Hello Andre,

Before you dive too deep into interpretations, I suggest you read the following articles (link below). Having a 91-level factor in your ANOVA is highly problematic statistically! Having such an ANOVA design dramatically inflates your error rates.

Article How to Get Statistically Significant Effects in Any ERP Expe...

Article Hidden Multiplicity in Multiway ANOVA: Prevalence and Remedies

Hi Brittany,

thank you very much. The first article you refer to helped me on my overall design. I changed from a 91x2x2 design to a non-parametric approach and a 14x2x2 design, which should be ok, if I apply Bonferroni correction. My question about the interpretation of between subject effects remains. It seems like SPSS is doing a two-way ANOVA on the grand mean of the within-subject factors here. As this does not make sence in my case, I decided to do one two-way ANOVA for each condition instead.

I'm assuming that with your new design (14x2x2) you still have a Group x Electrode interaction, and that there is no main effect of condition? If that is the case, you are correct that the ANOVA is finding some interactive effect regardless of the condition, and that there is no significant difference in overall coherence between conditions. You are correct that the means being tested in this interaction effect are averages across data from both conditions you entered in the ANOVA. Doing two separate ANOVAs (one for each condition) is not actually a relevant exercise, because the Group x Electrode interactive effect persists regardless of condition. If the effect were unique to one condition or the other, you would also have a significant Group x Electrode x Condition effect, and it sounds like that is not the case. What your data are telling you is that coherence generally differs between groups at certain electrode site combinations, but those differences exist regardless of the condition.