304 SECTIONAL TRANSACTIONS.— A. 



closely to straight lines for epicentral distances up to 12" or more. In addition to 

 P, P*, Pg, S, S*, Sg, discussed in detail by Jeffreys for the Hereford and Jersey 

 earthquakes of 1926, Conrad has found several other waves for which no immediate 

 explanation is forthcoming, and has given a very full account of one earthquake for 

 which P and S are sub-divided into Px, P„, and Si, S„ respectively ; Dr. Jeffreys 

 not3s on some of the records discussed a wave which is possibly a compressional wave 

 Ps in the sedimentary layer, and Mr. E. Tillotson finds both Pa and the corresponding 

 Sj in the records of a Balkan earthquake. All this work, however, has involved 

 careful scrutiny and measurement of the actual records. It is interesting to examine 

 the wealth of observations given by the International Seismological Summary. The 

 epicentres of the I.S.S. are sufficiently accurate for a preliminary examination, and 

 no adjustment of epicentre will explain awaj^ some of the very large residuals. 

 Earthquakes in the Balkans are particularly favourable for investigation in view of 

 the large number of European observatories. 



If the usual method is followed and the time of transit is plotted against epicentral 

 distance for all the observations printed (including the ' Additional readings,' which 

 are specially valuable) there is no difficulty in picking out the P, Pg, S, Sg graph.s. 

 and P* and sometimes S* may be drawn in, with close agreement between the 

 measured velocities and those found by Jeffreys and Conrad. Most of the large 

 residuals are at once seen to arise from wrong identifications : for instance, in 1926, Sep- 

 tember 3rd. 21 h. 59 m. 50 s., the residual of -|-28 s. in P for Athens arises because 

 Pg has been read for P, and the residuals of -|-55 s. for Lemberg, and 70 s. for Rocca 

 di Papa are attributable to the same cause. Likewise, the -f 70 s. for S at Zagreb 

 arises from the use of Sg ; the correct S wave is quoted as ' ePRi ? ' in the additional 

 readings. Vienna mistakes S* for S, and most of these incorrect identifications 

 could be inferred from tables for times for Pg-P, Sg-S, &c. Naturally, all this could 

 best be done from the original records, but it is possible to obtain fairly rapidly quite 

 a lot of geophysical information, not from a laborious examination of seismograms, 

 but from the readings and niisreadings available in the I.S.S. ' It is because of these 

 incorrect identifications that the published observations are so informative ! 



With the rather wide scatter of the plotted observations one might be tempted 

 to infer large errors of readings, to the extent of 20 s. or 30 s. This would be unwise : 

 with fairly large paper speeds observations can be trusted to a few seconds, and 

 timing errors are probably negligible in most cases, apart from mistakes of one whole 

 minute. With the large number of phases that are recorded up to 15° it is better to 

 assume provisionally that all observations are significant until examination of records 

 disproves it. The preliminary examination outlined here may indicate (a) the proper 

 identification of a given reading, (b) an improved time of starting and epicentre, by 

 the use of tables of Pg-P, Sg-S, and others, (c)an occasional large clock error, (d) an 

 epicentre of an earthquake for which readings are very few, (e) special earthquakes 

 that merit detailed study (in fact, one feels that every earthquake deserves special 

 study). 



It was certainly surprising that the readings in the I.S.S. gave for certain earth- 

 quakes values of the velocity of P, P*, Pg, S, Sg, and perhaps S*, that were close to 

 Jeffreys's values. In identifying a given wave it would be sufficient to assume these 

 values as a first approximation. All these six were found in the first shock examined, 

 viz. 1926, August 18 d. 17 h. 4 m. 52 s. (epicentre in or near Cephalonia, 38°N. , 20°. 5E.), 

 and most of the large residuals had evident explanations. Pg and Ss were suspected, 

 but not confirmed, but there seems to be a wave of velocity 6.9 km. /sec. and another 

 of about 4 km. /sec. In the Ziirich (E) record of the Jersey earthquake, reproduced 

 by Dr. Jeffreys in ' The Earth ' (second edition), these latter waves are definitely 

 present, with velocities 7.1 and 4.1 km. /sec. approximately-. If these waves are con- 

 firmed they would be particularly interesting, as the velocities fit the data for 

 crystalline basalt, which has not so far been shown to be present as one of the layers 

 of the earth's crust. 



The wave with a velocity of about 7 km. /sec. appears to be present in the shocks of 

 1926, August 17 d.l h., September 3d. 21 h., December 17 d. 6 h. and 17 d. 11 h., and it 

 is proposed to examine the actual records of one of these earthquakes ; further investi- 

 gation is certainly necessary before this suggestion can be advanced seriously. The 

 reason for associating these P and S waves together is that the ratio of the velocities 

 P/S is about 1.7, which is what is found for the pairs Pg/Sg, Pi=/S*, P/S and Ps/S., and 

 corresponds fairly closely to the case of an elastic solid for which X=[i, in the usual 

 notation, or a, Poisson's ratio, =|, when the ratio of the wave velocities is \^3. 



