222, ON THE ABSOLUTE MEASUREMENT OF HARDNESS. 
by trial that the values obtained in the last instance were more abnormal, 
but I did not use them. Fig. 2, supplied to corroborate these state- 
ments, represents the available part of the plate. It contains fully 28 
indentations, varying in size because lenses of different curvature were 
used, but it appears at a glance that there is room for many more expert- 
ments, 
WIGi oe 
VI. THE CONSTANTS AND SOURCES OF ERROR. 
The two chief constants of the apparatus are the scale value of the 
ocular micrometer and the ratio of the arms of the prime lever. The 
former was obtained in the usual way by comparing it with an objective 
micrometer. In my final and most accurate comparisons a tenth milli- 
meter scale was inserted in place of the plate p. Thus I found 
DAT seale parts ee 
with a deviatiun near the edges of the field presently to be discussed. 
In view of the form of the lever the ratio of the arms could not have 
been measured accurately by a direct method. Hence, I replace the 
plug Z by another containing a pin, on which a strong scale pan hang- 
ing in the main below the bedplate @ @/ could be pivoted. Observa- 
tions made for equilibrium between the new pan and the stirrup & 
(weights being suitably added to the former) showed the ratio in ques- 
tion to V = 9.8, with a probable error of + 0.01, 7. @, only about a 
tenth per cent of the value of V. It follows, therefore, that all measure- 
fod 
ments in terms of the ocular micrometer are to be divided by 27, and 
