When Mechanisms Are Not Enough: The Origin of Eukaryotes and Scientific Explanation

In recent years, mechanistic talk has become very popular among philosophers of science. Particularly, mechanistic talk has displaced the traditional approach to scientific explanation in terms of scientific laws (Nicholson 2012). Mechanists claim that scientific explanation consists of looking for a causal process –in this sense, the mechanistic movement is just the other side of the coin of traditional causal models of explanation– such that, through connecting the different entities and activities that participate in the process, the phenomenon that we aim to explain simply emerges. This claim is in contrast with the claim made by defenders of nomological expectability models of scientific explanation who generally claim that “to explain a phenomenon is to make it expectable on the basis of non-accidental regularities” (Díez 2014, 1414). Mechanists usually put forward biology as their main counterexample against defenders of nomological models: when biologists claim to have explained a phenomenon, they do so on the basis of having found a mechanism that brings that phenomenon about (Machamer et al. 2000). Biologists do not appeal to laws of nature, logical arguments, or any other kind of logic: they simply appeal to mechanisms. Thus, scientific explanation is, on this view, mechanistic explanation. In this paper, we contend this claim on its own terms, by presenting an example from biological practice. Specifically, we present the case of the origin of the eukaryotic cell and argue that the explanation of the salient features of this peculiar case is more suited to be understood in terms of a nomological expectability model of scientific explanation than in terms of mechanisms. For this purpose, we make explicit a kind of general regularity that biologists seem to be assuming when they provide explanations of the origin of the eukaryotic cell, and which forms the basis of the kind of proposals that they take as explanatory of certain facts that they consider particularly salient and in need of explanation (see Alleva et al. 2017, for a similar line of reasoning applied to the case of allosterism).

The paper is organised as follows: In Sect. 6.2, we introduce the symbiosis theory (ST, hereafter) of the origin of the eukaryotic cell, nowadays considered the canonical model for explaining the origin of eukaryotic cells, and we introduce a classification of the questions that ST provides answer to. In Sect. 6.3, we introduce the mechanistic account of scientific explanation complemented with Woodward’s account of causality and provide evidence that suggests that the appeal to mechanisms is not the most appropriate way to justify the explanatory character of ST of the origin of the eukaryotic cell and why this is so. In Sect. 6.4, we present a nomological expectability model of scientific explanation that we then use to provide an understanding of the explanatory character of the ST of the origin of eukaryotic cells by considering that ST appeals to scientific laws. Finally, in Sect. 6.5 we conclude by defending the superiority of the nomological approach over the mechanistic approach in providing an understanding of the explanatory practices of biologists in the context of the theories of the origin of the eukaryotic cell and we propose future lines of research.

The biological world is populated by different kinds of entities, ranging from cells, to all kinds of multicellular forms of life. Cells are normally taken to be the basic and most fundamental unit of life, of which all the other entities are made up (Archibald 2014; Audesirk et al. 2008; Stearns and Hoekstra 2000). There are two main types of cells, classified according to the location of their DNA: prokaryotic cells (subdivided into the domains of Archaea and Bacteria) and eukaryotic cells. The main structural difference between prokaryotic cells and eukaryotic cells is that in the former, the genetic material is dispersed throughout the cytoplasm; whereas in the latter it is encapsulated within a membranoid-structure called the “nucleus”. Apart from this, there are many other structural differences between the two types of cells, concerning aspects such as their size (eukaryotic cells generally being bigger), the types of membranes and the presence or absence of organelles. This last different constitutes a salient feature of eukaryotic cells, since only they host organelles within their bodies. Organelles are structural subunits, analogous to organs in humans, which perform certain functions within the body of the cell they belong to. Two of the organelles within eukaryotic cells are mitochondria (present in all eukaryotic cells) and chloroplasts (present only in plant eukaryotic cells); these two organelles bear their own DNA. Mitochondria are the site of cell respiration. Photosynthesis, in contrast, takes places within chloroplasts. Eukaryotic and prokaryotic cells are quite distinct from each other, and there does not seem to be any record of an intermediate form between the two types of cells, which is why certain biologists have referred to the origin of the eukaryotic cells as “the greatest single evolutionary discontinuity to be found in the present-day living world” (Stainer et al. 1963, quoted in Sagan 1967, 226). This immediately triggers a serious question for biologists: how did the first eukaryotic cell appear, given that all organisms share a common ancestor, and therefore eukaryotes and prokaryotes must have originated from the same ancestor?

