Why an fMRI Scanner Can’t Answer the Neuroscientist’s Question

This week’s Monday Workshop featured Ann Johnson and her stimulating presentation “Code is Fluid: Mapping the Circulation of Computational Fluid Dynamics Among Knowledge Communities.” As one might assume, she introduced her concept of “Knowledge Communities” (KCs) through a case study on computational fluid dynamics (CFD). Her fascinating integration of history and philosophy of science is methodologically of great interest to anybody studying the neurosciences (and so much more), and with regard to content, it is the computational part (software, models and animations) which struck me as relevant to my project.

Her KCs reminded me of Ludwik Fleck’s “thought collectives”: She defines these as fluid, informal and plural communities. Just like in Fleck’s framework, where one can be a member of several thought collectives at a time, Johnson’s KCs aren’t always easily distinguishable from each other and not necessarily institutionalized. Basically, her KC is a group of people who are interested in the same problem and share a similar approach to it. Important in this connection is: The KC is also an actors’ category, i.e., feminists and neuroscientists grappling with the question of the nature of sex/gender differences in the brain would not form a single KC, because neither are feminists parts of the neuroscientific solution set to gain more insight in male and female brains, nor are neuroscientists the feminists’ hope for an explanation of socially constructed sex/gender differences that manifest themselves materially in plastic brains.

What I just laid out is an extended version of Johnson’s response to my question aiming at clarification of her concept. Again, this characteristic of KCs seems very similar to Fleck’s notion of thought-style-specific and incommensurable vocabularies or sets of questions. (She did not answer my question on what is specific to her framework as compared to Fleck’s, though. But I didn’t want to push her, because in the end it doesn’t really matter what we call it, as long it’s as terrific as she laid it out.)

One aspect that sounded unfamiliar: She described how individuals can turn away from a KC; when they are no longer interested in the problem treated within the community, for instance. The question is: How sticky are norms and ideas acquired in the former KC? Are they already embedded in the individual’s seeing and reasoning?—Take Andreas Roepstorff, for example. If I am not mistaken, he started out as an anthropologist, and while working on the neurosciences, he became a neuroscientist himself. How far was he converted, and how far does he still pursue anthropological questions, only with other means? To what extent do all ethnographers or interdisciplinary scholars become proselytes when working on the neurosciences, or to what extent do we still ask questions that are not in the least relevant to scientists in the field we are studying?

Johnson didn’t talk about this form of stickiness in researchers’ heads and senses, but she did so with regard to tools/apparatuses/methods/technologies/devices: Laws and assumptions are embedded in methods and apparatuses that are used in scientific endeavors, she argued. When one KC takes up CFD from another one, the adopting KC is blind to layered original assumptions implanted through construction, use and modification in the original field. She illustrated that due to taking up this new technology with all its embedded features, the methods in the adopting KC change, but the reasons for this change as well as the function of the device or method become opaque in the new KC.1

Why this insight is very important with regard to fMRI and other technical devices used in the neurosciences, will become clear if we follow Johnson one step farther: She explained that tools and models have the constructors’ “fingerprints all over from the very beginning, but they become opaque and invisible” even to their creators themselves, because layer is put over layer in development and use. Consequently, these technologies might only be able to answer questions their constructors have already prepared them to produce; i.e., the tool’s response to a question will be the answer that was (unconsciously) anticipated in the very beginning, in constructing the tool/method in a certain way.

And here it comes: After travelling to another field, the constructor of a technology differs from its user. It doesn’t sound surprising that experimenters sometimes don’t escape a certain regress when they construct, use and calibrate their own tools, and thus obtain only results they expected in the first place. If fMRI technology travels from some KC to a neuroscientific one, however, it might answer neuroscientific questions according to another field’s presuppositions embedded in it! And this brings us back to my beloved question: How much thought style is embedded in the machine—and whose thought style? Neuroscience, engineering, software developing, …?

Johnson answered the question for her own case study: When CFD is taken up by traffic analysists, she said, “it carries with it that kind of assumptions, that kind of physics.” Thus, an fMRI is not a neuroscientific tool, but a set of computer and engineering specialists’ blackboxed set of assumptions—or isn’t it?

1 To be more precise: As time goes by, the same process also occurs within the field the technology is built, even if it is only one individual interacting with the tool. The very act of creation and revision adds more and more layers to the core of the technology and veils initially built-in assumptions, she explained.


Leave a Reply

Your email address will not be published. Required fields are marked *