Thanks for your suggestion.In fact ,I want to measure the extent of reliability of the bold response,So the effect estimates (beta values) of the five task fmri session were all the same.I have only 5 effect estimates (beta values) per subject.And the big problem is that the number of participants is four.If I calculate the mean 5 effect estimates (beta values) per subject for two way anova,it will lead to 5*4 two way anova. What is more,I want to compare the difference of effect of session and 5 task beta values between subjects. It is so bad that my repeated 5 task fmri is useless! how can I find a effective way to analyze this data structure? Thanks for your help very much!