To answer the set of question outlined above, two families of theories have been proposed: on the one hand, self-genetic or autogenous theories, according to which the organelles within eukaryotes appeared as a consequence of invaginations within the original pre-eukaryotic cell (Raff and Mahler 1972; Uzzel and Spolsky 1974; all reviewed in Sapp 2010, 130–131; O’Malley 2010; Archibald 2015, R912); and on the other, symbiosis or exogenous theories, whose main claim is that eukaryotic cells originated through the symbiotic merger of two previously extant prokaryotic cells (Margulis 1970; Martin et al. 2012; Cavalier-Smith 2013; Dolan 2013). In short, the proponents of ST argue that the eukaryotic cell evolved as a consequence of a phagocytic process in which prokaryotes “were swallowed but not digested” (Margulis 1970, 60). The difference between the two families of theories is radical, and so are the conclusions that one can derive from them. For instance, if one defends an autogenous theory, one has difficulties explaining the genetic affinities between mitochondrial DNA and the DNA of free-living prokaryotes, since one has to explain how this foreign DNA arrived in the mitochondria of present-day eukaryotes. However, if one defends a ST, this fact becomes easily explainable: the fact that in the origin of eukaryotes two different prokaryotic lineages merged makes it more likely that the primitive lineage associated with mitochondria still preserves part of its original DNA. The same logic can be applied to all kinds of questions that might be raised about the difference between prokaryotes and eukaryotes. So, the capacity to play a more effective role in scientific explanation proves to be a good feature for preferring one theory to another.

Nowadays, the ST family predominates among biologists, although the versions of it come in many different forms, with at least 20 different models that explain the origin of the eukaryotic cells appealing to symbiosis (Archibald 2015). What matters for the purposes of this paper is the general structure of the arguments that appeal to endosymbiosis to explain the origin of eukaryotes and to explain the set of why-questions that we have selected as relevant, more than the peculiarities of the different models.

Mechanistic explanation is the most influential approach to explanation in biology. The view was originally presented in direct opposition to the (previously) dominant nomological models of scientific explanation. Mechanists argue that in order to explain a biological phenomenon it is necessary to describe the mechanism that brings the phenomenon about (Glennan 1996, 2002; Bechtel 2011; Bechtel and Richardson 1993; Bechtel and Abrahamsen 2005; Machamer et al. 2000; Craver 2006, 2007; Darden and Craver 2002). Describing a mechanism, they claim, is not the same as presenting a scientific law that underlies a phenomenon. In fact, they deny the possibility of explaining a phenomenon by subsuming it under laws. In other words, the explanatory character of a mechanism does not lie on its supposedly underlying regularities, but in the identification of causal relations: “while it is sometimes the case that description of the inner parts of the mechanism will entail a description of the mechanism’s outward behaviour, the explanation lies not in the logical relation between these descriptions but in the causal relations between the parts of the mechanism that produce the behaviour described” (Glennan 2002, S348; see also Machamer et al. 2000 for a similar argument).³

There are several ways of describing what a mechanism is. For instance, Machamer et al. (2000, 3, our emphasis) claim that a mechanism is a set of “entities and activities organized such that they are productive of regular changes from starting or set-up conditions to finish or termination conditions”; Glennan (2002, S344, our emphasis) defines a mechanism by saying that it is a “complex system that produces the behavior by the interaction of a number of parts”; Bechtel (2006, 26, our emphasis) says that it is “a structure performing a function in virtue of its component parts, component operations, and their organizations”.

