44 SMITHSONIAN MISCELLANEOUS COLLECTIONS VOL. 76 



whence it must have been at least 3.4 km. high, to be seen over the 

 earth's curve — but how much higher cannot be guessed ; so the upward 

 Hne for Katmai is tipped with an arrowhead at about 5 km. 



In conclusion, thanks are gladly rendered to Dr. Harlow Shapley, 

 Director of the Astronomical Observatory of Harvard College, for 

 the privileges of the Observatory Library ; to the Librarian of Har- 

 vard University, and to Mr. W. B. Briggs, Assistant Librarian, for 

 privileges granted in the University Library ; to the staff of the 

 Boston Public Library ; and to Mr. L. Campbell and Miss L E. 

 Woods, of the Harvard Observatory, for assistance freely and often 

 given. 



Professor Alexander McAdie, Director of Blue Hill Observatory, 

 and Dr. Shapley, have been very generous in the matter of confer- 

 ences. But for selection of data, for methods, results and conclusions, 

 and for errors in this paper, the writer alone is responsible. 



SUMMARY 



The observational data on lunar eclipses, published by professional 

 astronomers and amateurs for the period i860 to 1922, have been 

 read, and have been collated, as far as possible, without reference 

 to any theory of the illumination of the eclipsed surface. 



A scale of brightness adapted to the data has been devised, con- 

 sisting of three grades ; grade 2, details on the eclipsed surface 

 visible with hand instruments or to the naked eye; grade I, details 

 visible with apertures of 2 to 6 inches, but not with less ; grade O, 

 apertures of 6 inches or more necessary to show detail. In assigning 

 grades the writer's bias has weighted positive statements of visi- 

 bility. 



This scale has been applied to all suitable data for each eclipse. 

 On account of curious discrepancies in the reports of both amateur 

 and professional observers, the grading has frequently been some- 

 what arbitrary. 



In the hope that the varying masses of air along the ray might 

 prove to have an effect in causing these discrepancies, the relative 

 air mass has been computed for each important report used in 

 grading, and the reports under each eclipse have been arranged in 

 order of increasing air mass ; but there seems to be no well-defined 

 connection. 



Assuming that the grades thus assigned are significant, the data 

 have been selected and arranged in several sequences, to test the 

 relations of various astronomical and accidental causes to the bright- 

 ness of the eclipsed moon. 



