Sektionsredner

I will compare and critisize two recent approaches to explanation and reduction in neuroscience: John Bickle’s ”ruthless” metascientific reductionism and Carl Craver’s mechanistic explanation. Both authors base their approach on actual scientific practice and appeal to the case of LTP (Long Term Potentiation) and memory consolidation for support, but the conclusions they draw are completely different: for Bickle it is ”ruthless reductionism”, for Craver multilevel mechanistic explanations. In the end, I will present my own approach, which is a kind of a middle way between the two positions.

The main idea of Bickle’s metascientific reductionism is that instead of imposing philosophical intuitions on what reduction has to be, we should examine scientific case studies to understand reduction. The reductionist methodology that thus emerges has two parts: (1) intervene causally into cellular or molecular pathways, (2) track statistically significant differences in the behavior of the animals. When this strategy is successful and a mind-to-molecules linkage has been forged, a reduction has been established.

According to Bickle, the case of LTP and memory consolidation is a paradigm example of an accomplished psychoneural reduction. He argues that the cellular and molecular mechanisms of LTP explain memory consolidation directly, setting aside psychological, cognitive-neuroscientific, etc., levels. What makes Bickle’s analysis ”ruthlessly” reductive is the claim that psychological explanations lose their initial status as causally-mechanistically explanatory when the cellular and molecular explanations are complete. He argues that accomplished lower-level mechanistic explanations absolve us of the need in science to talk causally or investigate further at higher levels.

Craver’s analysis is quite different: he sees LTP as a multilevel mechanism. What sets Craver’s position in direct opposition to metascientific reductionism is the thesis of causal and explanatory relevance of nonfundamental things. That is, he argues that there is no fundamental level of explanation, and that entities of higher levels can have causal and explanatory relevance. This is clearly in sharp contrast to Bickle’s view.

Craver’s defense of the causal and explanatory relevance of nonfundamental things relies heavily on Woodward’s recent account of causal explanation. According to Woodward, causal explanatory relevance is just a matter of holding of the right sort of pattern of active counterfactual dependence between explanans and explanandum, and invariant generalizations capture these patterns. Invariant generalizations are generalizations that continue to hold under some interventions or changes in their variables and background conditions. Invariance is a matter of degree: the range of interventions under which generalizations hold varies.

If we accept Woodward’s model of causal explanation, we see that Bickle’s claims about higher-level explanations losing their status as causally/mechanically explanatory are unwarranted. It is clear that higher-level explanations that are invariant will continue to be invariant even when cellular and molecular explanations are complete. In this sense, nonfundamental things can have causal and explanatory relevance even when the ”fundamental” cellular and molecular explanations are complete. In order to counter this argument, Bickle could argue that Woodward’s model of causation is insufficient, and that there is a stronger notion of causation that applies to the cellular/molecular level. However, a notion of causation like this does not emerge from scientific evidence, and Bickle is reluctant to provide philosophical arguments for his views.

I believe that Craver and Woodward are right in arguing that higher-level things can be causally explanatory, even when cellular and molecular explanations are complete. However, I think it is also crucially important to emphasize the ways in which lower-level explanations tend to be better than higher-level ones. I will argue that lower-level generalizations in neuroscience are generally more invariant than higher-level ones, and thus provide better or deeper explanations. I will show that also in the case of LTP, the generalizations at the cellular-electrophysiological and molecular levels are generally more invariant than the higher-level generalizations.

This leads to a restricted form of explanatory pluralism, according to which nonfundamental things can have causal explanatory relevance even when lower-level explanations are complete, but lower-level generalizations and explanations tend to be better than higher-level ones.