Mechanisms and Laws: Clarifying the Debate - Mechanism and Causality in Biology and Economics - page 135

proceed in the absence of p-law statements. 8 Second, he claims that mechanistic
knowledge is dispensable in our search for p-laws. For instance, by using
randomized experimental designs, one can control for disturbing mechanistic
factors without knowing what they are. This latter argument is supposed to show
that although (explanatory) mechanistic models are “epistemologically dependent”
on p-law statements (in the sense that the former require the latter), p-law
statements are not “epistemologically dependent” on mechanistic knowledge (L4)
(some p-law statements can be discovered without relying on knowledge about
mechanisms). Finally, he argues that if our knowledge of laws did depend upon
knowledge of mechanisms, then we would face an “infinite (and vicious) epistemo-
logical regress” ( 2010 , p. 333). Because knowledge of mechanisms requires knowl-
edge of laws, our knowledge of laws and mechanisms would never ground out in
fundamental facts. Fortunately, Leuridan claims, we can know the p-laws without
knowing anything about mechanisms, and this blocks the regress.
Leuridan's first point about explanatory models derives from the above discus-
sion (see Sect. 3 ). If cs-mechanisms require (c)p-laws, then an adequate model of
the cs-mechanism requires (c)p-laws. Above, we reject the antecedent. Given that
not all mechanisms produce a behavior in a regular way (granted, many do), there
exist cases of one-off mechanisms (Salmon/Railton mechanisms) in which the
mechanistic model for the irregular behavior necessarily involves neither a macro
p-law statement nor a micro p-law statement. Similarly, in cases where a mecha-
nism behaves regularly even though this macro-regularity is sustained by micro-
irregularities, a mechanistic model might involve a macro p-law statement but not a
micro p-law statement.
But what shall we do about the cases in which biologists explain a general
phenomenon in terms of general facts about components and their activities?
Mechanists should not, and do not, deny the existence of such explanations. Instead,
mechanists deny that an explanatory model must be formulated in terms of
generalizations. 9 One would think that explanatory models of cs-mechanisms
must include p-law statements if one embraced a covering law (CL) model of
explanation, according to which explanations are arguments that subsume
descriptions of events under general law statements. No mechanist, however,
accepts the CL model of explanation (see especially Salmon 1984 and Craver
2007 ). The reasons are too widely known to be repeated here, and it would be

8

One might have expected Leuridan to defend the epistemic claim that one cannot learn about
mechanisms in the absence of p-laws. One might hold that one can test causal connections only on
the basis of regularities. Such a claim would be false, of course, as we might make causal
inferences on the basis of temporal succession or spatiotemporal contiguity, for example. Leuridan
might claim (correctly) that such inferences are fallible, but all inductive inferences are fallible,
including those involving p-laws.
9 Thus, we deviate from the central thesis Holly Andersen ( 2011 ) defends in her short response to
Leuridan's paper: “The existence of stable regularities in nature is necessary for either model of
explanation: regularities are what laws describe and what mechanisms explain” ( 2011 , p. 325).