34 



REPORTS ON THE STATE OP SCIENCE. — 1916. 



when there is a first-rate clock-error available. There are thirty-four 

 cases of good or fair records of either P or S, including three cases 

 where an obvious S was recorded at the Observatory as P, but is easily 

 transferred : and there are only eight cases of some error at present 

 unclassed. The mean of the P errors is ±17^"1 and of the S errors 

 is +21«"3, part of which are undoubtedly due to errors in the tables. 

 If we omit errors, over 50* as in Table II. which follows, these become 

 + 17^"1 and ± 14:^'6. Now, this is very fair observing so far as it goes ; 

 but the important fact is that in one case only are both P and S success- 

 fully recorded (70'3, January 20). In seventeen cases P is recorded 

 and in seventeen cases S (the equality of the partition is remarkable), 

 but records of this kind which give no S-P are clearly not up to modern 

 requirements. 



Table I. 

 Records of Milne Seismograph at Edinburgh, 1914. 



In addition to these good or fair records there are the following, 

 some of which may be identified with other phases : — 



The following figures for some other stations will show how different 

 instruments compare in the present state of the tables : but it was soon 

 reahsed that the comparison is misleading, for many of the larger errors 

 are probably due to the tables, as the discussion in Section IX. indi- 

 cates. A more adequate discussion will therefore be given later. As 

 a rough method of treating the material at present, all residuals greater 



