*PC accounts

Fri 04 Apr 2014 11:48 AM

In a just-published article, Manolo Martínez tries to modify the Homeostatic Property Cluster (HPC) account so as to accommodate polymorphic species.* I have two comments about the relation between his discussion and my own work.

A. Martínez discusses my own attempt to solve the problem of polymorphism but substantially misdescribes what I had in mind. This might be the fault of my writing rather than his reading, but I want to clarify the point.

He summarizes my approach with this example:

1. We look around and find a female domestic cat, say.

2. We then look for the mechanism that is responsible for the fact that female

domestic cats keep recurring: it is sexual reproduction between cats.

3. Finally, we calculate the "scope" of this mechanism: it produces both

female and male cats.

That much gets me right. The general summary he gives, however, goes awry.

On my view, he writes, "once the mechanism is identified, it 'goes it alone'" and "everything that the mechanism that produces one morph produces should be counted in as a member of the kind." Yet this can't be correct! As he writes, "the solution to the problem of polymorphism cannot be to follow homeostatic mechanisms anywhere and everywhere, because homeostatic mechanisms can go places where we do not want species to go." To continue the example of the ongoing sexual interactions between cats: It produces mutants, too, but they shouldn't count as members of the species.

Martínez is entirely right to think that the causal mechanism can't go it alone. The problems with the idea go further than just mutants. The ongoing sexual interaction between cats also produces sleepless nights for apartment dwellers, when rutting cats in an alley keep them awake. But sleepy office workers the next day should not thereby count as members of Felis domesticus. Following out all the causal ripples produced by cat sex, the indiscriminate approach would end up including most of the planet.

So we should think that the properties of kind members still somehow count as part of the HPC, but in a way that allows for polymorphism. As I put it in my paper:

On my view the 'property cluster' of an HPC need not be a single list of characteristics which all members of the kind are even likely to possess. It will be whatever complex array of property regularities are sustained by the underlying causal process. These might be a single cluster (for an asexual species), distinct but overlapping clusters (as with ducks and drakes), or almost entirely distinct clusters (as with anglerfi sh and Osedax). What matters is the complex of related property clusters maintained by an underlying causal process.

The properties which get included in the complex for the kind are ones which are causally held together by the mechanism, but the mechanism doesn't have to go it alone. The kind is comprised by the complex of properties plus the mechanism. As I conclude, "What matters is the complex of related property clusters maintained by an underlying causal process."

So the properties involved determine whether some causal products count as part of the kind or not. Mutant cats might not count as cats, because of their divergence from typical features for any cats. Sleepy office workers don't count as cats and their properties don't get to be part of the property complex that characterizes the kind for systematic reasons, because of systematic constraints on how we construe biological kinds.

B. Martínez advocates generalizing the HPC account in terms of Informationally-Connected Property Clusters (ICPCs), where ICPCs are just defined in terms probabilistic correlations. The explicit requirement of causal connection goes away. The result is something like Matthew Slater's account of Stable Property Clusters (SPCs).

It may be that ICPCs or SPCs succeed as more general conceptions, but removing the requirement of causation comes at a cost. I have argued (several places) that it is helpful to distinguish between token-HPCs (in which all members are part of the same causal history) and type-HPCs (in which members might figure in distinct but sufficiently-similar causal histories). The advantage of token-HPCs is that we can start with some specimen members of the kind and try to figure out how far the boundaries of the kind go, relying on the constraint that any members will have to be historically connected to the specimen members. With type-HPCs, members might occur in distant and unconnected places, meaning that we have to worry not just about which products of the history count as members of the kind but also which histories count as the relevant kind of mechanism.

The ICPC and SPC accounts don't require causation and so can't draw this distinction in terms of whether the causal history is specified as a token or a type. Perhaps they can draw the distinction in other terms.


"Informationally-connected property clusters, and polymorphism"

Biology&Philosophy, March 2014

Published version (behind paywall)

Author's final draft (from his website)