292 Prof. Balfour Stewart and Mr. \V. L. Carpenter. 



observations. A description of this method has already been 

 published in the " Proceedings of the Royal Society," May 15, 1879, 

 bat in order to render what follows more clear, we shall to some 

 extent reproduce this description. 



Imagine, by way of illustration, that we have in our possession a 

 long series of temperatures of the earth's atmosphere at some place in 

 middle latitudes, and that, independently of our astronomical 

 knowledge, we were to make use of these for the purpose of investi- 

 gating the natural Inequalities of terrestrial temperature. We should 

 begin by grouping the observations according to various periods 

 taken, say, at small but definite time-intervals from each other. 

 Now, if our series of observations were sufficiently extensive, and if 

 some one of our various groupings together of this series should 

 correspond to a real Inequality, we should expect it to exhibit 

 a well-defined and prominent fluctuation, whose departure above 

 and below the mean should be of considerable amount. Suppose, for 

 instance, that we have 24 places in our series, that is to say, we 

 proceed to group our mass of temperature observations in rows of 

 24 each, and suppose further that the time-interval between two con- 

 tiguous members of one row is exactly one hour. The series 

 would thus represent the mean solar day, and we should, with- 

 out doubt, obtain from a final summation and average of our rows, a 

 result exhibiting a prominent temperature fluctuation of a well- 

 defined character, which we might measure (as long as we kept to 

 24 points) by simply adding together all the departures of its various 

 points from the mean, whether these points be above or below ; in 

 fine, by obtaining the area of the curve which is the graphical 

 representation of the Inequality above and below the line of abscissae 

 taken to represent the mean of all the points. 



4. Suppose next that, still keeping to rows of 24 places, we should 

 make the time-interval between two contiguous members of a row 

 somewhat different from one hour, whether greater or less, we should 

 uow in either case obtain a result exhibiting, when measured as 

 above, a much smaller Inequality than that given when the interval 

 was exactly one hour ; and it is even possible that if our series of 

 observations were sufficiently extensive we should obtain hardly any 

 traces of an Inequality whatever. In fine, when each row accurately 

 represented a solar day, the result would be an Inequality of large 

 amount, but when each row represented a period either slightly less 

 or greater than a day, the result would be an Inequality of small 

 amount. 



This process as far as we have described it is not new, having been 

 already used by Baxendell and probably by other observers of stellar 

 variability. In the present instance we should by its means, after 

 bestowing enormous labour in variously grouping in accordance with 



