Plumb-line in the Hawaiian Islands. 309 



These conditions were much more favorable than those that 

 would have resulted from carrying out the first conceived plan 

 of taking a stand to the top and swinging in a double tent. 

 No more perfect support could be had than a two inch plank 

 imbedded in solid masonry, and a cavern protected on three 

 sides by twenty feet of rock gives exceptionally good tempera- 

 ture conditions. The support in its essential features was the 

 same for all three stations. 



Three thermometers placed near the top, middle, and bottom 

 of the room were used throughout the work. The one below 

 was attached to a brass rod, and the rod and thermometer were 

 enclosed in tin foil — thus insuring as far as possible the same 

 temperature for both. The brass rod and the pendulum being 

 of the same metal, the lower thermometer was supposed to 

 indicate the true temperature of the pendulum at its lower 

 end. This thermometer was read continuously throughout the 

 swing. The middle and upper ones were only read when it 

 was necessary to enter the room to start the pendulum. By 

 this means we get a very accurate idea of what the real tem- 

 perature is, throughout the whole length of the bar. At one 

 station a fourth thermometer was employed which gave the 

 temperature of the air in the immediate neighborhood of the 

 attached thermometer and rod. When these means are em- 

 ployed it is hard to believe that an error in the mean tempera- 

 ture of the whole pendulum can be made as great as one-tenth 

 of one degree. The influence of such an error is quite within 

 the general range of errors of observation, or accidental errors 

 beyond our control. 



The oscillations were observed from the transit tent twelve 

 feet distant by means of a small theodolite. A window of 

 plate glass was built in the wall of the pendulum house, and 

 through it the observations of transits, amplitude of arc, and 

 lower thermometer were made. 



Each swing with heavy end down was made to consist of 

 15000 oscillations, beginning with an amplitude of -^ of the 

 radius. The instant marking the beginning of the swing was 

 determined by forty transits across the vertical wire of the tele- 

 scope. The probable error of the mean of this number of tran- 

 sits is considerably less than one hundredth of a second. After 

 the pendulum had swung for four or five thousand oscillations, 

 a few additional transits were taken in order that there might 

 be no mistake in the whole number of oscillations made. At 

 the end of every hour and a half is quite often enough to take 

 the intermediate transits, because the uncertainty in the deter- 

 mination of one oscillation, does not, in this time, accumulate 

 to be more than half a second at most ; and inasmuch as tran- 

 sits always begin with the pendulum moving in the same direc- 



