88 ROYAL SOCIETY OF CANADA 
with the usual three prism dispersion is in the neighborhood of 0.5 km. 
per second. When greater than this it indicates a spectrum with poorer 
lines and when less that special precautions and limitations were adopted 
in the making and measurement of the spectra. It is also shown that 
the probable error of determination with a thoroughly stable one prism 
instrument of one third the dispersion is for solar type spectra measured 
on the spectra comparator about + 0.70 km. per second and for spectra 
of earlier type varies from about + 2 to + 11 km. per second. 
IV.—Generally speaking the major part of the errors in solar type 
stars are due to systematic displacements of the lines as a whole owing 
to flexure, temperature changes, imperfect adjustments of the optical 
parts, faulty guiding or other causes. The accidental errors of pointing 
are responsible generally for only one third or less of the total error. 
In the case of early type stars the systematic displacements due to 
instrumental conditions will probably be approximately the same while 
the errors of pointing are correspondingly increased. 
