I am not hostile to the second idea (b) that levels of mechanisms exhibit a kind of ontological novelty, depending on how this novelty is understood and how one thinks that it is achieved. I suggest in the target article that a mechanism as a whole can do things that its parts (taken individually) cannot do. Lawnmowers mow grass; spark plugs do not. But I also claim that a mechanism as a whole cannot do things that its organized and interacting parts cannot do. This is because the behavior of a mechanism as a whole just is (or at least is ontologically intimate with) the organized interaction of its component parts. I don’t know how to make the notion of “ontological intimacy” precise here (though the behavior of a mechanism in context will surely supervene on the organized collection of interacting components in that context). In my hands, “ontological intimacy,” is meant to denote an exhaustive ontological grounding of the behavior of a mechanism as a whole in the organized interactions of its components in a given causal context, however that is properly to be unpacked. Everything has an ontic explanation in terms of the organized activities of the mechanism’s parts.
The term emergence strikes me as suspicious, and I hesitate to use it even in the context of levels of mechanisms, precisely because it suggests a severing of this ontological intimacy, a slide from the banal fact that mechanisms behave as they do because they are organized arrangements of interacting parts to the ontologically esoteric thought that something comes into being with causal powers ontologically inexplicable in terms of the organized interactions of the parts. Such claims about ontological novelty are almost always accompanied by claims that emergent things have a “downward” causal influence. I have labored to drive a wedge between levels of mechanisms and such ideas.
This connects with Martin’s description of Thompson’s view. According to Martin, Thompson requires that the parts in a system must be coupled and dynamically interacting to produce emergent properties and, further, that the properties of the whole should act “downward” on the dynamics of the components (2007). Take these in turn.
The idea of levels of mechanisms, by itself, requires only that the parts be organized and interact with one another; it does not require that the components interact non-linearly. Mechanisms with components that interact non-linearly are a subset of mechanisms more generally. It should not make any difference with respect to the novelty of higher-level properties that the parts interact non-linearly. Such dynamical interactions might make the behavior of the whole harder to predict on the basis of our understanding of the parts; but this is an epistemic, not an ontological, observation irrelevant to the ontological question to which I am responding. To be ontologically novel is not merely to be surprising.
Perhaps systems with dynamically interacting components are harder to idealize into separable interacting components; again, this would appear to be an epistemic issue. Thompson, at least in Martin’s description, is drawing a more fine-grained distinction than I draw. This is perhaps a terminological matter. The pressing question between us is whether non-linearity (or perhaps some other form of complexity) makes any ontological difference to the kind of thing that “emerges” from the “organized interactions” of the parts. I’m inclined to say no, but a full consideration of the issue would require a more detailed treatment than I can give it here.
Thompson also requires that the higher-level phenomenon should have “determinative” influence on things at lower-levels. “Determinative” surely cannot mean the existence of a universally quantified material conditional with the behavior of the mechanism as a whole as the antecedent and the organized interactions among the components as the consequent. Multiple realization precludes such an analysis. But “determinative” also cannot be understood in the causally productive sense in which something (such as charge or momentum) is passed from whole to part. And for reasons expressed in the target article, there are many widely shared assumptions about the independence of causes and effects that stand in the way of understanding such top-down relations in causal terms. This is why I suggest (as in Craver & Bechtel 2007) that the language of downward causation is misleading in the context of token mechanisms. We recommend that claims about downward causation are really shorthand ways of expressing an often complex web of constitutive (about relations between things at different levels of mechanisms and levels of realization) and causal claims (about things that are not at different levels from one another) about the system. However this is to be cashed out, the term “determinative” seems misleading.