It seems clear from the above definitions that all of them presuppose that a mechanism consists of a set of entities and activities (or parts and operations/interactions) plus their corresponding organization.⁴ To identify a mechanism, therefore, one has to disentangle its parts (the entities), individuated by their properties, and the activities it is involved in, “the producers of change”. Allegedly, the properties of the entities plus their organization are responsible for the way in which the activities come about. In the words of Machamer et al.: “Mechanisms are identified and individuated by the activities and entities that constitute them, by their start and finish conditions and by their functional roles” (2000, 6). This dualist reading of mechanisms in terms of entities and activities generates a new framework that should, in principle, be fruitful when it comes to clarifying notions such as causation, lawhood, function and explanation. In particular, the notion of activity is supposed to play the role of causes, laws and functions. For instance, if a law is supposed to be a regularity of something that acts in the same way under the same conditions, philosophers of a mechanistic bent can provide a similar reading of a mechanism: “a mechanism is the series of activities of entities that bring about the finish or termination conditions in a regular way” (Machamer et al. 2000, 7). According to such authors, these regular mechanisms are not accidental and can give support to counterfactual reasoning. Therefore, there is no need to talk of laws in biology, for their role is already played by the identification of activities within mechanisms. In the same vein, Glennan refers to the interactions within a mechanism as “invariant change-relating generalizations” which can support counterfactual claims (Glennan 2002, S344).

Given this characterization of a mechanism, we can now say that to give a mechanistic explanation of a given phenomenon consists of giving a description of the mechanism that brings the phenomenon about, such that the explanans includes the set-up conditions (arbitrarily taken as the beginning of the mechanism) plus the intermediate entities and activities together with their organization.

Nonetheless, there still remains the problem of providing criteria for identifying the different parts that compose a mechanism and that should be taken as relevant for the purposes of explanation. One possible way out of this problem, adopted among others by Craver (2007, 144), is to make use of Woodward’s manipulability criteria for identifying causes (Woodward 1997, 2000, 2003). Woodward’s strategy is to look for a “difference-making” clause in the explanans that, if we were to change it in various possible ways, would result in the final phenomenon being different. This strategy is mainly interventionist: if we want to identify the relevant factors for the production of a particular phenomenon, we must block certain parts allegedly involved in the causal path that terminates in the phenomenon to see whether this intervention has any consequence on the final output. Following this line of reasoning, one can say that “a part is causally relevant to the phenomenon produced by a causal mechanism if one can modify the production of this phenomenon by manipulating the behavior of the part, and one can modify the behavior of the part by manipulating the production of the phenomenon by the causal mechanism” (Nicholson 2012, 160).

Woodward is conscious that the interventions he requires to uncover the causes of phenomena might not always be available (think, for example, of historical phenomena). In order to resolve this difficulty, he argues that in those contexts where such manipulation is not feasible, the manipulability strategy takes the form of a counterfactual claim: “The notion of information that is relevant to manipulation thus needs to be understood modally or counterfactually: the information that is relevant to causally explaining an outcome involves the identification of factors and relationships such that if (perhaps contrary to fact) manipulation of these factors were possible, this would be a way of manipulating or altering the phenomenon in question” (Woodward 2003, 10). In other words, even in contexts where manipulation is not possible, it is “heuristically useful” to pursue or think of causes in the same way as we do when the relevant manipulation is available.

Biologists’ appeal to the notion of symbiogenesis, as we have argued, has the form of a general pattern: the biologists look for a general principle, which may be quite vague (in the sense that it might be applicable to a large number of entities, irrespective of their particular biological properties), that allows them to say not only how the first eukaryotic cell came about, but also why it has the properties it has (which are the answers to the four why-questions we have presented, plus other relevant questions that might be asked). It is convenient to specify at this point why we consider symbiogenesis to work as a regularity that might be used to account for certain facts (Archibald 2014; Douglas 2010).

First of all, symbiogenesis mere implies that the process by which an actual living organism has come about is a consequence of a symbiotic merger. Furthermore, in the case of the eukaryotic cell, it is always specified that this symbiogenesis gave rise to a case of endosymbiosis, whereby one organism lives inside the other. However, nothing about the particular nature of the organisms that interact endosymbiotically is specified, nor does it require to be specified in a general definition of symbiogenesis. Symbiogenesis just says something about how the mode of life of the organisms came about. Second, and related to the vagueness of the term, symbiogenesis is supposed to cover all the different cases of structures (and species) that emerge as a consequence of symbiosis between two different organisms. This entails that the entities that can interact symbiotically and give rise to a case of symbiogenesis are very different with respect to each other: bacteria, fungi, arthropods, mammals, etc.; they can all bear endosymbionts and/or enter endosymbiotic relationships with others. Third, by its very nature and its connection with the appearance of new biological structures, when it occurs through the acquisition of endosymbionts, symbiogenesis tends to trigger certain changes in the organisms involved: genomic decay, genetic assimilation, free exchange of genes between partners, vertical transmission, the appearance of particular bodily structures to bear the symbionts, etc. The evolution of these particular traits will differ depending on the particular relationship between the organisms and their necessities, and is normally what causes endosymbiotic states to be irreversible. Fourth and finally, symbiogenesis normally leaves some traces of the previously independent life of the partners. However, these traces vary quite a lot if we consider them on a case-by-case basis. Sometimes the traces will be biochemical pathways; others, molecular properties or chromosome structure, etc.

