22 PROFESSOR C. V. BOYS ON THE 
itself was found to be gogo inch. In this way there was no doubt as to the setting 
of the cross-wire accurately to a tenth of the thickness of the line. I was not, 
however, always able, owing to the circumstances to which I have already referred, to 
leave the headstock set and clamped so accurately in position, that I could detect no 
want of symmetry in the microscope. Ina few cases I left it with a + or — error 
of one-tenth of a division as estimated by the eye, which error I entered in the note- 
book at the time, and before I knew what the reading of the Whitworth machine 
would be. These were taken by Mr. Harrison, the very skilful assistant in the 
Physical Laboratory at Cardiff, and he did not know what correction I had entered, 
or indeed, if I had entered any correction. His reading was then entered into the 
book also. In this way I hoped to avoid that spurious appearance of accuracy that 
is apt to result from knowing during a process of adjustment when the last setting 
has been reproduced. The temperature, of course, was frequently taken, but it only 
rose half a degree during the day, and to avoid as far as possible differences in 
temperature in the apparatus, the bars that were used were kept during the measure- 
ments upon the bed of the machine. As an exampie, I give the figures of the middle 
inch exactly as they were entered. The whole numbers represent yodgoths, and the 

decimals ypponths of an inch. 





Division on | PB: Reading and Reading and Reading and 
av used. Bn sem , 
Scale. | correction. | correction. correction. 
| 315° 
21 | 9-inch Be 3149 315-0 
33 8-inch 3133 313°4 313-4 
24 9-inch 3144 3144 3146 
32 | 8-inch { She aes lie 22028 Aes 





Temperature 16° ‘9 C. 
The distance between 24 and 33 on the scale found by taking the simple mean of 
the above figures and subtracting, is 99983 inch, or allowing for the gradual change, 
due probably to temperature variation, which seems to have been going on at this part 
of the day (about 1 p.m.) more particularly, the distance is 99986. 
A determination, made early in the afternoon, of the same distance by direct com- 
parison with the l-inch bar, the 8-inch bar remaining in the machine, gave as the 
length of the middle inch of the scale 99979. Since the bars are not guaranteed to 
be nearer to the truth than zodo 9th of an inch, the agreement between the 1-inch 
bar and the difference between the 8- and the 9-inch bars is better than might have 
been expected. Meanwhile, I may consider for the present that the true length of 
the middle inch is known with an accuracy of 1 in 10,000 at least, I have taken it . 
to be equal to 99980. 
