248 FLUID MOTIONS •between them and the walls. The three pressures concerned are indicated on manometer tubes as shown, and the two differences of level representing head and suction can be taken off with compasses and referred to a millimetre scale. In starting an observation the water is drawn up in the discharge vessel, as far as may be required, with the aid of an air-pump. The rubber cork at the top of the discharge vessel necessary for this purpose is not shown. As the head falls during the flow of the water, the ratio of head to suction increases. For most of the observations I contented myself with recording the head for which the ratio of head to suction was exactly 2 : 1, as indicated by proportional compasses. Thus on January 23, when the temperature of the water was 9° C., the 2 : 1 ratio occurred on four trials at 120, 130, 123, 126, mean 125 mm. head. The temperature was then raised with precaution by pouring in warm water with passages backwards and forwards. The occurrence of the 2 : 1 ratio was now much retarded, the mean head being only 35 mm., corresponding to a mean temperature of 37° C. The ratio of Fig. 11. head to suction is thus dependent upon the head or velocity, but when the velocity is altered the original ratio may be recovered if at the same time we make a suitable alteration of viscosity. And the required alteration of viscosity is about what might have been expected. From Landolt's tables I find that for 9° C. the viscosity of water in •01368, while for 37° C. it is '00704. The ratio of viscosities is accordingly 1-943. The ratio of heads is 125 : 35. The ratio of velocities is the square-root of this or 1-890, in sufficiently good agreement with the ratio of viscosities. In some other trials the ratio of velocities exceeded a little the ratio of viscosities. It is not pretended that the method would be an accurate one. for the comparison of viscosities. The change in the ratio of head to suction, is rather slow, and the measurement is usually somewhat prejudiced by unsteadiness in the. suction manometer. Possibly better results would be obtained in more elaborate observations by several persons, the head and suction being.recorded separately and referred to a time scale so as to facilitate interpolation. But as they stand the results suffice for my purpose, showing, directly and conclusively the influence of viscosity as compensating a, change'.in otKe velocity. , .ies are small in comparison with that of sound.