ACOUSTICS AND GRAVITATION. 97 



together with its temperature correlative and the valleys r, due to release 

 of air from solution with rising temperature, stand out prominently. Toward 

 the end of May and in June this diver also remained permanently sunk 

 for long intervals. A new departure now presents itself, as the air charge 

 of the diver increases (g/m decreases) relatively fast when sunk. I attribute 

 this to the change of atmospheric air brought about by open windows, together 

 with the low water-head needed to float the diver. In figure 115 the line 

 has been drawn through a succession of points which seem to suggest it. For 

 this d(g/m)/dt would be 30 and r = 0.00039, somewhat larger than the pre- 

 ceding value. The loss of air here recorded is about 5Xio~ 6 gram per day. 



80. Stems of small bore. As a final test on the limitations of the original 

 design, swimmers with stems of finer bore were selected. These were pycno- 

 meter flasks of the following dimensions : 



B'" f" 



Bulb: Diameter, 4. cm. Diameter, 4 cm. 

 Total mass: M = 15.648 grams M = 15.399 grams 



Stem length: 4.8 cm. 4.7 cm. 



Internal diameter: .08 cm. .08 cm. 



They were charged with atmospheric air and adapted for observation under 

 slight pressure excess over the barometric pressure. The stand-glass F" 

 was rather too narrow, so that this diver did not move with the requisite ease 

 for speedy observation. Moreover, both were too heavy for the highest deli- 

 cacy, owing to the thickness of. the stems. But the bore was very uniform and 

 as the observations were made in a semi-subterranean chamber in the basement 

 of the laboratory, in the summer, i. e., without artificial heat, it was thought 

 that a smoother set of observations might be expected. At least there was an 

 opportunity of separating the effect of temperature from that of atmospheric 

 pressure in its bearing on the solution discrepancy. 



The results are given in table 4 and adequately reproduced as to g/m by the 

 graphs in figure 1 16, in which the upper curve shows the atmospheric pressure 

 the lower curve the temperature. The graphs for g/m (here in thourands) 

 are intermediate, the upper one referring to F'" and the lower to B'". In the 

 initial observations (before June 30) the micrometric composition of the im- 

 prisoned air may come into play, but after the long rest during July and 

 during August, these have been practically eliminated. In general, the ob- 

 servations for F'" and B'" run in parallel and the effects of temperature and 

 pressure, though by no means absent, are more obscure than previously, but 

 of the same nature. 



The time loss finally has been decidedly reduced as compared with figure 

 115 and amounts in the two cases to the following mean values : 



F'": d(g/m)/dt = g.s per day; r = d(log (g/m)}/dt= io~ 5 Xio.8 



B'": =6.2 per day; =io~ 6 X 7.2 



Apart from the kinks in the curves, the loss coefficient is not, it would seem, 



altogether too large for the purposes in question, for it can be applied as a time 



