408 Prof. S. U. Pickering on the 



the readings, calculated on the assumption that the alteration 

 in height of the column is directly proportional to the pres- 

 sure; these calculated values being deduced from the average 

 points given at the foot of the table. With the last-mentioned 

 instrument (65108) the effect of pressure on the bulb would 

 appear to cause a very regular expansion ; there are only 

 three observations which differ from the calculated values by 

 more than O'l millim., while the average difference amounts 

 to about 0*05 millim. only. With the other two instruments 

 the observations at the higher pressures cannot be much relied 

 on, since the pressures were ascertained by means of a Bour- 

 don's gauge, instead of a column of mercury as in the other 

 cases; and this gauge was afterwards found to be untrust- 

 worthy. Omitting these observations, we find that, in case of 

 No. 63616, where the coefficient of apparent expansion of the 

 bulb was twice as great as that of 65108, the error in the 

 readings is considerably larger, amounting to as much as 

 0'12 millim. on the average, while the grouping together of 

 the positive and negative differences is well marked. It is 

 only with No. 62839, however, which possessed a still greater 

 coefficient of expansion, that the differences become so large 

 as to render it quite impossible to attribute them to mere ex- 

 perimental error. As the pressure is increased above 500 

 millim., the bulb begins to contract far more than it should, 

 causing a difference of as much as 1*2 millim. between the 

 observed and calculated readings; it then contracts less, crosses 

 the line representing the calculated values and for pressures 

 from 1100 up to 1986 millim. (if these higher results may 

 be trusted) does not contract as much as it should do*. The 

 action of the bulb under pressure is evidently not regular. 



As the differences in the calorimetric results above men- 

 tioned were observed with the thermometers '616 and '08, 

 which behaved normally, or nearly so, under pressure, as well 

 as with other instruments with bulbs of considerably smaller 

 expansibility, it is impossible to attribute these differences to 

 the cause suggested ; at the same time, however, the present 

 investigation leads to results of considerable practical im- 

 portance. It is evident that where the coefficient of expan- 

 sion of the bulb is large, as with '39, irregularities in expan- 

 sion sufficient to introduce considerable errors may occur; 

 the bulb, when subjected to pressure, would appear to behave 



* Care was of course taken that the thermometer should not be read till 

 the column had attained a position of equilibrium. The top of the in- 

 strument was tapped throughout the experiments by means of the tapping 

 apparatus, and the column read at intervals of one minute till it was 

 found to be perfectly stationary. 



