when the barometers or thermometers were above or below certain marks, and so on.
In dealing with any complicated problem it is usually best to try and simplify it as much as possible, and if the problem even then appears too difficult, to attack a special case, and this is undoubtedly the best plan to pursue here. In order to avoid the rapid fluctuations which occur in the temperate zones, meteorologists have turned to places in the tropics where the types of weather during the course of a year are apt to be much more permanent and where the changes from one year to another are likely to appear more clearly than in other regions of the earth. It seems to be generally agreed that, owing to its political and geographical situation, India is the country best fitted to satisfy these various conditions. The most marked peculiarity of its climate, taking the country as a whole, is a rainy season extending over some two or three months, with comparatively little precipitation per month during the rest of the year. Independently of the scientific side of the question, the quantity of rainfall and the duration of the rainy season are of immense practical importance, since a deficiency in either may and usually does involve a famine. An examination of Indian records may therefore be of great value; and as a fairly continuous series of observations at several observatories has been obtained for some twenty or thirty years, it may not be too early to commence a systematic examination of them.
Several investigations in this direction have been made lately, unfortunately with but little success in obtaining positive results. Here it is advisable to set forth with a little more detail what is meant by 'success,' since it is necessary to have a clear conception of the value, both scientific and practical, to be attached to the results of investigations based on the theory of averages. A set of observations is recorded and an examination of them is made with a view of finding out the existence of a cycle. Suppose that one cycle is found, but that its effect on any individual observation is very small. For example, if the temperatures at any particular place vary in the course of a year from 20° to 90° and we find a cycle running through all its changes in eleven years and having a maximum effect of 1°, the daily temperatures are scarcely altered if we subtract the amount corresponding to this long-period variation. Thus, though the cycle may be thoroughly well made out, it tells very little about the variations of temperature. It is only when we are able to get a sufficient number of periodic changes whose combined effect will fairly well represent the individual observations that we can be said to be at all successful in our analysis of the latter, and it is only then that a basis is found sufficient to predict the future numbers of the series.
Of the various attempts to trace long-period changes in the
