Thanks. I used a Brunner Macro LD_F2, which enables a nonparametric analysis of doubly repeated measures. I discovered that a regular repeated measures anova on the doubly aligned Friedman ranks does the job. E.g. in a 2x3 fully within design, I analyzed the aligned ranks from 1 to 6, with the 'repeated a 2, b 3' (SAS) statement.