Dear Joelle,
You could go with
1) a single condition to test for average activations for your visuomotor task and add parametric modulators to test for changes in activation over time.
1a) With values 1:1:10 and depending on the order of the PM you could test for linear (1st-order), quadratic (2nd-order), cubic, ... changes. Note that initially, the vertex is at trial 1 for all the PMs. By default, the PMs are serially orthogonalized on each other, so the 2nd-order PM tests for that part of the quadratic effect which can't be explained by the 1st-order PM and so on. You might want to disable the orthogonalization, see http://imaging.mrc-cbu.cam.ac.uk/imaging/ParametricModulations . If you haven't worked with PMs/orthogonalization so far you might want to look into a recent paper by Mumford et al. (2015, PLOS ONE, https://dx.doi.org/10.1371/journal.pone.0126255 ).
1b) To test for exponential, logarithmic, ... changes go with a 1st-order PM and add appropriately scaled values (= testing for *linear* relationships with the *exponentially*, *logarithmically*, ... scaled values). E.g. for a u-shaped quadratic relationship with the vertex in the middle of trials 5 and 6 you could go with a 1st-order PM based on values (n-5.5)^2, with n corresponding to the trial number
2) ten separate conditions, which allows to directly contrast e.g. trial (condition) 1 vs. trial (condition) 3, which should be more flexible (but also possibly more difficult to interpret) than 1).
Whether to go with 1) or 2) depends on your hypotheses. Early/middle/late would be another option, but this distinction is artificial and thus very prone to criticisms.
Best
Helmut
|