52 ft. S. Woodward— Iced Bar Base Apparatus. 



Summary of results for length of _Z? 17 . 



No. of 

 Date. Measures. Mian length of bar. 



July, 1891 10 5 m — 11^-0 ±1^-4 



Feb. and March, 1892 20 5 —15 -2±0 -6 



April and May, 1892. 18 5 —11 -7±1 '6 



July and August, 1892 ... 20 5 —16 -8±0 *3 



Without desiring to discuss these data here, it may be said 

 that so far as is known at present the probable errors are 

 fairly trustworthy indices of the precision of the several results. 

 The range among them is but little in excess of the millionth 

 part of the bar's length, and is no greater than the probable 

 errors would lead one to expect. 



It may be stated also that the external air temperature 

 varied for the different groups of comparison from 5° to 40° C. 

 The average air temperatures for the two most important 

 groups, namely, those of February and March, 1892, and July 

 and August, 1892, were about 5° C. and 35° C. respectively. 

 Hence it does not appear that the bars in ice were affected 

 appreciably by the external air temperature. Finally, it 

 should be said that a systematic difference in the length of the 

 bar, according as the one or the other of its ends is to the 

 right in the Y-trough, is indicated by each group of compari- 

 sons. This difference appears to be an inequality of relative 

 personal equation of the observers at the two ends of the bar, 

 and may be due to the considerable inequality in widths of 

 the terminal graduation lines. 



Concluding remarks. — The question may be asked, does the 

 bar take the temperature of melting ice when fully packed in 

 it '? I am unable to give a decisive answer to this question at 

 present, but there appears to be no reason to suppose that it 

 takes a materially different temperature. Repeated observa- 

 tions on mercurial thermometers placed in the ice alongside 

 the bar show that they read zero within the unavoidable errors 

 of a few hundredths of a degree. That the bar assumes a fixed 

 length within very narrow limits is, it would seem, demon- 

 strated by the small range among the measures of the 100 m 

 comparator and the kilometer sections referred to above, and 

 especially by the recent work on the new comparator. This 

 latter work appears to justify the conclusion that the mean of 

 four determinations of the bar's length in terms of the Proto- 

 type meter, made in the manner described above, cannot have 

 a greater probable error than one micron. It is evident there- 

 fore, in view of the unavoidable errors of observation in such 

 work, and in view of the fact that the bar's expansion is about 

 55*" per degree Centigrade, that there is a very small margin 

 for change in the bar's length. 