We believe that these four characteristics of symbiogenesis justify consideration of the phenomenon as a general pattern that biologists use in order to guide their research and to explain certain features that would not be explained otherwise. Indeed, the key aspect of symbiogenesis, in relation to accounting for the features of the eukaryotic cell as mentioned above, is that it makes these “expectable on the basis of [a] non-accidental regularit[y]” (Díez 2014, 1414). Nonetheless, this pattern, though general, is not empirically empty: it says something about the past and the future of the organisms which interact, and this can be studied further (and proved to be true or false). We believe that symbiogenesis, understood as we have specified above, is a kind of scientific law in Mitchell’s sense (1997, 2000, 2003). In Mitchell’s account, laws are understood pragmatically, according to the role they play in scientific practice. In other words, laws are not interpreted in terms of necessary and sufficient conditions, as traditional normative approaches suppose, but in terms of what they allow scientists to do. In this vein, Mitchell argues that a scientific statement must be understood as a scientific law if it allows good predictions to be made, good explanations to be provided and feasible interventions to be designed. This flexible conception of scientific laws allows her to provide a new multidimensional framework to represent a whole set of scientific generalizations (Mitchell 2000, 259). Furthermore, scientific laws in this sense provide a certain non-zero degree of nomic necessity,⁹ which is established in terms of the stability of the conditions upon which the regularity is contingently dependent.¹⁰ Therefore, the degree of nomic necessity of regularities in physics is higher than that of regularities in biology, because the stability of the conditions upon which a regularity is contingent in physics and in biology are significantly different. However, both regularities in physic and in biology involve a certain degree of nomic necessity; which is what matters here and is relevant for considering these generalizations as legitimate scientific laws.

In the context of the symbiosis models of the origin of eukaryotes, the appeal to the concept of symbiogenesis seems to play the role of a scientific law in this sense. First, it is often supposed that endosymbiotic association between two different organisms will give rise to a tendency for a series of adaptations to evolve that will increase the tightness of the fit between the partners. These adaptations will tend to evolve due to the possible presence of “cheaters”, i.e. organisms that benefit from the association without providing any benefit to its partner. This is a consequence of the fact that endosymbiotic associations that are capable of evolving adaptations that prevent the possible presence of cheaters outrun those that are not. Second, it is also assumed that the partners in an endosymbiotic association will still preserve some traces of their previous free-living state, as a consequence of the special features of the symbiogenetic process. Indeed, symbiogenesis sometimes entails (and it definitely does so in the eukaryote case) a transition in biological individuality. But, as is well known, the framework of transitions in individuality assumes the existence of individuals whose evolutionary fates align and form a higher-level entity. It is precisely the existence of independent individuals whose individualities become combined into a higher-level unit what makes it reasonable to expect that certain features of their previously independent existence will be preserved. In addition, the features that are preserved could be studied in a lab, making certain predictions possible. It is in at least these senses that we believe symbiogenesis plays the role of a nomic pattern (a pragmatic law): it allows for certain predictions, makes a set of phenomena that can be empirically tested expectable and supports counterfactuals. This nomic character seems to be the aspect of the notion of symbiogenesis that biologists have in mind when they use it for explanatory purposes.

