FROM ORDINARY DAYS OF THE ELEVEN YEARS 1890 TO 1900. 
195 
origin, and is partly dependent on the type of day dealt with. In the average of the 
Astronomer Royal’s quiet days, D rose 0''044 (0'23y), II rose 3'34y and V fell 0'84y. 
Thus if all days had been quiet days, D and H would in a single year have increased 
respectively 16'‘1 (86y) and 1219y, while V would have fallen. 307y. The real 
average annual changes during the eleven years were — 5'79 ( — 30‘8y) in D, + 25'9y 
in H and — 22'6y in V. 
It seems desirable to look into the matter a little more closely. Take first the case 
of D. There were in the eleven years 209 disturbed and 19 incomplete days. None 
of the latter, so far as could be judged, were highly disturbed. If we regard them as 
ordinary, as we are fairly entitled to do, we have 3808 ordinary days. The mean 
observed n.c. changes on disturbed and ordinary days were respectively + Cf'327 and 
— Cf'0301. The total n.c. changes were thus + 68 /- 3 on disturbed and —114 /- 6 on 
ordinary days, leaving a balance of — 46 /- 3. If all measurements were exact, in the 
absence of instrumental change, we should expect this balance to agree with the 
observed secular change, but this at the observed average rate of — 5 /- 79 per annum 
amounted to — 63 /- 7. This leaves 16''4 unaccounted for, suggesting an instrumental 
drift of l /- 5 per annum. 
In the case of H the 10 incomplete days may reasonably be regarded as ordinary, 
making a total of 3775 ordinary days. The observed n.c. change on ordinary days 
averaging + 0‘797y, the total n.c. change from the ordinary days of the eleven 
years comes to + 3009y. Of the remaining 242 days, 209 were included in the 
original list of disturbed days. The mean n.c. change for these days was — 13'2y, 
giving a total n.c. change of — 2759y. The remaining 33 days were included in the 
subsidiary list of distui’bed days, which was made out when the H curves came to be 
treated. The sum of the n.c. changes on these 33 days was only — 24y. Thus the 
total n.c. change for the 4017 days of the eleven years was ( + 3009 —2759 —24)y, 
or + 226y. The observed secular change, + 25‘9 x 11 or +285y, exceeds this by only 
59y, suggesting the trifling instrumental drift of — 5'4y per annum. 
In the case of Y, including the incomplete amongst the ordinary days, we have 3669 
of these with an average n.c. change of — 0'420y, giving a total of — 1541y. Ihe 209 
disturbed days in the original list contributed +209 x 2'7y or +564y, while the 139 
disturbed days on the subsidiary list contributed +304y. Thus for the whole 4017 
days we have a balance of ( —1541 + 564 + 304)y or — 673y, as compared with a true 
secular change of — 22‘6 x lly or — 249y. This suggests an instrumental drift at the 
average rate of — 39y per annum. 
The chief importance perhaps of' these calculations is the light they throw on the 
trustworthiness of the magnetic curves and measurements. It is unnecessary to 
emphasise the fact that when instrumental creep is large it is a source of very 
considerable uncertainty. The results obtained above are not put forward as exact 
measures of the instrumental creep, but only as showing its order of magnitude and 
the general fact that it was small. Most of the quiet day curves were measured 
VOL. CCXVI.—A. 2 E 
