UNCERTAINTIES IN CALCULATION 195

adoption of possibly erroneous altitudes for the stations of observation. The altitude of K? was observed from Haramokh and other stations, but the altitude of Haramokh itself may be a few feet wrong, and the altitude of K? on this account may be thirty feet in error. Another element of uncertainty in determining the height of a peak is caused by the variation in the amount of snow on its summit. There is clearly more snow on the summit of a peak in winter than in summer, and in a hot, dry summer there may be less than in a generally cloudy, snowy summer. A more compli- cated description of error is introduced by the deviation of gravity from the normal in great moun- tain ranges. The attraction of the great mass of the Himalaya mountains and of Tibet pulls all liquids towards itself as the moon attracts the ocean, The liquid in levels on the theodolites with which observations of the peaks are made is similarly affected: the plates to the theodolites in con- sequence cannot be exactly adjusted, and when apparently truly levelled are in reality tilted upwards towards the mountains. At Kurseong, near Darjiling, they would be as much as 51” out of true level and at Mussouri about 87”.

But the most serious source of uncertainty in