Of course, defenders of mechanistic explanation might still question the alternative that we offer to mechanistic models of scientific explanation. As is well-known, the models that have traditionally appealed to scientific laws as the main explanatory “weapon” are conceptually flawed –they have to face numerous problems: flag-pole cases, contraception pills and male pregnancy, syphilis-paresis cases, vitamin C and flu recovery, etc.– and are not very popular among contemporary philosophers of science (Woodward 2017). Maybe, after all, we have to admit that, although not perfect, as our case illustrates, mechanistic explanation is the best theory of scientific explanation that we have for the moment. Nonetheless, Díez very recently proposed a new neo-Hempelian account that solves most of the conceptual problems that have been raised against nomological expectability models and –allegedly– would include mechanistic explanations as a specific subcase satisfying additional conditions (Díez 2002, 2014). As this is the only nomological alternative we know of that has these features, we now proceed to evaluate whether Díez’s model can accommodate the case of the origin of the eukaryotic cell.¹¹

Díez’s model takes as a point of departure Hempel’s thesis that “to explain a phenomenon is to make it expectable on the basis of non-accidental regularities” (Díez 2014, 1414). This expectability, however, is not as strict as it has traditionally been in deductive/inductive nomological models (one of the possible forms that nomological expectability models can take), where the cases in which the explanation is based on a low-probability relationship between the explanandum and the explanans were excluded. The reason for this exclusion was that explanations were taken as logical inferences; thus, in the case of inductive inferences, they demanded high probability (Hempel and Oppenheim 1948; Hempel 1965). In contrast, Díez substitutes the notion of logical inference for the less demanding notion of “embedding”: according to Díez, to explain a phenomenon is to embed it “into some branch of a net of theoretical constraints” (the explanans) such that they make the phenomenon expectable (Díez 2014: 1419). The idea of embedding is the structuralist counterpart to the positivist notion of implication and it presupposes a distinction in scientific models/structures between data models and theoretical models (Balzer et al. 2012). A data model is a structure that describes the phenomenon to be explained; whereas theoretical models are the structures defined by their satisfying a theory’s laws. A data model is embeddable in a theoretical model when the former “fits” the latter, i.e. the relevant values of the phenomenon square with those of the theoretical model. Embedding is thus a relation between models, not a relation between sentences, which allows for a weakening of the positivist demand for logical inference (for instance, making room for embedding in increasing yet law probability cases) but still preserves the core intuition behind Hempelian expectability. To put it in Díez’s words “[e]xplanations are (at least) certain kinds of predictions” (Díez 2014, 1420).

We will now provide an example of embedding. Suppose we want to explain the movement of the Moon using Newtonian mechanics. Our data model would include the Earth, the Moon, and the space and time functions that describe the kinematic trajectory of the Moon around the Earth, [DM_E,M = <{Earth, Moon}, space, time>]. The theoretical model would include, apart from the aforementioned components, the functions of mass and force, [TM_E,M = <{Earth, Moon}, space, time, mass, force>] defined by their satisfying Newtonian laws. The idea of the embedding of the data model within the theoretical model would be the following: by using the “machinery” of classical mechanics (laws of motion) plus the relative positions of the Moon and the Earth at a particular time, the theoretical model includes the relevant positions at other times; if such values fit the measured values of the data model, the former successfully embeds the latter, otherwise the embedding fails (and the theory has a Kuhnian anomaly). In this sense, model-theoretical embedding expresses the core intuition of nomological expectability.

However, as Díez explains and the case of the Moon’s trajectory exemplifies, nomological embedding, though necessary, is not sufficient for explanation, since we may still fail to have explanatory embedding in two kinds of cases. First, one may have embedding by merely descriptive/phenomenological theories that systematize data with laws describing general phenomena without explaining them (e.g. Galilean kinematics or Kepler’s laws). Second, in theories with what Kuhn calls “general schematic principles” such as Newton’s Second Law (Kuhn 1970), one can always construct ad hoc trivial “successful” embedding that cannot count as explanatory. To exclude these cases, Díez adds two further conditions: the embedding has to be ampliative and specialized. Its ampliative character is based on the notion of T-theoreticity (Balzer et al. 2012; related to Hempel’s distinction between “characteristic” and “antecedently understood”, and Lewis’s distinction between old and new vocabulary). T-theoretical concepts are those introduced by a theory such that, in order to determine their extension, one has to use/accept some T-law (e.g. mass and force in classical mechanics); whereas T-non-theoretical concepts are those which are already available and that can be determined (at least on some occasions) without the help of T-laws (e.g. space and time in classical mechanics). Explanatory embedding is ampliative, as in the case of classical mechanics: classical mechanics explains why the Moon is in location X at time t through embedding the phenomenon and introducing new T-theoretical concepts/entities (masses and forces) that do not appear in the data model DM_E,M. Thus, for embedding to be explanatory, it must make use of laws that (as in classical mechanics and not in Galilean kinematics or Keplerian astronomy) appeal to new concepts/entities. Specialization, on the other hand, requires that we introduce non-ad hoc “special laws” in order to account for the phenomena.¹² As Díez points out, we always require that our explanations include something more than merely schematic, very general principles such as ∑f = ma. In the case of the Moon–Earth system, for example, we need to introduce the law of universal gravitation, f = G*mm’/r ² , if we aim to explain the positions of the Moon over time.

