Tuesday, May 20, 2008

MRI - the new phrenology

Having just done a posting on MRI, I thought it appropriate to point to a discussion by Michael Shermer in his "Skeptic" column in the Scientific American on the misuse and over-interpretation of MRI data.

It is a reminder that seeing scans with highlighted (usually in red) areas where your brain “lights up” when thinking about X (money, sex, God, and so on) should not seduce us into buying the Swiss Army knife model of the brain, with specialized modules for vision, language, facial recognition, cheating detection, risk taking, spirituality and even God. There is the minor problem of reversing the causal inference:
...where people see some activity in a brain area and then conclude that this part of the brain is where X happens. We can show that if I put you into a state of fear, your amygdala lights up, but that doesn’t mean that every time your amygdala lights up you are experiencing fear. Every brain area lights up under lots of different states. We just don’t have the data to tell us how selectively active an area is.
As Patricia Churchland points out:
Mental modules are complete nonsense. There are no modules that are encapsulated and just send information into a central processor. There are areas of specialization, yes, and networks maybe, but these are not always dedicated to a particular task.” Instead of mental module metaphors, let us use neural networks.

3 comments:

  1. It is a reminder that seeing scans with highlighted (usually in red) areas where your brain “lights up” when thinking about X (money, sex, God, and so on) should not seduce us into buying the Swiss Army knife model of the brain, with specialized modules for vision, language, facial recognition, cheating detection, risk taking, spirituality and even God.

    So it's a fallacy to regard area V1 as a "specialized module" for vision?

    ReplyDelete
  2. Churchland's comment covers your point. Of course V1 is dedicated largely to vision, but it also links to, and can be activated by or send information to, virtually any other part of the brain. The argument is against autonomous modules.

    ReplyDelete