F 



► 



OF ARTS AND SCIENCES. 351 



probable errors deduced by tlie metliod of least, squares generally 

 prove too small, whenever we can bring them to the test of different 

 methods : this, of course, is due to the presence of unknown constant 

 errors peculiar to each method. 



Yours very respectfully, 



Chas. a, Schott. 



Carlile p. Patterson, Esq., 



Superintendent U. S. Coast Survey. 



I 



