Recall that in a multi-level scenario, where particles are nested in collectives, we can distinguish between emergent and aggregate characters of the collectives (cf. Chapter 2, Section 2.2.1). Many theorists have suggested that the aggregate/emergent distinction is somehow related to the levels-of-selection problem (Vrba 1989; Nunney 1993; Maynard Smith 1983). One version of this idea says that genuine collective-level selection, that is not reducible to selection at the particle level, can only operate on collective characters that are emergent. The intuitive force of the emergent character requirement is clear. Emergent characters are often complex, adaptive features of collectives, which it is hard to imagine evolving except by selection at the collective level. By contrast, aggregate characters are mere 'sums of the parts', so are natural candidates for arising from lower-level selection; certainly, they do not make convincing examples of collective adaptations.
One objection to this idea was noted in Chapter 2: aggregate characters often play a role in formal models of multi-level selection. This objection is not decisive, since formal models may not accurately reflect what happens in nature. But a deeper objection is that the emergent character requirement conflates product with process (cf. Williams 1992; Sober and Wilson 1998). The causal process of natural selection, at whatever level, must be distinguished from the products it can generate, such as complex adaptations. Emergent collective characters may be evidence of selection at the collective level, but should not be made preconditions of it; as Williams says, this is to put the cart before the horse (1992 p. 26).
This conclusion is bolstered by the analysis of particles collective by-products in Chapter 3. That analysis aimed to make explicit the idea that a character-fitness covariance at the collective level might be a byproduct of particle-level selection. The emergent character requirement can be thought of as a rival way of identifying particles collective byproducts; it says that the covariance at the collective level can only not be a by-product if the character in question is emergent. But our analysis implies that this is wrong. In both MLS1 and MLS2, whether a given character-fitness covariance is a cross-level by-product has nothing to do with whether the character is aggregate or emergent; these questions are independent in both directions. Knowing whether Z is aggregate or emergent tells us nothing about whether non-zero values of Cov (W, Z) or Cov (Y, Z) are by-products of lower-level selection, nor vice versa.
However, this does not close the case on emergent characters altogether. For as is clear from the causal graphs in Chapter 3, our analysis helps itself to the notion of a 'causal connection' between a character and fitness; it uses this notion to explain how cross-level by-products can arise in multi-level settings. It could be argued that where aggregate characters are concerned, it is impossible for there to exist a causal connection between the character and fitness, of either particles or collectives.1 So the fact that the previous chapter's analysis deems the aggregate/emergent distinction irrelevant is beside the point; for it
1 Lewontin (2000) makes a remark to this effect in the course of arguing that the average value of a trait in a group is not a real property of that group. He writes: 'averages are not inherited, they are not subject to natural selection; they are not physical causes of any events' (p. 33).
reaches that verdict by assuming that aggregate characters can be causally efficacious, which begs the question.
This is a coherent line of argument, but I do not think it is correct. The idea that a collective character has got to be emergent, if it is to causally influence the fitness of either the collective itself, or its constituent particles, does not seem plausible. If this were true, it would be a substantial metaphysical thesis in need of explanation. In general, the fitness of a biological unit can be affected by any of its characters, and by the environment. To single out some subset of these characters—the aggregate ones—and declare them incapable of causally affecting fitness is to claim a priori knowledge of what appears to be an empirical matter. Substantiating this claim would require a precise account of how to draw the aggregate/emergent distinction, and some indication of the source of the alleged difference in causal potential between the two sorts of character.
In Chapter 3, I suggested that the emergent character requirement stems from confusing two questions. The pertinent question is whether a given character-fitness covariance is a side effect of selection at a lower level, not whether it is a side effect of some causal processes or other at a lower level. If these two questions are not kept distinct, it becomes tempting to appeal to emergent characters to try to block the superveni-ence argument. Distinguishing the questions thus removes one possible motivation for thinking that genuine collective-level selection can only operate on emergent characters.
These considerations all suggest that the aggregate/emergent distinction is not of fundamental importance for the levels-of-selection question, a conclusion reached previously by Lloyd (1988), Grantham (1995), and Damuth and Heisler (1988). However, in one respect the emergent character requirement is on the right lines: it stems from a realization that a character-fitness covariance at a given level may not reflect direct selection at that level. This point is fundamentally correct, even though appealing to emergent characters is not the right way to accommodate it. Below I examine two further suggestions for how to identify the level(s) of selection that stem from the same realization; in a way, they constitute refinements of the emergent character requirement.
Was this article helpful?