FROM ORDINARY DAYS OF THE ELEVEN YEARS 1890 TO 1900. 
239 
§ 17. The possibility that the phenomena owe something to instrumental causes 
must be recognised. If there is any instrumental cause at work, it is presumably of 
thermal origin. In the case of I, any sensible temperature effect on the absolute 
observations is difficult to imagine. The strength of the dip needles, it is true, may 
not be the same, when they are stroked at 25° C. as when they are stroked at 10° C. 
I have no statistics on the subject. Also the distance of the centre of gravity from 
the axis of suspension will naturally increase with rise of temperature. Thus there 
would be nothing very surprising if the inclination observed with one particular end 
of the needle dipping varied slightly with temperature. But the reversal of the poles, 
invariably observed at Kew, ought to eliminate this as a first approximation, leaving 
only second order terms, which one would not expect to be sensible. The absolute 
observation hut at Kew is heated by a lamp in cold weather, so that the difference 
between summer and winter temperatures of observation is only of the order of 10° C. 
For the diurnal inequality in I we are dependent on the H and V magnetographs, 
the latter of which has a large temperature coefficient. An error in the value 
accepted for that coefficient, or in the results accepted for the diurnal variation of 
temperature of the V magnet, might introduce a differential error as between summer 
and winter. Any such error would affect the accuracy of the corrections applied to the 
observed inclinations to reduce them to the mean value for the day. The hour at 
which the observations is taken is, however, one at which the change is not rapid 
either in the inclination or in the temperature of the magnetograph room. Thus any 
considerable error seems improbable. 
§ 18. In the case of H, the ways in which temperature might come in are more 
numerous. There might be an error in the temperature coefficient of the collimator 
magnet, but the consequences of this would not be nearly so serious as might appear 
at first sight. What really concerns us in this connection is not the mean temperature 
of the whole H observation, but only the difference between the temperatures in the 
vibration and deflection experiments, which is usually only 1° or 2° C. In the present 
case, moreover, the error would come in not on the average size of this difference, but 
only on its seasonal variation. Temperature practically always rises during an H 
observation, and so is higher during the deflection experiment, which comes last, than 
during the vibration experiment. If there were no artificial heating the difference 
would naturally be greatest in summer, so a seasonal differential error is conceivable. 
If it existed, however, it should show itself in an apparent seasonal variation in 
the values found for the magnetic moment of the collimator magnet reduced to 0° C., 
and a special investigation showed no trace of this. 
Errors in the values assumed for the variation with temperature of the length of 
the deflection bar and the moment of inertia of the collimator magnet would come 
in on the full annual range of temperature in the magnetic hut, but coefficients of 
thermal expansion in brass and steel are small quantities, and any large percentage error 
in them is most improbable. The temperature coefficient of the H magnetograph is 
