Neuroimaging data: Individual or group?

Traditionally, it seems that the power in your study is directly correlated with your sample size (bigger sample size=more power). Along those lines, some people focus on the generalizability of the results in one study. What are the pros and cons for analysis at the level of individual participants compared to group level? Can anyone elaborate on this or point me toward relevant literature?

The question can be rephrased to - what do I want to generalize over.

  • If you only study individuals you can generalize over behaviour of a given individual (for example predicting their future responses to stimuli).
  • If you study groups of individuals you can generalize across people (for example predicting behaviour of a new unseen participant).

I don’t know of a good methodological comparison of the two approaches (not sure how it would look like), but maybe others can help.