— 4 — 



station. This building was used for keeping boats, and had no floor. Stone 

 piers were set up in this house for penduhim and magnetometer. The transit 

 instrument was sot near the house in Dpi-n grouuil with a t('m]iorary shelter of 

 sail cloth as a protection from rain. We began our work on the night of the 

 i:Uh, and finisiied on the raoniing of the 18th of August. Leaving tlie Island 

 on the 19th, we reached Tokio on the 31st. A chock series of experiments was 

 made between the 5th and 0th of September. 



CO-ORDINATES OP THE TWO PLACES. 



Tokio. 



Lat 35° 42' 40" N. 



Long 139° 45' 45" K. 



Height above the sea level 5 metres 



OOASAWAKAJIMA. 



(BONIN.) 





Lat 



27° 4' 



11" N. 



Long 



142° 11' 



54" E. 



Height above the sea level ... 





2.2 raotros 



RESULTS OF THE PENDULUM EXPERIMENTS. 



B 



Time of a single oscillation at Tokio (before). . . 

 „ „ ,. „ (after) ... 



„ „ „ ,, (moan)... 



„ „ „ Ogasawarajima 









Time 



of 



a single osci 



illation 



at Tokio 



(before)... 



n 





>i 



it 



>» 



ft 



(aftor) ... 



n 





» 



It 



H 



tt 



(meau)... 



» 





»» 



it 



it 



Ogasawarajima 



Tima of a single oscillation at Tokio (before) . . 

 » $, .. M (after) ... 



„ >> .. •> (mean) ... 



» » » Ogasawarajima... 



1. 000708 + +.000001 2 



1.000729+ .0000030 

 1.000718 



1.000906+ .0000010 



0.999891+ .0000007 



0.999897+ .0000012 

 .999894 



1.Q00087+ .0000014 



1.000088+ .0000013 



1.000089+ .0000011 

 1.000089 



1.000271+ .0000021 



• A reference to the reduction •liccU will «how that thc«> «re not, ttrirtlT «pnikinir. ppnboMe 



error» in as much a» the clironomcUT mtf in aupixm«! to In- cimntuit tliroii^hniit tlio time fur «hiih only 

 an averapc rate i« determined. Apiin. if the fiK'urc of Iho |Kwluhini •uff.r a ilicht d.-fonnation 

 during the transportation there is no mean« i.f jud|.:inK whctlicr that liap)>i>ni-d in piing or n'tnniintr. 

 From these consideration«, we abandon the former |ilan of (rivin({ weight« to the hrfore and «flor 

 results inveraelj proportional to the s-itiar.-« of the probable errors, and we believe that the »iropi« 

 arithmetical mean gires a better approximatiou to the truth. 



