There are some basic conceptual problems hovering about the widespread use of brain scans as pictures of mental activity. As applied to medical diagnosis (for example, in diagnosing a brain tumor), a brain scan is similar in principle to an X-ray: it is a way of seeing inside the body. Its value is straightforward and indubitable. However, the use of neuroimaging in psychology is a fundamentally different kind of enterprise. It is a research method the validity of which depends on a premise: that the mind can be analyzed into separate and distinct modules, or components, and further that these modules are instantiated in localized brain regions. This premise is known as the modular theory of mind.

It may in fact be that neither mental activity, nor the physical processes that constitute it, are decomposable into independent modules. Psychologist William Uttal contends that rather than distinct entities, the various mental processes are likely to be properties of a more general mental activity that is distributed throughout the brain. It cannot be said, for instance, that the amygdala is the seat of emotion and the prefrontal cortex is the seat of reason, as the popular press sometimes claims. For when I get angry, I generally do so for a reason. To cleanly separate emotion from reason-giving makes a hash of human experience.

But if this critique of the modular theory of mind is valid, how can one account for the fact that brain scans do, in fact, reveal well-defined areas that "light up" in response to various cognitive tasks? In the case of functional magnetic resonance imaging (fMRI), what you are seeing when you look at a brain scan is actually the result of a subtraction. The fMRI is usually interpreted as a map of the rate of oxygen use in different parts of the brain, which stands as a measure of metabolic activity. But what it actually depicts is the differential rate of oxygen use: one first takes a baseline measurement in the control condition, then a second measurement while the subject is performing some cognitive task. The baseline measurement is then subtracted from the on-task measurement. The reasoning, seemingly plausible, is that whatever remains after the subtraction represents the metabolic activity associated solely with the cognitive task in question.

One immediately obvious (but usually unremarked) problem is that this method obscures the fact that the entire brain is active in both conditions. A false impression of neat functional localization is given by differential brain scans that subtract out all the distributed brain functions. This subtractive method produces striking images of the brain at work. But isn't the modular theory of mind ultimately attractive in part because it is illustrated so well by the products of the subtractive method?