116 Dr. GC. Chree on the Law 
As to the actual formule for P and Q, if we suppose 
R negligible we easily find from (3) 
pe Wyri* (737124) + Woret(rit — rs") + Wars'(72' — 71°) (8) 
Wiryt (722 — 732) + Wor's*(773? — 752) + War3h(ry2?—7e" 
es ryPrsrs'{ Wry (7? — 13°) + Wore?(1'3?— 71") + Wsrs'(ryP — 75) 
is Wines ts) + Wart(rs are) HW ors'(rs' 2 (9) 
Here W,, W., W3 may represent results from a single 
observation, but it is highly desirable they should represent 
means from a large number- ef observations. Only those 
days’ results should be used on which observations have been 
taken at all three distances, | 
_ It will be noticed that P being of dimensions 7°, and Q of 
dimensions #*, they are not strictly speaking independent 
of the temperature. Thus, theoretically, there is an objection 
both to combining observations taken at different tempera- 
tures, and to treating the functions of the distances in (8) 
and (9) as constants. In practice, however, the contributions 
from P and Q to the value of H are so small that these 
objections may be disregarded. 
The most satisfactory way of investigating the degree of 
variability of P and Q under ordinary conditions, would be 
to take observations at three distances at frequent intervals 
during several years, with the same instrument, and compare 
the results deduced by formule (8) and (9) from different 
groups of the observations. We shall presently consider 
existing evidence as to the variability of P and Q, and the 
question how these quantities are related to the dimensions 
and other peculiarities of the magnets employed. But before 
doing so we shall discuss the uncertainties in the determi- 
nation of P and Q, and in the value of H, arising either from 
errors of observation or in the assumed values of the deflexion 
distances. 
Effect of Errors in the Setting of the Deflecting Magnet or in 
the Values accepted for the Deflexion Distances. 
§ 4, The temporary induction correction is not absolutely 
unaffected by errors in the deflexion distances, but assuming 
ordinary care any error so introduced is too small to require 
consideration. Thus, for the purposes of the present inquiry, 
we may leave induction and temperature corrections out of 
account. As will be seen later, the coefficient R is seldom 
likely to be of practical moment in the case of English 
magnetometers, so to avoid undue length in the formule and 
discussion I shall in general neglect it. Gs 1S Bay 
