54 JAfe and Letters of Fronds Gnlton 



ocean l)eiiig divided into areas of 5° angle in longitude and latitude, and 

 the ship returning all its observations, sulfject to the sole condition of an 

 interval of eight hours between observation and observation, a ship will give 

 more ol>servations when the wind is unfavourable than when it is favourable ; 

 accordingly there will be an error produced — since favourable and un- 

 favourable winds are peculiar to certain areas, and ships outwiird and inward 

 l>oinid follow different courses — in talking not only the mean direction of the 

 wiiul for certain ai-eas but also in other meteorological variates highly corre- 

 lated with the wind, such as temperature and dampness. The remedy would 

 be to enforce not only an interval in time, but an mterval in distance of the 

 positions of successive ol)servations. 



Galton's criticism is of less importance now that steamships have replaced 

 sailing vessels, but the paper is of interest as marking probably the first 

 occasion on which Galton exhibited publicly his tine instinct for the discovery 

 of statistical fallacies. 



The reader will not appreciate Galton's work at this period unless he 

 remembers that Galton's earliest travels were associated with sailing ships; 

 it wtis in such a vessel, the Dalhousie, that he sailed for Africa; and he 

 thought for many years of his life in terms of wind and not steam as a 

 motive power'. Thus it came about that when Galton turned his study of 

 meteorology in the direction of ocean travel, he thought in terms of sailing 

 vessels. The wind had for Galton a singular fascination, and for him the 

 problem always was : What can we learn from the wind, how can we makd 

 it of greatest service? 



Three or four of his papers touch on wind problems, and these we will 

 now briefly consider. 



The first one that may be refeiTed to is entitled: "Barometric Predictions 

 of Weather." and the paper was read at the British Association Meeting in 

 1870'. fJalton's paper is suggestive, because, what he is actually seeking 

 for in his linear prediction formula of the velocity of the wind in terms of 

 barometric height, temperature and damp is what is now familiar to statis- 

 ticians as a multiple regression formula. Galton very properly saw that the 

 relation of barometric height to wind-velocity dici not depend upon the 

 instantaneous wind, and he accordingly experimented with average wind- 

 velocity for a series of two, three, etc. hours. He came to the conclusion 

 that the best period for the average was about twelve hours. He con- 

 sideretl that twelve hour averages should also be taken for temperature and 

 damp. Galton ettsily found his averages from the automatic record of con- 

 tinuous temperature, wind-velocity and damp. He explains clearly why he 

 takes an average, namely the barometric pressure acts in sympathy with 

 a much larger wind-velocity area, than that immediately in its own neigh- 

 bourhood. The pressure (as in the case of water) is affected some time 



' I tliink this is true even a,s late as the early 'seventies when Gallon wan busy with his 

 "wave eiiffine'' (see p. 51). Such an engine as a propulsor would hardly have occurrecf to one 

 who had grown up in an era of steam vessels. 



» Brit. Amoc. Report, 1870, Tran«. Sectiow,, pp. 31-33; Nature, Vol. ii, Oct. 20, 1870, 

 pp. 501-3. 



