Hello neuroimagers [this is a long post, but please stick with me],
I am working on a set of analyses in FSL, and I want to make sure 1) I am entering the different variables/covariates appropriately to test my hypotheses and 2) I interpret the contrast maps correctly. I am using the FSL 5.0.9 FEAT GUI to set up these analyses.
Design: Participants receive acute pain stimulations during separate English and Spanish scanning runs. There are other events during the task (e.g. pain ratings, image viewing, etc.) but for the purposes of my questions, I'm focusing only on the pain during the task. There are 4 task runs, two in English and two in Spanish. Two runs of the same language are always consecutive, with starting language counterbalanced across participants (i.e. either EESS or SSEE, never ESES). Additionally, I am interested in looking at how participant's "cultural orientation" (a continuous variable reflecting how much they prefer Hispanic or US-American culture) might affect their responses to the pain in different languages. So pain vs. rest is the primary trial-level contrast of interest, language in which they receive the pain is a run-level variable of interest, and "cultural orientation" is a person-level variable of interest while counterbalance order (EESS or SSEE) is a person-level covariate I would like to control for.

I am currently conducting the analysis as follows:
1st level- I have the timing of the different events happening throughout the run in the design matrix being used to predict the neural activity associated with each event type (contrasted against the implicit baseline). Main contrast: Pain > Rest
2nd level- Here the results from the 1st level analysis are combined across runs for each subject, with run language being included as a 2nd level covariate. This results in 3 maps for each contrast from the first level: 1) map showing the mean activity across runs (subject mean; collapsing across languages), 2) map showing areas with greater activity during Spanish runs (Pain in Spanish > Pain in English), 3) map showing greater activity during English runs (Pain in English > Pain in Spanish)
3rd level- Combining second level contrasts across all subjects. Here I add two covariates: each subject's counterbalance order (EESS or SSEE; discrete), and each subject's "cultural orientation" score (continuous on a scale from -1.5 [more Hispanic] to +1.5 [more US-American] with a meaningful 0-point denoting equal identification with both cultures). Without considering the covariates, this analysis creates the same 3 maps as given by the second level, averaged across all participants. So I have one brain map showing areas with higher activity when receiving pain vs rest (grand mean), one map showing areas which show more pain-evoked activity in Spanish vs English (S > E), and one map showing areas which increase activity to pain more in English vs Spanish (E > S).

Questions:
1a. I can reasonably assume that subjects' responses to the pain stimulations will decrease from run to run (habituation effects). I think the inclusion of "counterbalance order" would account for habituation in the language contrasts. If I just add the counterbalance EV at the third level (coded as +1 for EESS or -1 for SSEE), does that achieve the goal of controlling for order effects when looking at the language contrasts? Or do I need to include a third-level contrast with the counterbalance variable to create maps of the areas affected by the counterbalance order?
1b. Should the "counterbalance order" covariate be orthogonalized with respect to the grand mean EV at the third level?
2. I hope to address "whether and how the strength or distribution of pain-evoked neural activity changes across languages". Am I appropriately addressing that question with the English>Spanish and Spanish>English maps at the third level (assuming I am correctly controlling for the counterbalance order)?
3a. Further, I want to ask "does the nature/strength of language effects on pain-evoked neural activity differ as a function of subjects' cultural orientation". I am including cultural orientation as a 3rd-level covariate and then requesting a contrast map of that EV (i.e. +1.0 for orientation EV and 0's for the grand mean and counterbalance EVs in the "Contrasts and F-tests" tab of the "Full Model Setup" window). Does that cultural orientation covariate need to be orthogonalized with respect to either the grand mean EV or the counterbalance EV?
3b. Since the 0 point of this orientation variable is already meaningful, I don't think I should mean-center it. Is that correct?
4a. Including a contrast for cultural orientation then produces an additional 3 "cultural orientation" brain maps from the third level analysis (one for the grand mean, one for S> E, one for E>S). Am I correct in interpreting the grand mean-associated cultural orientation map as those areas that modulate their increase in pain-evoked activity across the different levels of cultural orientation in the sample (i.e. main effect of cultural orientation)?
4b. Is it correct to interpret cross-level contrast maps (the effect of the cultural orientation covariate in the S > E and E > S contrasts) as addressing the research question in #3a above (i.e. the language-cultural orientation interaction effect)?
4c. Since FSL only shows areas of increased activity, would I need to make one contrast with a +1.0 on the orientation EV and a separate contrast with a -1.0 on that EV? As it stands now, I am interpreting the cultural orientation contrast map under the Spanish > English contrast as "those areas with increased pain-evoked activity in Spanish compared to English whose activity differs [increases?] as participants in the sample show greater US-American cultural orientation". Does that interpretation sound correct?

I would be happy to respond to any clarifying questions or provide more detail on the design or analysis. Any help or advice you can offer is greatly appreciated. Thank you