April 6, 1883.] 



^CTENCEi 



241 



was moved to No. 6. Now, it will be seen from 

 the comparisons, that the influence of the cold 

 wall was such, that despite the care 

 with which the bars were protected, 

 and their nearness together, each one 

 had a different length for each posi- 

 tion that it occupied. 



When the comparisons had con- 

 tinued for two or three da3's, it was 

 seen that the range of individual re- 

 sults was greater than should be from 

 mere errors of comparison. The true 

 cause was not, however, suspected 

 until the set had been completed by 

 running the bars through each posi- 

 tion of the rack, returning to the ar- 

 rangement with which the set started. 

 The following table gives the relation 

 of the standard to the steel metres, 

 the differences being expressed in 

 microns (one micron equals one- 

 thousandth of a millimetre) . 



The variations from the mean are magnified 

 sixteen hundred times. 



"While the regularity of - the change is ap- 

 parent in this table, it is much more readily 



seen in a graphic projection, 

 in Fig. 2.. 



This is shown 



In this diagram the vertical' lines represent 

 the mean values ; and the points in the curves, 

 are obtained by using the differences from the 

 means as offsets to the right or left, for posi- 

 tive and negative differences. The greatest 

 length of each bar is found when the bar is 

 farthest from the outer wall, and the least length 

 when nearest it. If the differences be shown 

 graphically in parallel projection, the similarity 

 of the curves is still more forcibly shown. This 

 form is given in Fig. -3. 



That the variation of temperature within so 

 small a space so carefullj' protected should 

 have shown so marked an effect, was entirely 

 unexpected. It is susceptible of much 

 more accurate determination through 

 the bars themseh'es than bj' the use of 

 thermometers. Tn the case uudei", 

 consideration, the difference between 

 the extreme positions corresponds to 

 a difference of temperature of about 

 0.7° F. 



To lessen the effect of the influence 

 of the outer wall, other piers were 

 built at double the distance from the 

 wall, and a large screen was placed 

 between the comparator and the wall. 

 The screen was made of a framework- 

 of wood, covered on each side with 

 heavy paper. Another series of ob- 

 servations upon the same bars was 

 then begun. The results . show the 

 same influence to have been at work ; 

 but the effect is very much reduced.. ■ A graphic 

 representation is given in Fig. 4. 



