350 



HENRY A. KOWLAKD 



diately after; some fifteen minutes, however, being necessary for the 

 thermometer to entirely cool. 



TABLE I. SHOWING CHANGE OF ZERO POINT. - 



The second 100 reading was taken after boiling for some time. 



It is seen that the zero point is always lower after heating, and that 

 in the limits of the table the lowering of the zero is about proportional 

 to the square of the increase of temperature above 25 C. This law 

 is not true much above 100, and above a certain temperature the 

 phenomenon is reversed, and the zero point is higher after heating; 

 but for the given range it seems quite exact. 



It is not my purpose to make a complete study of this phenomenon 

 with a view to correcting the thermometer, although this has been 

 undertaken by others. But we see from the table that the error can- 

 not exceed certain limits. The range of temperature which I have 

 used in each experiment is from 20 to 30 C., and the temperature 

 rarely rose above 40 C. The change of zero in this range only amounts 

 to 0-03C. 



The exact distribution of the error from this cause throughout the 

 scale has never been determined, and it affects my results so little that 

 I have not considered it worth investigating. It seems probable, how- 

 ever, that the error is distributed throughout the scale. If it were 

 uniformly distributed, the value of each division would be less than 

 before by the ratio of the lowering at zero to the temperature to which 

 the thermometer was heated. 



The maximum errors produced in my thermometers by this cause 

 would thus amount to 1 in 1300 nearly for the 40 thermometer, and 

 to about 1 in 2000 for the others. Eather than allow for this, it is 

 better to allow time for the thermometer to resume its original state. 



Only a few observations were made upon the rapidity with which 

 the zero returned to its original position. After heating to 81, the 



