I actually can read the covariance scores on the plot. I just looked at it and can identify the high scores for relation with the class variable. At this point I may pick a cut off value of say .0999 and take all the variables relating better than this and try a next step with this. These would be at´s 1,2,3,4,5,6,9,11,17,21. This is a rough covariance tree to decide which variables look promising for use in the model. Using a covariance score cut off of .199 I would use at´s 1,2,3 only. This makes sense of the data which also showed structure when looking at the first at´s in ggobi. So I think this is progress.