ON SEISMOLOGICAL INVESTIGATIONS. 53 
all those marked doubtful) only 39, can be regarded as true readings 
of P; say 40 per cent. at most. With the other machines there are 
87 out of 134, or 65 per cent. 
Coming now to §, and correcting the results (which include those 
transferred from P) as for other instruments, we find 12 large errors; 
the others are distributed as below :— 
TaBxeE LY. 
38 Errors for S in Milne Seismographs tn 1918. 
Distance from Epicentre 
| oo — 40° — 80° — 130° All 
. m. m. | 
+3:3 to +3-7 — | 1 1 2 
+2-8 to +3-2 Sar Seda _— — 
+2:3 to +2:-7 | — — 3 3 
+1:8 to +2:-2 — 1 1 2 
esta ei-72 > PS = 1 1 
+0:8 to +1:2 — 1 2 3 
+0:3 to +0-7 2 2 2 6 
—0-2 to +0:2 | — 3 3 6 
—07to-03 | 9 — 2 5 7 
—12to-08 = — ~ 2 2 
Pepto 9-3 ak ist) vig 2 4 
| —2-2 to —1:8 —_— — 1 1 
—2-7 to —2:3 — 1 — 1 
The mean of the errors ig + 1‘1m.; for other instruments it was 
+0°73m. The ratio of these is about the same as in the case of P. 
But it will be seen that there are acceptable readings of S in 38 cases, 
whereas for the same earthquakes there are only 39 of P at most. 
It is usually considered that the Milne instruments show P but not S. 
The evidence here tabulated points to the conclusion that S is shown 
at least as well as P. It is true that the five earthquakes considered 
are large ones; but it might reasonably be argued that P should 
therefore have the better chance of asserting itself. It seems probable 
that in some cases P could be recovered from the records when it was 
realised that the reading formerly given was that of S. The important 
point is that without any great difficulty it can be settled when 
we have an § reading, for the cases of doubt are few. We may now 
give the 12 large errors excluded as mistakes; they are +35°2m., 
+11:9m.,:+10°3m., +91m., +8'7m., +86m., the smallest of 
which exceeds the maximum error (+3°5 m.) accepted as S by over 5 m. ; 
and on the negative side we have —4°4m,, —4°4m., —51m., —80m., 
+118m., and —14:2m. Here the separation is not so marked; but 
there is a full 2m. interval. Some or all of these negative errors may 
be readings of PR, but the two largest, which both occur on January 11 
(Toronto —11°8 m. and Stonyhurst —14°2 m.), are supported by 
several other readings and probably refer to a preliminary shock. As 
the performance of the Milne pendulums is the main point under 
