62 J. M. Baldivin: 



Each of the two standard yards being compared was supported, 

 by a system of eight rollers, connected in groups of four, equi- 

 spaced as described on p. 629 of the Phil. Trans. Vol. 147, the 

 interval being 38/^/63 inches. The main support of each system. 

 of four rollers was a casting resting on three screws, the points 

 of the screws being fixed relative to the tables by a point slot 

 and plane arrangement. 



With the limited means at my disposal it was impossible to> 

 have a constant temperature bath, but provision against rapid 

 change in temperature was made by enclosing the whole of the 

 supporting tables and the standards in a box, of which the moving; 

 table formed the bottom, the sides and top being wooden frames 

 with panels of zinc outwards, and thick strawboard inside. The 

 top was in three sections, to leave space for the microscopes- 

 to pass through. Two thermometers were supported horizon- 

 tally midway between the standard bars. Throughout the whole- 

 comparisons the greatest care was exercised to eliminate the 

 effect of any progressive change, and the bars were measured, 

 in every arrangement. Thus in comparing two bars, A and B,. 

 eight series were made. 



North/South: A/B, A/g, v/a, V/B, B/y, B/A, g/A, a/v, 

 so that any constant difference in temperature caused by the-, 

 presence of the observer, who always was to the north, should 

 have no effect on the final result. A series consisted of eight 

 sets, the pointings in a set being in the order a, b, c, d, d, c, b, a 

 where a, b, are the terminal lines on one bar, c, d those on the 

 other bar. In the sets the first pointing was made on each line 

 in turn. A series occupied about half an hour, and during this- 

 time the temperature of the thermometers in the box rose about 

 0.3 °C. At the close of a series, the bars were placed in position 

 for the next series, and a minimum time of about two hours 

 elapsed before the next series was started. It is hoped that 

 with the precautions observed any difference in temperature is- 

 entirely eliminated from the final mean. 



There were thus in all 64 comparisons between any pair of 

 bars, and in each comparison eight pointings were made, ar- 

 ranged symmetrically so as to eliminate any linear progressive 

 change. There is no need to give full details of the readings; 

 it will suffice to state that in no case did the difference between 

 the extreme readings in the 32 comparisons of a group of four 

 series exceed .00020 inch, this including all sources of error 



