Mechanisms in biology

(02 Jul 2020) biology

A recurring meme on biology twitter is Oded Rechavi’s “What is the mechanism?”. In a response to one of his threads, someone linked to The Concept of Mechanism in Biology by Daniel Nicholson (2012) where Nicholson examines the notion of “mechanism” as it has been used historically, and contrasts it with current notions of mechanism in biology. I’ve recorded some of the more striking arguments in the paper.

[Relevant: Arjun Raj’s thread on Mechanisms in biology]

In a nutshell, Nicholson categorizes three meaning of mechanism in biology - mechanicism, machine mechanism, and causal mechanism. By these, he hopes to separate reductionist notions of mechanism arising from a sum-of-parts description of biology, to a more general epistemic notion of mechanism which may or may not correspond to “real” strucutures.

More crucially, it presents an understanding of the postulation of causal mechanisms in science that distinctly characterizes the mechanismic program today, namely that the appeal to causal mechanisms in scientific practice does not imply a commitment to the reductionistic agenda of mechanicism.

Issues at hand: “causal worldview” where effects are tied to causes vs a “machine mechanism” conception which assumes a reductionism to sum of parts.

A mechanism is not the phenomenon (the explanandum) itself, but the explanation behind the phenomenon (the explanans). This is the hard part:

To fully account for the explanandum phenomenon (Craver’s normative requirement for a good mechanismic explanation) it becomes necessary not just to specify, but also to explain how this organization is generated and maintained.

He recommends an abstraction of the living system

The conception of causal mechanism that I argue best fits biologists’ mechanism-talk is that of a contingent explanatory description which heuristically abstracts away the complexity of a living system sufficiently to describe some localized causal process within it which leads to the realization of some function of interest


One of the advantages of the epistemic view of causal mechanisms is that it is no longer necessary to postulate additional epistemic notions like ‘mechanism sketch’ and ‘mechanism schema’ to make sense of mechanismic explanations. Depending on the degree of abstraction, causal mechanisms may constitute what mechanismic philosophers call ‘sketches’, ‘schemas’, or ‘mechanisms’.

On the manupalibility theory of causation:

a part is causally relevant to the phenomenon produced by a causal mechanism if one can modify the production of this phenomenon by manipulating the behaviour of the part, and one can modify the behaviour of the part by manipulating the production of the phenomenon by the causal mechanism.

In a nutshell

If, as I suggest, the notion of causal mechanism is understood epistemically, then it can be characterized as an explanation where the explanans and explanandum are sorted out from the context of its formulation. However, if causal mechanisms keep being conceived as ‘real systems in nature’ (Bechtel, 2006, p. 33), it becomes exceedingly difficult to specify exactly what these ‘systems’ actually are, not to mention what they all have in common.

I disagree with this, if I’ve understood it correctly:

In other words, an ontic characterization of causal mechanisms can only increase its domain of applicability at the expense of sacrificing the concreteness of its formulation.


Consequently, the only way mechanismic philosophers could encompass all the different ways in which the notion of causal mechanism is employed in scientific practice would be to propose an ontic characterization so general and so abstract that it would be effectively vacuous.

I don’t understand this. Specifically, “increase its domain of applicability” is unclear - in what way can a causal mechanism be generalized when it is tied up to a phenomenon that it tries to explain? In my opinion each causal characterization is independent of more fine grained descriptions, and such scale independence presupposes that there is an intermediate causal characterization “linking” the different scales.


However, when we adopt an epistemic view of causal mechanisms, the tensions generated by the efforts to ontically reconstruct this causal mechanism disappear.

I need to read the cited paper, but I am uncomfortable with this notion that causal mechanisms are “epistemic all the way to the bottom” if we accept that there is a sort of a DAG of causal explanations through spatial and temporal scales.

The author insists on a stronger epistemic view that I am comfortable with, but I am willing to grant his final point

The ontic-epistemic dispute concerning the nature of causal mechanisms will not be settled by simply listing examples of the usage of ‘mechanism’ in the scientific literature, but by considering how best to make philosophical sense of the role played by mechanism-talk in scientific reasoning and explanation.