capture not only the relations between functional parts of an individual organism but also, and crucially, its interactions with the environment as an agent (Barandiaran and Moreno,
Second, organizational descriptions are necessarily partial, despite their possible complexity. This limitation implies that many aspects are neglected, be they other functional parts, or aspects of the environment, or other organisms. In section 2.2, we referred to this implication as the
Third, the genealogical strategy steps in and provides a procedure for dealing with the aspects that the organizational framework does not make explicit. The procedure considers as candidates for membership to an identity class those organisms which share the same past. Often, in experimental biology, organisms have a controlled,
To illustrate how the genealogical strategy fills in the gaps of the organizational one, let us focus on the treatment of specific functional constraints. A constraint is a relational concept, defined by its mathematical structure and its link with the constrained process (Montévil and Mossio,
Fourth, the organizational conception focuses on constraints closure, which contributes to explain how biological organisms can maintain themselves over time by constraining the thermodynamic flow. In particular, closure brings about an inherent tendency of organisms to stabilize existing functional constraints by removing many variations and by regenerating them in a fundamentally unaltered form. Such a tendency to conservation (what we have previously labeled “organizational inertia” in Mossio et al.,
Fifth, the tendency to conservation emphasized by the organizational framework
Sixth, the integration between genealogical and relational conceptions leads us to advocate a
Yet, the hybrid nature of the definition is not the end of the story. Indeed, our theoretical framework relies on the principle of variation, according to which individual organisms do undergo variation over time. The main implication here is that, even though an individual organism satisfies the hybrid conception at a given moment, there is no guarantee that it will do so as time passes. Consequently, although a population of organisms shares the same hybrid identity during several generations, sooner or later, some of these organisms will undergo variations that will contravene their membership to that identity class. As a result, our conception of organism identity is not only hybrid but also
The conceptual framework outlined above would gain clarity if it were expressed by an adequate formal language, which, to our knowledge, is currently lacking. Let us take some preliminary steps in this direction.
We first introduce a new symbol, χ, which represents the
In any characterization or model complying with the hybrid conception of organisms' identity, χ realizes organizational closure in combination with the constraints explicitly represented in relational terms. The overall characterization does not make the closure entirely explicit, precisely because it contains χ. A group of organisms that meet the hybrid model—and would, therefore, share the
Since there is no theoretical invariant specified by χ, its status is fundamentally different from that of a variable, as used in physics. Variables are defined through formal relations, while χ refers to a genealogical connection with a specific object, a particular. As a result, it is ultimately a symbol in the etymological sense of the word, bridging the formal description and the part of the world under study.
How is χ formally integrated into an organizational model or diagram? The general idea is to represent χ as a
Integration of a historical symbol and organizational closure. Since χ and the relational constraints have a different epistemological nature, we use different arrows for constraints and processes related to χ.
Diagrams integrating χ to organizational closure raise the question of the connection between χ and the explicit relational part. Depending on what the modeler knows and ignores about the organisms, the diagram has a different meaning and form, in particular with regards to the dependencies between χ and other constraints. Besides, if the diagram does contain a partial closure, specific organizational patterns become visible, and further general challenges arise. For instance, as one can see in
When considering the relations between χ and the constraints in a diagram, we can distinguish several cases. Without trying to be exhaustive, let us mention a few significant ones. It is worth noting that these cases are not supposed to be mutually exclusive: the very same χ in the same diagram can carry aspects that are relevant for several of these cases.
The first case is a generalization of the situation that we discussed earlier for
In the second case, we postulate that some aspects of χ are equivalent to aspects explicitly described in relational terms. The underlying hypothesis is that a constraint may have a single generic effect on a class of processes having different roles in the organizational diagrams. For example, cell membranes constrain the diffusion of a broad class of molecules similarly, or ribosomes constrain the translation of most RNAm similarly. In particular, a constraint can act in the same, generic manner on a process contributing to the partial closure
The third case refers to a situation in which, although χ could be “unpacked,” as discussed above, the resulting organizational model would be extremely specific, and therefore unfit to sustain generalization and reproducibility. In other words, the transition from an initial diagram
The fourth and last case that we discuss here concerns the situation in which χ includes intrinsically diachronic constraints. As such, these constraints may involve novelties that have not appeared yet and whose nature may be unprestatable (Longo et al.,
The integration of χ within organizational models covers a variety of situations. Following the specific scientific objectives and depending on the available knowledge, the relational part of the diagram can be more or less detailed, and generate more or less restrictive hybrid classes of identity (together with the genealogical control on χ). Yet, it is worth underscoring that, as we discussed in section 2.2, we maintain that an organizational description is never complete (be it for contingent or principled reasons), which means that whatever model of an organism does include χ. Organisms' historical identity possesses irreducibility that cannot be captured by any given organizational model.
