This paper proposes a new version of the orthogonality test where, the difference of spouses’ predictions over total net family income is regressed on variables that are common knowledge between the two. The test is consistent for the linex (and lin-lin)