338 Prof. Spencer U. Pickering on Delicate 



III. 



Kegistered. Theory. Difference. 



o o 



Immersed after being heated 10° 45 - 845 



cooled 10 45-755 45-804 --049 



heated 10 45763 45718 +045 



cooled 10 45-680 45717 -063 



heated 10 45-670 45-643 +-027 



cooled 10 45605 45"655 -'045 



heated 10 45-640 45-595 + 045 



cooled 10 45-585 45-635 -'050 



heated 10 45-630 



Mean -046 



Without exception, therefore, a reading when the thermo- 

 meter is falling to a given temperature is higher than when 

 it is rising to this same temperature, and the difference between 

 these two readings appears to be independent of the extent of 

 the foregoing fall or rise; a fact which will be shown more 

 clearly in a series of similar experiments with another ther- 

 mometer to be quoted below. The difference in the two 

 readings, however, is considerably influenced by the position 

 of the mercury in the stem ; from these experiments it would 

 appear to be directly proportional to the height of the column, 

 the total difference in each case being yooo" °f this height. 



Although this difference is very appreciable, it must be 

 remembered that it corresponds to a difference of very small 

 dimensions in the capacity of the bulb. The o, 015 measured 

 in the first series is equivalent to o, 0009 C, and represents 

 but 0*0000004 c.c, or but little over one millionth part of the 

 total capacity of the bulb (2*64 c.c), a quantity so small that 

 there is no wonder in its having hitherto escaped observation. 

 The considerable force necessary to push a column of mercury 

 through a long very fine tube is, I should say, quite sufficient 

 to account for this expansion of the bulb ; it is well illustrated 

 by the fact that a quick fall of but a small amount causes the 

 mercury to separate at the neck of the bulb. 



The question now arises whether this instrument, possessing 

 such serious imperfections, is really workable, or whether we 

 have not pushed thermometric delicacy beyond the limits 

 compatible with accuracy. 



The various determinations of the heat of dissolution of salts 

 which were made with this thermometer give the means of 

 answering this question. 



As shown in the previous communication, ±13'3 cal. is the 

 average error of each experiment when compared with a 

 duplicate one performed under the same conditions ; of this 

 7*59"cal. are error due to manipulation and 5*71 cal. to error 

 of reading. The manipulation-error will be independent of 



