A rigorous geometrical proof is given for the statement that meridian observations of declination involve the Earth's axis of figure, so that nutation should be computed for this axis and not for the instantaneous axis of rotation as has always been done hitherto. A pair of observations of a star at upper and lower transit is analysed to show that the classical expressions for the fundamental declination, and the latitude of date, which had originally been acceptable, needed amendment after the work of Oppolzer, but become valid again if the nutation is changed as suggested.