Mechanistic Constitution in Neurobiological Explanations

Abstract

This paper discusses the constitution relation within the framework of the mechanistic approach to neurobiological explanation. It develops a regularity theory of constitution as an alternative to the manipulationist theory of constitution advocated by some of the proponents of the mechanistic approach. After the main problems of the manipulationist account of constitution have been reviewed, the regularity account is developed based on the notion of a minimal type relevance theory. A minimal type relevance theory expresses a minimally necessary condition of a given type that consists of a disjunction of minimally sufficient conditions of that type. Afterwards, the attained definition is successfully applied in an analysis of the logical structure of the explanation of spatial learning and memory in rats as a paradigm neurobiological explanation. The overall result is a more robust and more precise version of the mechanistic approach to neurobiological explanations.

Acknowledgements

I would like to thank Michael Baumgartner, Vera Hoffmann‐Kolss, and two anonymous referees of this journal for their valuable criticism and advice. An earlier version of this paper was presented in 2009 to the research colloquium of the Institute for Cognitive Science Osnabrück, to the theory colloquium of the Institute for Frontier Areas of Psychology Freiburg, to the Seventh International Conference of the Society for Analytic Philosophy in Bremen, and to the research seminar of the Cohn Institute for the History and Philosophy of Science and Ideas, Tel Aviv. I am grateful to the audiences at these events for their comments.

Notes

[1] The general view has received both a descriptive (Machamer, Darden, and Craver 2000, 23; Craver and Darden 2001, 118–120) and a normative version (Craver 2007, vii, defends both). The first contends that all paradigm neurobiological explanations mirror this general picture while the second demands that good neurobiological explanations must comply with the general picture.

[2] Instead of using the term ‘constituent’ Craver uses the term ‘component’. We take it that nothing substantial is lost by the replacement undertaken here for the sake of terminological unity.

[3] The terminology used to express the notion in question is actually not unified in the literature on the mechanistic approach. The terms frequently occurring are ‘composition’ (Machamer, Darden, and Craver 2000, 13; Bechtel and Abrahamsen 2005, 426; Craver 2007, 164), ‘constitution’ (Craver 2007, 153), and ‘constitutive relevance’ (Craver 2007, 139). However, it is safe to say that the authors intend these terms widely synonymously.

[4] Craver takes this more formal definition of constitution as expressing the same idea as the more colloquial definition mentioned in Section 1.

[5] Craver evidently believes that there are many levels in neurobiology (Craver 2007, 13, 164, 177). In his view, with each level there is associated a unique set of mechanisms and a unique set of laws governing these. That is, levels of mechanism are ‘… loci of stable generalizations, and consequently can be seen as local maxima of regularity and predictability’ (Craver 2007, 190; cf. also 164).

[6] Virtually all of the works taking up the central ideas of Machamer, Darden, and Craver 2000 use the notion of a ‘level’ in a similar way. As two examples, Bechtel and Abrahamsen point out that ‘[m]echanisms may involve multiple levels of organization’ (Bechtel and Abrahamsen 2005, 424), and Craver and Darden emphasize that ‘[m]echanisms in contemporary neurobiology are organized into multi‐level hierarchies’ (Craver and Darden 2001, 117).

[7] Note that the placeholders ‘X ₁’, ‘X ₂’ etc. are not contained in C. Rather they should be considered formula schemata quantifying over conjunctions of sentences in C.

[8] Note that ‘F X ₁ → G’ is not a formula in C as it contains the symbol ‘X ₁’ and as it does not express a complete sentence.

[9] Chalmers (2000) has proposed a minimality condition for neural correlates of consciousness as well. The analogy to the present case should be obvious. As a side remark, note that apart from the monotony problem, the minimalization requirement disarms a variant of a general problem discussed in Haugeland (2000, 212–214). Suppose a TV set is divided up into 100,000 cubes of equal size. Intuitively, the cubes are not all genuine constituents of the TV, because what’s going on in some of the cubes is completely irrelevant to the activities of the TV. If Definition 1 were applicable not only to neurobiological constitution but also to TV activity constitution, the minimalization condition would ensure that no irrelevant mechanisms, and hence no cubes hosting irrelant mechanisms, will be considered as constituting the activities of the TV.

[10] The parthood relation intended here is defined by the system GM; cf. Varzi (2009 [2003]).

[11] Due to lack of space, a full justification of condition (iii) cannot be provided here. However, to point out one further important aspect of condition (iii), note that a more formal description of conditions (ii) and (iii) would be the following: ∀x∀y(ϕx ∧ X ₁ y ∧ Rxy → ∃z(ψz ∧ Pxz ∧ Pyz ∧ ¬∃w(Pwz ∧ (¬∃Pwx ∨ ¬∃Pwy)))), where ‘R’ corresponds to the notion of a ‘co‐instantiation’ mentioned by condition (ii) and ‘P’ designates the parthood relation. If it is the case that the z referred to by the alternative formulation has no parts that are not a part either of the x or the y, then z must be identical to the fusion of x and y. Only this condition allows the desirable identification of ϕ X ₁ and ψ whenever ϕ X ₁ empirically turns out to be the sole minimally sufficient constitutional condition of ψ (cf. Section 4.4). The reason is that only then can it be inferred that ϕ X ₁ and ψ, in addition to being necessary and sufficient for another, are also co‐extensional.

[12] The proofs have been developed in Harbecke (in prep.).

[13] A question not discussed in this section is whether the types related by constitution must always be logically independent. If they need not be, it may seem as though constitution can be had much too easily. There are some preliminary reasons, however, suggesting a cautious response to this question. For one thing, constitution as defined by Definition 4 is a reflexive relation and, hence, sometimes relates logically dependent types. Secondly, if the hypothesis of an a priori physicalism should be defensible (cf. Chalmers and Jackson 2001), types such as ‘NMDA receptor activity’ and ‘spatial memory formation’ may turn out to be logically dependent after all. Since we wish to remain neutral on the question of a priori physicalism here, it is desirable to reject logical independence as a criterion for constitution.

[14] A case we have to exclude here is that ψ is just a conjunction of which ϕ is a conjunct. In this case it would still be true that Cϕψ and ¬Cψϕ, but ϕ and ψ should still be considered as inhabiting the same level.

[15] Note that this table is a copy of the one at the end of Section 2.

[16] For the notion of a difference test and the inference rules used to determine causal relevancies explicated, see Baumgartner and Graßhoff (2004, ch. 9). The rules determining causal relevancies have directly analogous reformulations that can be used to determine constitutive relevancies, although their precise explication must be left for future research. It is the latter rules that are presupposed here. As the rules allow us to determine constitutive relevancies in a finite number of steps (in difference tests under homogeneity), the regularity definition does not induce the epistemic regress diagnosed for the manipulationist approach.

[17] As mentioned in Section 4.2, Urban and Barrionuevo (1996) argue that it is not empty as there are forms of LTP in hippocampal area CA3 that are not constituted by NMDA receptor activation.