Confusing sorting with selection: sorting and selection must not be

In the previous section, I presented the main criticisms that have been raised against HCE, conceived as a thesis about the role of the holobiont as a unit of selection. Concretely, the criticisms were raised against the thesis that the hologenome is a replicator which, as I explained in chapter I, is just one of the many theses about biological individuality that HCE entails. In this section, I will present a criticism that, although suggested in different ways in the critical literature about HCE, has never been explicitly presented in the way in which I will present it here. I will argue that one possible interpretation of the widespread disagreement about the role of holobionts as units of selection consists in the fact that defenders of HCE, by confusing the concepts of the ‘ecological holobiont’ and the ‘hereditary holobiont’ (chapter I, section 3.1.4), are confusing the notions of sorting and selection, two notions that should be

55 _{In population genetics, additive genetic effects occur when the interactions between the}

genes are ‘linear’, meaning that all genes make the same contribution to the final outcome. Non- additive effects include dominance (when the action of an allele at one single locus can silence another allele in the same locus) and epistasis (when the expression of one gene at one locus depends on its genetic background, i.e. on the action of genes at different loci).

carefully distinguished in debates about the units of selection (Sober and Lewontin 1982; Sober 1984; Vrba and Gould 1986; Lloyd 1988; Okasha 2006).

Let me start by defining the two notions. In biology, and specifically, in the way in which the Modern Synthesis recasts Darwinism, evolutionary change in conceived at the level of the population. Populations are collections of causally connected individuals from the same species, amongst which there is variation. Different individuals in the population have different rates of death and birth, and as a consequence the composition of the population changes with time. This process of differential death and birth of the individuals in a population that ultimately leads to the changes in their representation of in the population that is considered to be evolutionary change is simply a process of ‘sorting’. In itself, sorting is simply the process of change in biological populations, but the concept does not specify any of the causes of why this process happens: sorting is simply what results from differential death and birth.

Given this definition of sorting, an important question for biologists is thus the following: what causes the process of sorting? Or, in other words, taking sorting for granted, what causes that the population will shift in one direction (i.e. towards a specific set of changes) rather than another? Here is where natural selection enters the picture: selection one of the causes of the process of sorting. Concretely, it is the non-random cause of differential birth and death of the individuals in a population. Other causes, normally qualified as ‘random’, include the different phenomena of drift, including genetic drift. Now, an obvious question arises at this point, and it is important to tell selection and sorting apart, namely: in what sense is natural selection a non-random cause of sorting? Here, a quote from Elisabeth S. Vrba and Stephen J. Gould is particularly illuminating:

‘Selection encompasses those interactions between heritable, emergent character variation and the environment that cause differences in rates of birth or death among varying individuals’ (Vrba and Gould 1986: 219)

Notice that the definition includes three elements: first, the localization of a focal level which is distinct from the environment; second, the localization of variation in character (phenotypic variation) at that level; third, the heritability56 _{of that}

variation in character. If these three properties are localized, then selection can arguably be understood as a cause of sorting, i.e. as a plausible cause for the differential rates of birth and death of the individuals in a population. However, and importantly, the biological world is usually considered to be hierarchically organized. That is to say, it is possible to find different focal levels where one can observe individuals that satisfy the three properties included in Vrba and Gould definition. Thus, the observation of sorting in one level of the biological hierarchy can always be the result of a process of selection acting on the upward or on the downward level, which generates a random (i.e. not due to selection) distortion in the level under study (Sober and Lewontin 1982; Vrba and Gould 1986; Lloyd 1988; Figure 16). Therefore, sorting and selection should not be conflated.

Figure 16. Table suggesting the relationship between selection at a focal level with its upward

and downward effects on the sorting of the entities at the upper and lower-levels. The table shows how what might look as selection at one level (e.g. effects of sorting on the organismal phenotypes) is just a result of the selection at a lower level (e.g. selection of ‘selfish DNA’). (From Vrba and Gould 1986: 220, Table 1)

56 _{Heritability should not be understood in its technical meaning in population genetics, but}

Take the following, and conventional example of selection: the case of selection for darker phenotypes—industrial melanism—in Biston betularia moths due to industrial pollution (Kettlewell 1958). In this case, the population is formed by two phenotypes: darker and brighter moths. In a non-polluted environment, darker months are less successful, because their pigmentation makes easier to identify by predators, and thus more susceptible to be eaten due to their phenotypic characteristics. However, in a polluted environment, the opposite is the case, and thus the trait distribution in the population shifts due to the higher susceptibility of brighter moths to be eaten. Selection is acting on a very specific trait (pigmentation), and thus might generate a distortion in the genetic makeup of the population (downward level) or on the macroevolutionary pattern of moths (upward level). However, the upward and downward level effects are mere sorting, which is due to the effect of selection on the focal level of moths. The changes in the genetic distribution, as well as the changes in the macroevolutionary patterns of moths (if any) are the distortions that result from the causal effect of natural selection on a different level of the biological hierarchy. Therefore, to prove selection at one level it is indispensable to prove that the change in the phenotypic character occurs at that level, that it can be inherited, and that it causally leads to the differential death and birth of the individuals at that level.

