i7, 6 Fisher: Low-sun Phenomena in Luzon, III 609 
Table 1 — Continued. 
Duration. 
Contacts. 
No. 
Date. 
Ob- 
served. 
Com- 
puted. 
Differ- 
ence. 
Mirage 
Type. 
i 
2 
Remarks. 
1920 
Seconds, 
Seconds. 
Seconds. 
39 
Apr. 4 
134.5 
134.8 
— 0.3 
? 
A 
a 
Colton phenomena 
marked. 
40 
Apr. 5 
134.5 
134.5 
0.0 
0 
A 
a 
a 
San Fernando. Colton 
phenomena weak. 
41 
Apr. 6 
134.9 
134.6 
+ 0.3 
0 
A 
b 
b 
Colton phenomena? 
42 
Apr. 24 
158.8 
136.9 
+ 21.9 
? 
? 
bl? 
a 
Mount Mirador. Colton 
phenomena very 
strong. End a thin 
blue line. 
No. 1 was made with a stopwatch whose rate was unknown; 
Nos. 2 to 17 were made with the seconds hand of an ordinary- 
watch, whose error was constantly checked by the time ball of 
the Manila Observatory, and with either a field glass, a galvano- 
meter telescope, or the naked eye, with suitable protection. 
When a field glass (a Lemaire 4-power, Galilean) was used 
with an ordinary watch (Nos. 10 to 17) the left-hand eyepiece 
was removed and the watch held in front of the left-hand ob- 
jective, at such a distance that to the left eye the dial, to the 
right eye the horizon, were both clearly visible. It is difficult 
to attend to both images at once, but with practice an accuracy 
of better than a second can be attained. 
Nos. 18 to 36 were made with a i second stopwatch (works 
marked Leonidas W.'Fy., Switzerland), and a galvanometer tele- 
scope with smoked plane spectacle glass covering the objective. 
This helped greatly in observing lower (first) contact, during the 
cool months when the sun is very brilliant even near the horizon, 
but tended slightly to obscure upper (second) contact and so to 
shorten the observed duration of sunset. 
The correction of the stopwatch was found by averaging five 
runs of 1 hour each, 30 minutes each, 10 minutes each, and 1 
minute each (twenty runs in all), and was found to be exactly 
proportional to the length of the run within limits of starting 
and stopping errors, and equal to +0.2 sec./min. In all tests 
and observations initial error (flyback) was allowed for. 
Feeling that Nos. 1 to 36 had established the existence of dif- 
ferences between computed and carefully observed durations of 
sunset, and as I believed that the explanation must lie in peculiar- 
ities of atmospheric refraction, the month of April, 1920, was 
