Please see Rik Henson's excellent discussion of fMRI DesignEfficiency
MatthewBrett gave a talk at HBM2006 on avoiding errors in design and analysis. The text below is a slight expansion of the last few slides.
The importance of having an anatomical hypothesis
This is my (MatthewBrett) personal view for what it's worth.
- An interesting imaging experiment has to go further than where brain activity is, and tell us something about what the brain activity means in terms of how the brain works
- As a consequence, a brain imaging experiment without an anatomical hypothesis rarely tells us much about how the brain works
- It is very rare for an experiment to have no anatomical hypothesis atall, but an experiment with a weak anatomical hypothesis asks a question like "where in the brain is activity for task T"
- An strong anatomical hypothesis usually leads to a question like "what does activity in region R tell me about the contribution of R to the computations of task T"
How to make sure you have a strong anatomical hypothesis
- Sketch your ideal of how the paper will turn out while you are designing the study
- Draw a cartoon or your expected brain activation and brain activation if your hypothesis is not correct
- Make sure that you know how to use the difference between the activations above to be able to test your hypothesis
The following may be signs that you have a weak anatomical hypothesis:
- "Neural correlates" in your title!
- Introduction that concentrates on the task much more than the brain
- Long list of activations in your discussion with lists of previous studies activating these areas