60 THE ROYAL SOCIETY OF CANADA 
PROBABLE ERRORS. 
26. As Adams * has already compared his errors of measure- 
ment with those of Dunér and Halm, showing the marked advantage 
of the photographic method, it will suffice here to give the Ottawa values 
and compare them with Adams. 
The mean probable error of measurement of the velocity from 
a single line determined by the use of all the lines on all the plates is 
Series I — = 0.024 km. per sec. 
Series III — + 0.015 
) ) 22 
The probable errors in Series I vary for the different plates from 0-010 
to 0.040 and in Series IIT from 0.006 to 0.023. As the number of lines 
measured in each plate in the two series have been 19 and 15 respectively 
the probable error of an average plate as determined from the internal 
agreement of the measure is 
Series I — = 0.0055 km. per sec. 
Series IIT — + 0.0038 
” ” ” 
” 
The average probable error of a plate determined from comparisons 
of the velocities of all plates at the same latitudes and for all the latitudes 
is for 
Series I — = 0.028 km. per sec. 
Series ITI — + 0.026 Ue Sa 
or 5 and 7 times the probable error as determined from the internal 
agreement of the lines. 
These somewhat anomalous results are however not unusual as 
about the same ratio of probable errors is obtained in stellar radial 
velocity work and in many other astrophotographic methods, but the 
cause of this comparatively high ratio cannot be satisfactorily explained. 
One can imagine that changing instrumental conditions might 
cause differences in displacement in plates taken on different dates 
but where, as in the example previously cited, differences of from 0.05 
to 0.07 km. were found on exposures taken one immediately after the 
other on the same plate on the same region of the sun and under, so 
far as known, identical conditions, no explanation, except that of 
rapidly changing proper motions on the sun, can be assigned. 
27. In comparing these probable errors with those of Mt. Wilson, 
only series III which is in the same region, À 4250, as the Mt. Wilson 
plates must be considered for, as the relative probable errors indicate, 

* Adams & Lasby, p. 117. 
