Composition, Realization, and Constitutive Explanation: An Ontologically Egalitarian Account of Multilevel Mechanisms

(Note for connoisseurs: I consider this “aspect view” of realization an improvement over the “subset view,” although they share many virtues. As far as I know, the subset view is due primarily to Jessica Wilson.)
It also explains why higher-level sciences, such as psychology, are not autonomous from lower-level sciences, such as various branches of neuroscience. To understand a higher level, we need to know which aspect of its realizer it is and how it’s constitutively explained by its realizer. Thus, understanding a complex multilevel system requires identifying its mechanisms at all levels and integrating them into a multilevel mechanistic explanation.
There are several metaphysical themes that come up again and again in the theory of cognition and the philosophy of the special sciences more generally: reductionism and causal exclusion, autonomy and multiple realizability, scientific integration and multilevel mechanistic explanation. To do justice to these themes, we need an adequate metaphysics of composition and realization.
I begin the book with an account of composition and realization that avoids ontological hierarchy—the notion that some levels are more fundamental than others. I argue that levels are real, but they are not grounded in a fundamental level. How could this be, you ask? It’s because wholes are just invariants under some changes in their parts, while higher-level properties are invariant aspects of their lower-level realizers. Invariants (wholes) are neither more nor less fundamental than what varies (parts). Aspects are neither more nor less fundamental than what they are aspects of.
This “egalitarian” ontology is not reductionist in the traditional sense, because higher levels are not identical to lower levels. It’s also not antireductionist in the traditional sense, because higher levels are not completely distinct from lower levels either.
Composition is the relation between parts and wholes—the parts compose the whole. Realization, as I use the term, is the relation between properties of parts and properties of wholes—the properties of the parts realize the properties of the whole.
For example, consider a whole ship (e.g., the ship of Theseus). During any short enough time interval, enough of its parts and properties are retained to ensure persistence of the ship as such. Meanwhile, the properties of the ship are aspects of the properties of the parts of the ship, because the parts of the ship do many things (such as holding each other together) that the ship as a whole does not do but, collectively, the parts of the ship do everything that the ship does.
Thanks to Dan Burnston and Nick Byrd for this opportunity to introduce Neurocognitive Mechanisms: Explaining Biological Cognition (OUP 2020). This book is the culmination of my main research program over the last 20+ years; it offers a comprehensive foundation for the science of biological cognition. In these posts, I’ll outline some of the main themes of the book in a way that cuts across several chapters and, hopefully, adds value to what you can find in the book itself.
It solves the causal exclusion problem, because higher level properties are causally efficacious without being redundant.
This is also why a full constitutive explanation of a phenomenon is mechanistic; mechanistic explanations provide the greatest explanatory depth relative to merely “functional,” “dynamical,” or otherwise not fully mechanistic explanations. Of course, mechanistic explanations are also functional and dynamical, so there is no loss in going for mechanistic explanation.
This egalitarian ontology yields many philosophical benefits:
It provides a clear account of multiple realizability, because tokens of the same higher-level property type can be aspects of many different lower-level mechanism types.