(ar. ae 
ON SEISMOLOGICAL INVESTIGATIONS. 9 
the observed quantity, and A is only inferred. But the consequent 
change in (S—P)+-(S—SR) is small and in the wrong direction ; for though 
S$ is increased by 14 seconds, SR, is increased by 18 seconds. After 
altering A we have, however, still to apply the errors of 8 and of SR,. 
The complete process may be represented thus, when the corrections 
used are those of the 22nd Report :— 
i= A LA +8 -8R, (S—P)+(S—SB,) 
s. a a 8. s. 8. 
Tables . 538 68:0 34-0 +1202 —1520 +220 
A _ +12 +06 + 14 — 18 — 4 
Errors . — 3 + 38 + 35 
We see at a glance that the efficient corrective is the error in SR, which 
is taken to be twice that of S at 4/2. The error of S at 34°-6 is found 
in the 22nd Report to be —19 seconds; if this were only larger numeric- 
ally, say —34 seconds, we could completely destroy the ++ 64s. of Table 
V., column O—C,. 
The observations of SR, are not, however, so numerous as those of 
PR, to which we now return. Let us neglect the small change in A. 
Beginning at the bottom of the table, let us assume that we know that 
the error at A=93° is —7 seconds, from the 22nd Report. Then, since 
P—PR, =+1s. 
PR, =P—1s.= —7s.—ls.= —8s, 
é.e. the error at A = 46°-5 = — 4s., which so far accords quite well with 
that deduced independently in the same Table II. of the 22nd Report. 
But as we proceed upwards we shall not find the same good accordance. 
The errors for the smaller values of A deduced in this way from PR, 
are as in Table VI. :— 
TaBLE VI. 
Errors of P deduced from PR,. 
A From PR, |From (22) Old Tables 5P, New Possible 5P.,| Corn. 
tes = 2 
® es: Ss. 8. S. 8 
11°5 +10 +1 | 172 ; 182 184 +12 
/ 34 | 35 36 
14:0 +11 +1 | 206 217 220 +14 
; 33 29 31 
16°5 +7 +2 | 239 | 246 251 +12 
30 | 27 25 
19-0 + 4 +2 | 269 273 276 +7 
30 26 18 
21:5 0 +1 | 299 299 294 — 5 
29 26 21 
24:0 — 3 0 | 328 325 315 —13 
25 24 25 
26°5 —4 —1 353 349 340 —13 
; 25 23 26 
29:0 — 6 —3 378 372 366 —12 
24 22 25 | 
31:5 — 8 —4 402 394 391 —ll 
23 21 24 
34:0 —10 —6 425 415 415 —10 
