390 PROCEEDINGS OF THE AxMERICAN ACADEMY 



The following include errors of the class (c) : — 



From T""- — C. S. (meter) r= ±OMfi 



rpa,_p^a, « r=±1.44|t4 



a S. — R^^' (yard) r=± 0.99 /* 



r—± 0.49 II 

 Its we have, — 

 e=±0.07/* 

 e= ±0.06(M 

 e= ±0.06/* 

 e= ±0.09(M 

 e= ±0.22 |t< 

 6= ±0.20jt« 

 e= ±0.15|M 

 e = ± 0.07 jM 



The only conclusion which can be safely drawn from these results 

 is, that the errors due to temperature are between three and four 

 times as large as the accidental errors of observation. The errors 

 due to imperfect focus may be considered as eliminated in a long 

 sei'ies of observations. 



The writer regards the values given for the probable errors of the 

 final results as entirely illusory. The fact that the formula for proba- 

 ble errors takes no account of constant errors, needs to be strongly 

 emphasized in this connection. Errors of classes (h) and (c) are 

 probably of this sort. 



Sdbdiyision op Standards. 



The subdivisions of the various standards were made with some- 

 what unequal precision. In the transfers from one standard to another, 

 the errors of subdivision are in a general way reduced ; but I have 

 thus far found it impossible to reduce them to zero when the system 

 of corrections applied, includes a change in the entire length. 



