As I mentioned in two previous posts (here and here), I recently had the opportunity to attend several of the sessions of the “Future Directions in Genetic Studies” workshop at Washington University in St. Louis.
One of the speakers was Carl Craver of Washington University. Craver’s talk was titled, “The Reductionist Distortion of Behavioral Genetics,” a push-back against those who consider genetics to be essentially a gene-centered reductionistic enterprise.
Craver illustrated his talk with two examples of progress in genetic research in contemporary neuroscience: the study of mechano-sensation in C. elegans, and the study of learning in mice. Here his own synopsis of his basic points:
Although these examples involve the effort to explain behaviors in terms of lower-level mechanisms, they violate the assumption that behavioral genetics is or ought to be reductionist in the above sense. First, one must distinguish genetic intervention, in which genetic mechanisms are commandeered to intervene into higher-level systems implicated in a behavior, from genetic explanation, in which the genes are components in the mechanism’s behavior. Second, one must distinguish cases in which genes play a crucial background role in the development and maintenance of mechanisms underlying behavior without themselves being components in the mechanism from cases in which the genes are components in the mechanism. Third, even in cases where genes are components in the mechanisms underlying a behavior (as in the LTP case), their contribution is intelligible only in light of facts about higher-level mechanisms; explanatory reduction to genetics is not even a coherent possibility. The reductionist distortion of behavioral genetics introduces biases into our understanding of neural mechanisms that are distracting and misleading us in the effort to understand the neural basis of behavior.
According to Craver, genetics is not inherently reductionistic, nor should it be. In order to best make progress in the neurosciences, we need to take an integrated view of how genes contribute. Taking this integrative approach would serve as an “antidote to reductionism.”
At the beginning of his talk, Craver distinguished neurophilosophers (“who use findings from neuroscience to address traditional philosophical puzzles about the mind”) with Craver’s own work as a “philosopher of neuroscience (who “study neuroscience to address philosophical puzzles about the nature of science”). [After hearing Craver speak, I purchased his recent book, Explaining the Brain: the Mechanisms and the Mosaic Unity of Neuroscience (2007). Some of the material in this post is from Craver’s book].
According to Craver, neuroscience, which attempts to link various levels of explanation with double-sided arrows of causality, is not destined to be a reductive science. It is important to recognize this, because characterizing any science as reductionist slants science policy decisions in a nonproductive way. Characterizing a field as reductive affects the allocation of resources. For instance, what university departments and projects should be funded? To consider a field to be reductionist pushes away from what Craver considers the true goal of neuroscience: to improve the human condition-essentially, neuroscience is medical science, or at least it should be considered as such.
Craver described three versions of reductionism. For instance, consider “metaphysical reductionism,” which considers everything to be “physical.” Craver finds this approach “puzzling” and “esoteric,” as well as irrelevant to real life neuroscientists. Working neuroscientists actually do their work at this level. Consider also “explanatory reductionism,” which considers low-level explanations to be automatically blessed as privileged and fundamental. A third version is “methodological reductionism,” which advocates experiments in which we “intervene low and detect high.” For instance, we can alter a gene and strive to detect higher level behavioral changes. Craver nominated John Bickle as a good example of a methodological reductionist, in that Bickle tended to “leap from molecules straight to behavior.”
What are the alternatives to reductionism? One of these alternative approaches is the position Craver termed “autonomy, advocated by David Marr and Jerry Fodor.
The other major alternative approach would be integration, which actively seeks explanations that span multiple levels. Reductionism is “anti-integration.” Craver discussed some of the work with C. elegans, a small worm that “God created for neuroscientists,” in that it has only 302 neurons, it is completely sequenced and it is transparent. Craver recommended a website containing much information regarding the experimental work on C. elegans.
Using the integration approach, one will encounter many senses of “level,” when one considers possible “levels” of explanation. There are dozens of these, including levels of control, description, theories and implementation. Whenever Craver discusses explanatory integration, however, he’s referring to mechanistic levels, the more primitive levels of mechanisms being embedded within the higher levels.
According to Craver, Integrative neuroscience provides a clear sense of A) what is a “level” and B) how to integrate levels. In the actual practice of neuroscience genetics is frequently not reductive. This is true because many neuroscientists are not truly obsessed with reductionism, and “what you find often depends upon what you’re looking for.” In order to provide more impressive explanations, “trans-genetics needs to be an integrated study,” because reductionism brings on a series of troubling biases.
For instance, reductionism emphasizes the internal over the external. It considers a simplified version of the environment. It causes complexity at higher levels to be legislated out of existence. It encourages us to buy into a simplistic “additivity,” causing us to assume that what we have learned about tiny parts in isolation can be scaled up automatically when considering behavior as a whole; reductionism thus causes an oversimplification of phenomena. Reductionism also encourages “disciplinary myopia,” encouraging the assumption that phenomena can be described exhaustively from merely one perspective.
Craver offered additional arguments for using the integrative approach:
a. Brain systems are likely to have hierarchical organization, and the integrative approach is the most promising approach.
b. The integrative approach encourages exploration–interest in multiple levels of explanation.
c. The integrative approach is robust; its findings are more likely to withstand scrutiny when attacked from the multi-disciplines.
d. Craver’s “clinical argument”: interpersonal research is less intrusive and it provides ways to restructure the environment, ways that are often at least as promising as the reductive approach of manipulating the genes.
In sum, Craver makes a strong case that we should pitch our research at multiple levels of argumentation using the integration approach, therefore avoiding the tendency toward reductionism.
I’m in the process of reading Craver’s book, which I find to be written persuasively and with precision. This is also the way Craver speaks and it is thus delightful to listen to his ideas, which are based on his intimate familiarity with both the science and history of neuroscience.
Anyone who has been following my interests at this site likely knows of my interest in trying to pin down what it is that constitutes a meaningful “explanation” (not simply in the sciences, but, e.g., in ordinary conversation and political advocacy). I will be posting on this topic repeatedly in the coming months. Like Craver, I often have an aversion to reductionistic explanations (I should be clear–I don’t object to explanations that involve lower level mechanisms, only those that do this to the exclusion of higher level explanations and emergent phenomena). And like Craver, we need to do the sort of work Craver is doing, to keep explanations meaningful, to prevent them from devolving into expressions of pure emotion or dead-end reductionism, which comes in many forms.