186 
PROCEEDINGS OF SECTION A. 
In Australia, however, the results of 1888, found by exchange of 
galvanic signals, may be said to confirm those obtained in Melbourne, 
Sydney, and Windsor by lunar observations. For accepting the last 
values given by Dr. Auwers — (3) page 315 — which are based on the 
latter methods, the amount of discordance is only a fraction of a 
second of time. 
The telegraphic method has in some instances produced astonishing 
results. The interval Greenwich-New York was measured three 
times in different years through different cables, and the three values 
obtained did not differ by more than one-hundredth part of a second 
— (4) page 181, But this is not to be taken as a typical example, 
for in the majority of cases the uncertainty of the results is much 
greater, often amounting to as much as half a second. 
The last telegraphic measurement, Greenwich- Paris, in which 
the precautions to guard against error were equal to its importance, 
gave independent results differing by 0’4 seconds — (5). 
Pive independent measures of the interval Sydney- M elb our ne 
vary in their extreme values by 0*6 second — (6) page 24. These, 
however, may be considered the limits of error in telegraphic 
operations of longitude, and the meihod may be assumed to give 
results generally true within much smaller amounts than those last 
mentioned. So that if it were feasible to exchange galvanic signals 
directly between Greenwich and Melbourne, provided that the signals 
occupied the same time in passing from one terminus to the other in 
both directions (a condition not yet ascertained in any case), wo 
might be able to determine the position of an Australian meridian 
within a very small fraction of a second of time. But each interval 
measured is more or less affected by some error, and there are at least 
ten such intervals to consider between the prime meridian and the 
meridian of Melbourne. Therefore we must not too readily assume 
that we know our longitudes within 100 or 200 feet, even though 
such may be the amount indicated by the theoretical probable error. 
These considerations seem to point out very strongly that, in 
attempting to find out what values we are to adopt as the most 
probable from the many combinations that can be formed, it is 
advisable to take into account those derived from moon culminations 
and oecultations. The American astronomers and others justly 
adopted the telegraphic values iu preference to all others for their 
longitudes; but our conditions are different, and we have the fortunate 
circumstance that the older results based on absolute methods differ 
but a little from those of 1883, aud that therefore the former cannot 
vitiate the latter to any appreciable extent. 
The telegraphic operations to be considered for my purpose 
extend over a period of thirty-three years, and their records are 
necessarily scattered in several volumes, reports, aud papers, some of 
which, as stated before, came to light only recently. They contain 
full accounts of the circumstances under which the work was carried 
out, show complete details of the observations and deductions in 
nearly every case, and for the intervals included between Greenwich 
and Aden the theoretical mean errors computed by Dr. Auwers and 
Dr. Gill are attached to the values therein adopted — (7). 
It was my first care to collect all the independent values of 
intervals actually measured, and to compute their mean errors (if not 