How to apply the distinction between sorting and selection to the case of HCE? I think it is possible to argue that defenders of HCE are confusing the two notions and are therefore taking cases of mere sorting as cases of selection or, in other words, confusing an upward effect with a real cause acting at a lower- level in the biological hierarchy. First, it seems unquestionable that holobionts, defined as a host plus its microbiome, have differential rates of birth and death. For instance, if holobionts are the metabolic-wholes of Dupré and O’Malley (2009; chapter I, section 6), it seems easy to argue that some holobionts live longer than others, and some holobionts produce more offspring (i.e. more metabolic-wholes) than others. Second, it seems at least plausible to think that holobionts have different phenotypic characteristics, and their different rate of birth and death can arguably be attributed to these differential phenotypic characteristics. But now the crucial question to decide whether selection can

possibly act at the level of the holobiont is whether inheritance can be detected, so that phenotypic features have a way of getting fixed, if they are advantageous. Hologenome defenders argue that this is possible, on the basis that the hologenome is the replicator. Hologenome detractors, on the other hand, argue that it is not possible to tell apart selection from sorting in the case of holobionts: how to distinguish cases of genuine host-microbiome coevolution and cases of ecological filtering? Assuming that the holobiont can be singled out as a genuine focal level that can be distinguished from the environment (which is something that some detractors of the holobiont would neglect, section 2), what are the criteria to tell apart cases of sorting from cases of genuine selection, i.e. differential rates of death and birth due to heritable phenotypic characteristics? The argument here would be that, since this is not empirically possible, then the holobiont cannot be singled out as a unit of selection, and thus HCE is empirically false. Defenders of HCE are thus conflating the indirect effects of selection on the differential rates of birth and death of the holobiont with the direct effects of selection on the individual species that interact in the holobiont and that produce sorting of holobionts as a byproduct.

5.1. But should sorting and selection really be distinguished? A brief reflection

The argument presented above seems a knock-down argument against HCE and seems even more strict that the argument based of SoS that I presented in section 3. The case I have made here is not that SoS is not detected among holobionts. It is that even if it were, the authors cannot really tell apart which are the cases of genuine selection where the holobiont is the focal level, versus cases of sorting, where the species that compose the microbiome are the focal

level. It is just another (more elaborated) way of expressing Moran and Sloan’s

worry that you can never tell apart cases of convergent host-microbiota phylogeny and cases of mere ecological filtering. I think reflecting their worry in terms of the distinction between sorting and selection is important, because it reflects one of the most genuine empirical worries to consider when selection for holobionts is being studied.

However, is this worry legitimate? From a conceptual point of view, it seems important to consider that not all causes of differential births and deaths of the individuals in a population are due to selection for their characteristics. Also, it is empirically important to single out the mechanisms, or causal agents, that are producing the selection, in case its effects are detected. But it is not so clear that the possibility of the effects of selection at the lower-level should be taken simpliciter as a knock-down argument against selection at the upward level, because doing so might bear the risk of a regression ad infinitum. In a hierarchical-nested view of the world, it is true that holobionts are made of different species, but so is it true that genomes are made of different genes, genes are made of different molecules, molecules are made of different atoms, and so on and so forth. If the possibility of distortion is taken ‘too seriously’, one might end up with no level where the effects of selection can be studied. There must be one level where one can stop, and the reason for stopping at that level must be somehow independent from the possibility of distorting effects from the lower level, or no empirical research could be pursued. Godfrey-Smith suggests the idea that we must start ‘afresh’ at each level where we aim to study individuality and the levels of selection, and Okasha suggests that we must break up with the necessity of a nested hierarchy in biology. I will pursue these two thoughts for holobionts in chapter IV and chapter V. For the moment, it is enough to say that the argument based on the distinction between sorting and selection must not be taken as an argument that rules out the possibility that holobionts are units of selection, and thus the possibility that some formulation of HCE is true. The purpose of the project is precisely to figure out which formulation of HCE is true, and in virtue of what it can be considered true.

6. Brief summary of chapter II: Three levels of criticism to holobionts and