460 ME. K. D. OLDHAM ON THE CONSTITUTION [Aug. IO,o6 y 



Secondly the disturbance, instead of beginning abruptly, as is some- 

 times the case, may come in gradually : and when this is the case, 

 it is easy for the times of commencement of each phase to vary by 

 a minute or more, on the records of different instruments, or even in 

 the reading of the same record by different individuals. Either of 

 these causes will make the recorded time late, but it also happens, 

 not commonly, though often enough for the contingency to be borne 

 in mind, that one station or a group of stations is affected by some 

 small local disturbance, which almost coincides with a distant 

 earthquake, and leads to the apparent commencement being too 

 early. Apart from these sources of error, there is also that which 

 may easily occur in determining the time of origin of the earth- 

 quake : this will introduce an error into all the intervals, which is 

 constant for each earthquake but varies for different ones both in 

 amount and direction ; it will, consequently, be eliminated when 

 an average from a sufficiently-large number of earthquakes is taken, 

 and will be partly eliminated even with the few which are at 

 present available. The other sources of error are partly eliminated 

 by averaging, but it is necessary to reject any records which are 

 abnormally early or late, and to take only those which, by their 

 close concordance with each other, show that they refer to the same 

 phase of wave-motion. The average so obtained will naturally 

 incline to be a little late, but is likely to be nearer the truth than 

 any individual record, taken at random, and for this reason I shall 

 deal, as far as possible, with averages rather than with single 

 observations. 



These averages may be obtained in two ways. In Table I 

 (p. 472) are given all the group-averages that I have obtained : 

 that is to say, averages of the records of groups of stations and instru- 

 ments, each average being that of observations of a single earthquake, 

 each group consisting of at least five distinct records from distances 

 differing by less than five degrees of arc from each other. 



In Table II (p. 473) a different treatment is used. The whole 

 series of records from all earthquakes, excepting those numbered 

 13 and 14, were tabulated, and the average, of each group covering 

 5° of arc from the origin, taken. In this way we get a group 

 of averages for each 5° from 45° to 95°, which are on the whole 

 better than those in Table I. 



For distances of 100° or over the possibility of averages is small, 

 and I have given in Table III (p. 473) all the available records of 

 first and second phases. The figures given are those originally 

 determined by me, those enclosed in parentheses are times which, 

 from their discordance with other records, were evidently mis- 

 interpreted or are otherwise doubtful. 



It is evident that, having prepared these tables, we are no longer 

 confined to the consideration of those earthquakes of which both 

 the time and the place of origin are known ; for, if the latter be 

 known with even approximate accuracy, we can determine the 

 former from distant records. This has enabled me to utilize two 

 other disturbances for filling in gaps in the series of records. 



