OF ARTS AND SCIENCES. 887 



due to accidental errors of observation, on account of the limited 

 number of comparisons made. 



It should be stated also, that, judging by the color of the metal, the 

 composition of iraud Q^ differs from that of Q., and Q^. 



It now only remains to describe some observations which were made 

 to determine the extent to which the variations in the length of the 

 standards under variations of temiierature correspond with the com- 

 puted values from the indicated readings of the thermometer when 

 placed upon the upper surface of the bars. 



First, a series of comparisons were made with a rising temperature. 

 When reduced to 16°.67, the value of B.^— IF should be 522.9 div. 

 The following are the observed values. 



It therefore appears that, for a rising temperature of about 1° for 

 each hour of time, the length derived from the reading of the ther- 

 mometer will be from 2 jx to 3 fx too short. 



The lengtli of W was then compared with an end-meter in melting 

 ice with a falling temperature. The temperature of the cnmparing- 

 room had for several hours remained at 15°. 4 C. At 8h. 27ra. A. M. 

 both windows were opened. The relation *S' — W had been previ- 

 ously found to be — GI8.3 div. The comparisons were now continued 

 every three minutes with the following results. 



Time. S — IV Deviation Time. S— TV Deviation 



A.M. 8612 at 16^67 from -CIS 3 div. a.m. 8012 at 16' 67 from —618.3 div. 



h. m. o div. div. /j. h m. o div. div. ju. 



830 14.55 —G34.2 +15.9 =+ 8.0 8 54 11.07-681.0+127=4-6 4 



8 33 14.25 —033.0 +14.7 =+ 7.4 8 57 11..39 —644.7 +26.4 =+13.3 



836 13.67 —645.6 +27.3 =+13.7 9 11 10 —640.2 +21.9 = + 11.0 



8 39 13.41 — G39.8 +215 =+10.8 9 3 10.94 -0.36.0 +17.7 =+ 8 9 



8 42 13.07 —630.3 +21.0 =+10.6 9 6 10.95 -023.3 + 5.0 =+ 2.5 



8 45 12.76 —646.1 +27.8 =+14.1 9 9 10.90 —023.5 + 5.2 =+ 2.6 



8 48 12.57 —039.9 +21.6 =+10.9 9 12 11.00 —615.9 — 2.4 =— 1.2 



8 51 12.19 —643.1 +24.8 = + 12.5 



It appears, therefore, that the measured length will be found to be 

 too great under a rapidly falling temperature by about lO/ixor 12 fx 

 for at least half an hour after the change of temperature has occurred. 



