In two papers on multi-level selection, Heisler and Damuth (1987) and Damuth and Heisler (1988) provide an insightful discussion of emergence, and link it to the contextual approach to MLS1. They argue that emergence is relevant to the levels of selection, but not in the way that advocates of the emergent character requirement have thought. The crucial question is not whether a given character is emergent rather than aggregate, but whether the relation between the character and fitness is emergent, Damuth and Heisler argue.
The idea of a relation being emergent may sound unusual, but there is no reason why it should not make sense, at least to the extent that emergent characters make sense. For often in metaphysics, a distinction between types of property will be matched by an analogous distinction between types of relation. Damuth and Heisler say that a character-fitness relation is emergent if it 'cannot be accounted for' by a character-fitness relation at a lower hierarchical level. They argue that contextual analysis provides a way of identifying such emergent relations, in the MLS1 case. Since our treatment of cross-level by-products in MLS1 was also based on contextual analysis, it ties in with Damuth and Heisler's notion of an emergent relation.
To see this, recall that contextual analysis asks whether a particle's fitness depends only on its own character, or is also affected by a collective character (aggregate or emergent). If there is a 'collective effect' on particle fitness, then the statistical association between collective character and fitness remains after controlling for particle character. In Damuth and Heisler's terms, this means that the relation between collective character and fitness is emergent, because it cannot be fully explained by the relation between particle character and fitness. By contrast, where a particle's fitness depends only on its own character, the relation between collective character and fitness (if any) is not emergent, for it is a side effect of lower-level selection; in our terms, there is a particles collective by-product. Understood this way, the relevance of emergent relations to the levels-of-selection question is immediate.
One consequence of Damuth and Heisler's definition is that 'emergent relation' actually becomes a better defined notion than emergent character. For recall that contextual analysis treats all collective characters equally—they are all just possible sources of causal influence on particle fitness. So on the contextual approach, the distinction between aggregate and emergent characters is of no particular significance. A theorist who held that this distinction is not a principled one, or cannot be satisfactorily explicated, could still accept Damuth and Heisler's account of when a given character-fitness relation is emergent. For their account does not presume that the notion of emergent character is antecedently understood.
Interestingly, Damuth and Heisler restrict their discussion of emergent relations to multi-level selection of the MLS1 variety.7 But there seems no reason why it cannot be extended to MLS2. For cross-level byproducts can also arise in MLS2; putative examples from the literature were discussed in Chapter 3. Since Damuth and Heisler's concept of an emergent relation is intimately linked to the concept of a cross-level by-product, this suggests that emergent character-fitness relations can also occur in MLS2. The only way to deny this would be to argue that cross-level by-products cannot occur in an MLS2 framework, which is implausible. For this would be to say that, in principle, natural selection at the particle level cannot be the explanation for why collective character and fitness2 covary; but this is surely an empirical issue. Admittedly, cross-level by-products in MLS2 are difficult to analyse, for the reasons explained previously, but we should not deny their existence.
In a recent discussion, Gould (2002) describes what he calls an 'emergent fitness' approach to multi-level selection; he contrasts this with the 'emergent character' approach. The same contrast is drawn by Lloyd and Gould (1993), and Gould and Lloyd (1999). The inspiration for the 'emergent fitness' idea comes from Damuth and Heisler, though Gould and Lloyd's concern is with multi-level selection of the MLS2 sort. This is not in itself a problem, if I am right that Damuth and Heisler's notion
7 Damuth (personal communication) argues that the concept of an emergent relation is not well-defined in MLS2, and would not be of importance even if it were well-defined.
of an emergent character-fitness relation can be extended from MLS1 to MLS2. However, in one place Gould states that his emergent fitness criterion, as applied to species selection, implies that genuine species selection occurs wherever there is a species character such that 'the fitness of the species covaries with the character' (2002 p. 661). Lloyd and Gould (1993) make a similar remark (p. 598). But this is clearly wrong. The whole point of the notion of an emergent relation, for Damuth and Heisler, is to take account of the possibility that character-fitness covariance at a level might not be due to direct selection at that level. If Gould's description of what his 'emergent fitness' criterion amounts to is taken at face value, then the criterion is incorrect.
Was this article helpful?