Hi Markus, >[...] a model with two covariates. this > doesn't look right to me, because both variables (pain and anxiety) seems > to be regarded as equal and not one as the one which i want to ask the > correlation and the other as the variable i want to control for. I think this is simply a matter of using a contrast that tests the one you are interested in (e.g. zeros over all other columns, with a 1 corresponding to this column), while ignoring the other. Simply having the other in the design matrix will control for its effect. Hope that helps, Ged.