There is much that I applaud in Smaldino's article, in particular his emphasis on the importance of division-of-labour, role differentiation, and group-level functional organization in explaining the success of human (and other) groups. These features have of course been noted before, but I agree with Smaldino that they have been somewhat neglected by traditional models of social evolution (both genetic and cultural), which tend to focus on the conditions needed for an altruistic trait to spread in a group-structured population. These models are clearly important, but I share Smaldino's sense that they are not the whole story.
A theme that runs through Smaldino's article is the inadequacy of what might be called an “individualistic” or bottom-up approach to understanding the evolution of emergent group-level traits. Smaldino argues that such traits can only be expressed by groups, are not mere aggregations of individual-level traits, and evolve by a multilevel selection process that cannot be reduced to or redescribed as individual selection in a structured population. A number of points about this argument merit discussion.
Firstly, I agree that it is intuitive to regard some group-level traits, in humans and other species, as “emergent.” However, the emergent/aggregate trait distinction is notoriously difficult to make precise (cf. Okasha Reference Okasha2006). After all, any group trait is presumably ultimately dependent on the traits and behaviours of the constituent individuals in the group, in the sense that once the individuals' traits are fixed, this then fixes the group trait. (Philosophers call this a relationship of “part-whole supervenience.”) In cases where the pattern of dependence is particularly complicated, or where the underlying individual traits are heterogeneous, it may be natural to think of the group trait as “emerging” from them, but this is inevitably a matter of degree; thus, “emergent traits” do not constitute a sharply defined category.
Secondly, the question of how best to understand the evolution of a group trait, be it emergent or not, depends heavily on how the group trait gets transmitted down the generations. This in turn depends on whether reproduction occurs at the individual or the group level. In a human group, however functionally organized it is, the constituent individuals all retain reproductive capacities; so presumably, the group trait gets transmitted via individuals transmitting, to their descendants, the individual-level traits on which the group-level trait supervenes. (“Descendants” can be understood to include cultural, as well as genetic, descendants.) I say “presumably” because in principle, one could imagine a process of “group-level reproduction” in which some groups beget other groups, for example, by fission, and directly transmit the group trait to their (group) offspring. However, such a process seems rather unlikely in the case of human groups, and Smaldino does not appear to envisage it.
If it is right that the group traits in which Smaldino is interested are ultimately dependent on individual traits, and if no process of “group-level reproduction,” decoupled from individual-level reproduction, is being envisaged, then in principle it must be possible to understand the evolution of the group trait in an individualistic way, by tracking the evolutionary dynamics of the underlying individual trait(s) on which the group trait depends. Thus, I am unconvinced by Smaldino's argument that an individualistic approach is in principle impossible.
Thirdly, I do not entirely agree with Smaldino that extant theories of multilevel selection and kin selection are silent about emergent group traits. The modelling approach known as “contextual analysis,” for example, introduced by Heisler and Damuth (Reference Heisler and Damuth1987), explicitly treats an individual's fitness as a function of its own traits and the traits of the group to which it belongs, which can include “emergent” traits that are not mere aggregations of individual-level traits (cf. Frank Reference Frank2013; Okasha Reference Okasha2006). However, Smaldino is right that the majority of models of multilevel selection do not explicitly incorporate emergent group-level traits into their formal analyses; the “group traits” that feature in such models are typically the frequencies of individual traits (or genes) within the group. But because all group traits, emergent or not, must ultimately depend on individual traits, one cannot assume that such models are incapable of shedding light on the evolution of the group traits that interest Smaldino.
Finally and relatedly, Smaldino argues against the widespread view that multilevel selection and inclusive fitness (or kin selection) models are essentially equivalent because the evolution of any (individual) social trait or gene can be expressed using either. Smaldino argues that this equivalence breaks down when a “between-levels” perspective is adopted, which takes suitable account of emergent traits. I remain unconvinced by this for the reasons given above, but I agree with Smaldino that social evolution theorists have been too quick to endorse the equivalence of multilevel and kin selection. Although it is true that the two are formally or predictively equivalent, in that both yield identical conditions for an allele to increase in frequency, it does not follow that they are causally equivalent. In forthcoming work, I argue that in some cases, multilevel selection yields a better causal representation of social evolution, which in other cases kin selection does better (Okasha, in press).
These critical points notwithstanding, Smaldino's article offers many valuable insights into what an improved theory of human cultural evolution might look like.
