DIMENSIONS OP RINGS AND RODS OP IRON AND OTHER METALS. 219 



nearest to which the image was deflected was quickly noted.* The difference 

 between the two readings gave the deflection in scale divisions. The first of these 

 readings was believed to be correct within less than a quarter of a scale division ; the 

 second within half a scale division. It is clear that the results of two such observations 

 made under similar conditions and not unequally affected by any source of disturb- 

 ance, such as temperature changes, ought not to differ by more than one scale division, 

 and their mean should be correct to half a scale division. Now half a scale division 

 corresponds to O'OOOOOU mm., or about 1 three-millionth of an inch, which is equivalent 

 to l ten-mil lionths of the diameter of the iron rings and (approximately) of the 

 length of the cobalt rod, and to a little less than 1 ten-millionth of the length of 

 the rods of iron and nickel. This was the degree of precision aimed at. How nearly 

 it was actually attained may be gathered from the detailed notes of some of the 

 experiments contained in Tables III., IV., and V., which are given as specimens, and 

 from the diagram fig. 5. 



The tables consist of a transcript from the note-book of the actual figures recorded 

 in the experiments to which they relate. Nothing has been altered or omitted, and 

 nothing added, except the two final columns showing the results when reduced to 

 magnetising forces and ten-millionths of length. The figures printed in heavy type 

 are those which appear in Table II. Such of the figures as have been rejected (or, 

 rather, amended) owing to evident mistakes are printed in italics, t This occurs in 

 only two instances. 



The experiments selected are those which relate to the " Iron rod No. 1 " and the 

 cobalt rod. The first is chosen both because it goes further than any experiment 

 with the rings, and also because it affords an opportunity of comparing measurements 

 taken on two different occasions. The figures relating to the cobalt are given in full, 

 on account of the novelty and probable importance of the results obtained with that 

 metal. 



* This could only be done by a momentary glance, the circuit being necessarily closed for no more 

 than about half a second to avoid heating : it was, therefore, not attempted to read to a fraction of a 

 division. 



t The error which most frequently occurred in making a rapid reading was to mistake a long " 5-line " 

 for the long " 10-line " which came next above it on the scale. This mistake was generally detected and 

 rectified the moment it was made, but in three or four cases, all of which are specified in the text, the 

 wrong number was recorded. In these cases the observation was not actually rejected, but was corrected 

 by subtracting five. 



2 F 2 