In short, we might now say that a theory explains a phenomenon if: (1) we can embed the phenomenon in the theory, in such a way that the theory makes the phenomenon expectable; (2) the theory includes and makes use of at least one T-theoretical term; and (3) the theory incorporates and makes use of at least one special law in order to account for the phenomenon (Díez 2014, 1425). We will show that the appeal to symbiogenesis that biological theory makes to explain the origin of eukaryotes and the different phenomena laid out in Sect. 6.1, which does not fit the mechanistic account, is nevertheless perfectly legitimate and can be explicated by applying Díez’s re-elaborated model of explanation as nomological expectability.

First, the appeal to symbiogenesis provides a theoretical model that allows the embedding of the phenomena that constitute our data model. In the case of the origin of the eukaryotic cell, the data model would include objects such as membranes –of both cells and mitochondria– or genomes –again, both cell and mitochondrial genomes– and their respective biochemical properties –those of the lipid components of the mitochondrial membrane versus those of the lipid components of the cell membrane; circular, single-strand DNA versus linear, complex DNA, etc.– (DM_G,M = <{genome, membrane}, biochemical properties of both>). The theoretical model would include these objects plus entities/functions that correspond to the notions of fitness and symbiogenesis, which are purely theoretical and associated with particular features of symbiosis relationships and the theory of natural selection (TM_G,M = <{genome, membrane}, biochemical properties of both, fitness, symbiogenesis>).¹³ The embedding is possible in this case because DM_C,M happens to actually be a submodel that squares with TM_G,M, and TM_G,M makes the phenomena we aim to explain expectable (as reviewed in Sect. 6.1 in response to questions 1–4).

Furthermore, TM_G,M includes a couple of T-theoretical entities/functions, fitness and symbiogenesis, that play an ampliative role. Biologists do not explain the features of the mitochondrial genome by appealing to features of free-living bacteria. They explain them by appealing to the idea of symbiogenesis (and its specific endosymbiotic form): certain formerly free-living bacteria (that we can indicate through phylogenetic analysis) were at some point endosymbiotically acquired by an archaeon and, symbiogenetically, gave rise to the organelles that nowadays we call mitochondria. The preservation of the genetic and biochemical features of the mitochondrial membrane is explained by appealing to its symbiogenetic origin plus the fact that they confer fitness advantages. In this sense, it seems clear that the embedding is ampliative in the sense Díez’s account requires.

Finally, the explanation in terms of symbiogenesis includes an element of specialization in relation to ST (or the concept of symbiosis): an appeal to a special law which plays a non-trivial role in the explanation of the particular features of mitochondria. Symbiogenesis is a particular form of integration that two symbiotically associated organisms could enact, if the circumstances were favourable. It is well established that there are different types of symbiotic relationship (mutualism, commensalism and parasitism); some might be long-term evolutionarily relationships that are not conducive to integration, whereas others are. If they are conducive to integration and they have the desired fitness effects (i.e. they do not lead to the extinction of the integrated lineages), then they would trigger certain changes in the lineages that evolve symbiogenetically (mosaicism, genomic decay, loss of independent modes of life, etc.), giving rise to the appearance of new biological structures (they would fall down an evolutionary “rabbit hole”, as some biologists describe it, e.g. Moran and Sloan 2015). In contrast, if the symbiosis relationship does not lead to integration, even if it is a long-term relationship, it would lead to a different kind of changes that would affect to both organisms independently, such as certain phenotype changes, changes in behaviour, etc. In this sense, symbiogenesis plays the role of a special law concerning a more general principle of the expected outcomes of long-term symbiotic associations.