There is much that I applaud in Smaldino's article, in particular his emphasis on the importance of division-of-labour, role differentiation, and group-level functional organization in explaining the success of human (and other) groups. These features have of course been noted before, but I agree with Smaldino that they have been somewhat neglected by traditional models of social evolution (both genetic and cultural), which tend to focus on the conditions needed for an altruistic trait to spread in a group-structured population. These models are clearly important, but I share Smaldino's sense that they are not the whole story.
A theme that runs through Smaldino's article is the inadequacy of what might be called an “individualistic” or bottom-up approach to understanding the evolution of emergent group-level traits. Smaldino argues that such traits can only be expressed by groups, are not mere aggregations of individual-level traits, and evolve by a multilevel selection process that cannot be reduced to or redescribed as individual selection in a structured population. A number of points about this argument merit discussion.
Firstly, I agree that it is intuitive to regard some group-level traits, in humans and other species, as “emergent.” However, the emergent/aggregate trait distinction is notoriously difficult to make precise (cf. Okasha Reference Okasha2006). After all, any group trait is presumably ultimately dependent on the traits and behaviours of the constituent individuals in the group, in the sense that once the individuals' traits are fixed, this then fixes the group trait. (Philosophers call this a relationship of “part-whole supervenience.”) In cases where the pattern of dependence is particularly complicated, or where the underlying individual traits are heterogeneous, it may be natural to think of the group trait as “emerging” from them, but this is inevitably a matter of degree; thus, “emergent traits” do not constitute a sharply defined category.
Secondly, the question of how best to understand the evolution of a group trait, be it emergent or not, depends heavily on how the group trait gets transmitted down the generations. This in turn depends on whether reproduction occurs at the individual or the group level. In a human group, however functionally organized it is, the constituent individuals all retain reproductive capacities; so presumably, the group trait gets transmitted via individuals transmitting, to their descendants, the individual-level traits on which the group-level trait supervenes. (“Descendants” can be understood to include cultural, as well as genetic, descendants.) I say “presumably” because in principle, one could imagine a process of “group-level reproduction” in which some groups beget other groups, for example, by fission, and directly transmit the group trait to their (group) offspring. However, such a process seems rather unlikely in the case of human groups, and Smaldino does not appear to envisage it.
If it is right that the group traits in which Smaldino is interested are ultimately dependent on individual traits, and if no process of “group-level reproduction,” decoupled from individual-level reproduction, is being envisaged, then in principle it must be possible to understand the evolution of the group trait in an individualistic way, by tracking the evolutionary dynamics of the underlying individual trait(s) on which the group trait depends. Thus, I am unconvinced by Smaldino's argument that an individualistic approach is in principle impossible.
Thirdly, I do not entirely agree with Smaldino that extant theories of multilevel selection and kin selection are silent about emergent group traits. The modelling approach known as “contextual analysis,” for example, introduced by Heisler and Damuth (Reference Heisler and Damuth1987), explicitly treats an individual's fitness as a function of its own traits and the traits of the group to which it belongs, which can include “emergent” traits that are not mere aggregations of individual-level traits (cf. Frank Reference Frank2013; Okasha Reference Okasha2006). However, Smaldino is right that the majority of models of multilevel selection do not explicitly incorporate emergent group-level traits into their formal analyses; the “group traits” that feature in such models are typically the frequencies of individual traits (or genes) within the group. But because all group traits, emergent or not, must ultimately depend on individual traits, one cannot assume that such models are incapable of shedding light on the evolution of the group traits that interest Smaldino.
Finally and relatedly, Smaldino argues against the widespread view that multilevel selection and inclusive fitness (or kin selection) models are essentially equivalent because the evolution of any (individual) social trait or gene can be expressed using either. Smaldino argues that this equivalence breaks down when a “between-levels” perspective is adopted, which takes suitable account of emergent traits. I remain unconvinced by this for the reasons given above, but I agree with Smaldino that social evolution theorists have been too quick to endorse the equivalence of multilevel and kin selection. Although it is true that the two are formally or predictively equivalent, in that both yield identical conditions for an allele to increase in frequency, it does not follow that they are causally equivalent. In forthcoming work, I argue that in some cases, multilevel selection yields a better causal representation of social evolution, which in other cases kin selection does better (Okasha, in press).
These critical points notwithstanding, Smaldino's article offers many valuable insights into what an improved theory of human cultural evolution might look like.