Because of deviations from a truly cylindrical surface at the capillary wall or to an abnormal orifice at one end of the capillary, the value of the coefficient,m,of the kinetic energy correction may be abnormally high with one direction of flow while it remains quite normal with the other direction. This variation may even cause the value of the principal instrumental constant,C,to be slightly greater in one direction of flow than in the opposite. This effect has been shown to be definitely greater than experimental error and is thought to be due to a slight difference in the ``effective'' capillary radius in the two directions of flow. This study shows the necessity of determining bothmandCfor each direction of flow through the capillary. This suggests the use of two viscosity formulas: &eegr;=Clp0t−mlC′&rgr;/tfor ``left‐hand data,'' and &eegr;=Crp0t−mrC′&rgr;/tfor ``right‐hand data.'' Although a well‐constructed viscometer may not require this unusual precaution, with it any commercially made capillary viscometer may be used with the highest possible accuracy.