By characterizing the identity of organisms for modeling and experimental practices, organizational diagrams integrating χ can also represent a typical experiment. Before concluding this section, let us have a brief look at this application of the framework (
Theoretical representation of a typical experiment.
Overall, the diagrams represented in
Biological organisms are a very peculiar kind of natural systems. They are familiar to us and, at the same time, resistant to a comprehensive scientific understanding. As claimed in the Introduction, they are complex objects.
The characterization of organisms' identity faces their complexity. It is a notoriously difficult task to tell whether a group of organisms that look similar at first sight does not hide substantial differences, which may be revealed after in-depth scrutiny. Similarly, it is difficult to make explicit the conditions at which it is legitimate to claim that an organism remains the same over time. Despite these challenges, a workable notion of organisms' identity is required, because of its pivotal role in grounding generalization and reproducibility in science.
In this paper, we have discussed the strengths and weaknesses of two broad conceptions on identity. The genealogical conception builds identity classes by reference to the past, especially by linking individual organisms to a common ancestor. Experimental biologists routinely use this strategy to work on hypothetically equivalent organisms. While it tends to work, genealogical identity does not provide its conditions of validity for experimental purposes. The relational conception, in turn, defines identity by referring to a set of relations possessed by individual organisms. While its conditions of validity are explicit, it faces the widespread problem of biological variability.
To overcome this situation, we have put forward a hybrid conception of organisms' identity. We have argued that the identity of biological organisms should be construed by integrating both genealogical and relational conceptions. In short, we suggest that individual organisms belong to the same identity class when they share the same specific organization of functional constraints
Even though the hybrid definition of identity was deemed to be useful and fecund in the biological domain, we have also underscored that the validity of identity classes cannot be but limited in time. Because of their inherent tendency to vary, individual organisms that meet the criteria of an identity class at some moment may contravene these criteria as time passes, and their offspring will presumably do the same after some generations. Therefore, organisms' identity is not only
All authors listed have made a substantial, direct and intellectual contribution to the work, and approved it for publication.
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
1By contrast, properties that do not stem from relations are arbitrary. For example, in classical mechanics, both stillness and uniform movement correspond to no force, thus ultimately to the same situation. In Galilean relativity, the difference between the two situations stems only from the arbitrary choice of a reference frame; choosing a different reference frame can transform the stillness of an object into uniform motion.
2Current debates in physics concern the alternative between the possible use of absolute concepts (such as the absolute time of Newton) or the adoption of a
3A full-fledged discussion of constraints identity goes beyond the scope of this paper. As detailed in Montévil and Mossio (
4Let us mention that the issue is complex since mathematical descriptions, especially equations, precisely subsume a diversity of situations under the umbrella of a single mathematical frame. As a result, different views coexist. Two systems may be considered different on quantitative bases, either by their states (different positions) or their parameters (different mass). On the opposite, they may also be different if the overall equation representing them is different. Last, there are situations in between. For example, physical morphogenesis or bifurcation are situations where a change of state corresponds to a qualitative change of the trajectory or structure of the object.
5The idea behind genealogical proximity can be understood from a more general perspective in terms of
6There are several theoretical scenarios in which such functionally significant variations can appear. One possibility consists of a significant geometrical change (as neovasculogenesis in the case of the vascular system) or a mutation (in the case of DNA) affecting a constraint. There are other scenarios, which include more general changes of organization (Montévil,
7The necessity of this justification makes a principled difference from biophysical relational models discussed above. While the latter can focus on some local constraints or constraints dependencies and could acknowledge that, “somehow,” these local phenomena are connected to the global organization, organizational models cannot focus on local phenomena if they cannot exhibit and justify the connection between the parts and the whole.
8Note that we write that “almost” all other cells depend on oxygen transport. χ refers here to the historical identity of organisms (they are mammals), and, as discussed, it can include variations. In cancers, for example, cells switch to the glycolytic metabolism that does not require oxygen, a phenomenon called “the Warburg effect” (Vander Heiden et al.,
9Note that the genealogical specification of χ may also be more restrictive.