We believe this appeal to a special law is the crucial step in ST, it is what provides the main explanatory power and as we argued, it does not have the form of a mechanism. The special symbiosis law certainly is such in Mitchell’s pragmatic sense: it provides a certain degree of nomic necessity, therefore providing biologists with a guide to what they might find. For instance, appealing to a symbiogenetic origin makes it expectable that organelles, i.e. mitochondria, still preserve a certain degree of biological individuality that might be manifested, for example, by the possibility of in vivo replication. It is important to bear in mind that this would not be expected if the origin was self-genetic: in this latter scenario, we would never expect mitochondria to have any degree of biological individuality. Furthermore, if the origin of mitochondria is symbiotic, we will not expect to find intermediate forms in the fossil record, since symbiosis gives rise to saltational evolutionary events, which would not be the case if the origin was self-genetic. This same line of reasoning might be applied to all the features that ST makes nomically expectable and, in this sense, we have something similar to a pragmatic law that provides the research field with some order.

We should still note something about the special character of the law. As we said before, the condition is introduced in order to avoid counting as explanatory cases in which we merely apply general principles to trivially justify why certain phenomena occur (using ad hoc mathematical functions in ∑f = ma to explain intentional movement, for instance). One might argue that the appeal to symbiogenesis is still trivial in this last sense: it is just one general principle we could use to justify every feature we find in an organism. Nonetheless, this is not the case: the appeal to symbiogenesis rules out certain possibilities and it makes a difference (as does the appeal to f = G*mm’/r ² , in the case of planetary movement). It specifies the manner in which evolutionary innovation can arise, and this is in contrast to other possibilities, such as mutation, recombination, methylation, changes in the developmental matrix, or even other types of long-term non-integrative symbiotic relationships. It specifies a very particular pattern followed by the organisms that experience this mode of generation of evolutionary novelties and precludes triviality by ruling out the appearance of certain features that other evolutionary pathways would make expectable.

In conclusion, we have provided a (partially weakened, partially strengthened) nomological expectability framework as a possible alternative to a mechanistic framework of scientific explanation that explicates why biologists consider ST a legitimate explanatory theory of the origin of the eukaryotic cell by appealing to the notion of (pragmatic) scientific laws. In this sense, we have provided reasons to justify why an account of scientific explanation in terms of laws (in the restricted sense we have given) might be appealing to gain an understanding of the explanatory practices of biologists in certain contexts; an understanding that –we have claimed (Sect. 6.3)– mechanist philosophers cannot provide.

In this paper we have presented the symbiosis model of the origin of the eukaryotic cell together with a set of questions (phylogenetic, biochemical, etc.) that any theory of the origin of the eukaryotic cell must provide answers to. We argue that the notion of symbiogenesis, understood as the process by which a new biological structure (organ, metabolic pathway, etc.) originates as a consequence of a long-term symbiotic relationship, plays the entire explanatory role when biologists aim to provide an answer to the different questions we mention (Sect. 6.2). This said, we defend the idea that the mechanistic account of scientific explanation is not well-suited to understanding why the notion of symbiogenesis plays the entire explanatory role in these cases. First, we argue that every attempt to offer a mechanistic explanation to the questions previously mentioned turns out to be unsatisfactory, since they move to a level of detail which turns out to be unnecessary for the matters discussed; moreover, many of the causes that should be mentioned in a mechanistic account seem orthogonal to the type of phenomena that demands an explanation. Second, we show that the notion of symbiogenesis is far from being a mechanism as they are conventionally understood in the literature (in terms of parts, activities and organization): symbiogenesis is a regularity or general pattern that cannot be suitably captured in mechanistic terms (Sect. 6.3). Finally, we present Díez’s nomological expectability model of scientific explanation as an alternative to mechanistic models of explanation and defend the notion that Díez’s model helps in understanding the explanatory character of symbiogenesis, despite its not being a mechanism but a general pattern (Sect. 6.4). If our argument is sound, it shows how and why the appeal to general patterns –that might well be considered scientific laws in Mitchell’s sense, as we argue– might be explanatory in some contexts, thus challenging the universality of mechanistic explanations. It remains to be explored, however, whether the nomological expectability approach to scientific explanation we have defended here could also be applied to other biological contexts, either as a complement to (e.g. Alleva et al. 2017) or as a substitute for mechanistic accounts.