34 PROCEEDINGS OF THE AMERICAN ACADEJMY 



from which it is obvious that, as the longest column of mercury read 

 in the gauges was about 100 mm., no appreciable error would be 

 introduced into any of these columns through errors less than 1° in 

 their temperature measurements. With the arrangement of screens 

 used, and taking the temperature by means of a thermometer dippino- 

 into mercury in a tube of similar dimensions and similarly exposed to 

 the gauge tubes, no error of this magnitude could enter into t. It is 

 also obvious that, as the gauges were placed side by side and under 

 similar conditions in every respect, the same proportionate error would 

 be produced in each, and this would therefore be largely eliminated. 

 Also, for the reason given in considering tlie effect of barometric errors, 

 the effect of errors in the determination of its temperature would be 

 wholly imperceptible, and indeed all temperature correction both of 

 the gauges and of the barometer might be omitted without seriously 

 affecting the accuracy of the results, for the reasons already shown, 

 and because the temperatures of the mercury columns differed but a 

 few degrees throughout the whole research. 



II. A K. 



Constant errors in the determination of A' may arise from any of the 

 sources of error in measurement of jy,, p.,, andjo.,, just discussed, which 

 can occur under the conditions which hold during the determination 

 of A'. These conditions have already been described in detail, and it 

 has been shown that 



when both capillaries are at the same temperature, conveniently that 

 of melting ice. By the method used in a previous paragraph on varia- 

 ble barometi'ic errors, it may be shown that a constant error even as 

 great as 1 cm. would have an imperceptible effect on A", and this effect 

 would be counterbalanced and eliminated by a similar and opposite 

 effect of this error in measurements taken with the same barometer to 

 determine y. To a less extent the same is true of constant errors of 

 the cathetometer scale, and it may be asserted that both of these 

 sources of error are practically eliminated from the results of this in- 

 vestigation. Further sources of constant error have been sufficiently 

 discussed in the paragraphs I. 1-7, and it should be added that, in 

 the packing of ice about the tubes, particular care was used, not only 

 to obtain thorough contact and surrounding at first, but to see that this 

 was maintained. 



