232 



NA TURE 



[January 6, 1898 



from the camera, for the reasons indicated by Fig. 3, which is 

 an extreme case in which the camera, or eye, R is supposed to 

 be very near to the side of an animal, so fat that his cross 

 section has to be represented by a circle. 



The summit of the outline as seen from R is H, giving the 

 idea that the spine of the animal is as high as L, whereas it is 

 really at K. The ratio of lo to KO is of course the same as that 

 of LO/OR to Ko/OR, that is to ho/or ; in other words as the 

 tangent of i. R o to the sine of the same angle. The values to 

 be dealt with in reality, are very different from those in the 

 diagram. OR is 240 inches and KO may be taken as 15 inches. 

 It results that L R o is only 3° 35'. Now the tangent and sine of 

 such a small angle are so nearly alike that LO : KO :: 100 ■00 : 99'85, 

 which corresponds to a difference of less than i/8th inch in a 

 horse of 15 hands high, and is quite negligeable. 



In some fat stock, however, the backs are flat like tables. 

 Here some artifice would be necessary to obtain the true height, 

 such as by fixing a stud of say 2 inches in height to a surcingle. 

 The top of the stud would then be the point of measurement, 

 and 2 inches would be subtracted from the result. 



We now come to the effect of obliquity of the median plane 

 of the horse to the long sides of the rectangle. The 

 hoofs of a thoroughbred horse are some 4 inches wide, 

 and 4 inches apart, so that the closest distance between 

 either s or T, and the nearest side of the rectangle, is 

 6 inches ; therefore the utmost cross distance between s 

 and T is (20-12) or 8 inches, and the length of ST in a 

 horse of 15 hands in height may be taken as 60 inches. There- 

 fore the maximum obliquity within the strip is as 8 to 60, or 

 0"I333, which corresponds to an angle of 7° 39'. 



The foreshortening of the length s T ( = 60 inches) is such that 

 its foreshortened value must be multiplied by the secant of 7° 39' 



Fig. 3. 



to obtain the unforeshortened value, that is by i '009, which 

 makes it nearly i per cent, longer. This is the greatest error to 

 be feared under the conditions, and it is further much diminished 

 by determining the actual obliquity. We can do this easily by 

 measuring the distance lengthways between the points where s T 

 produced cuts the opposite sides (not ends) of the rectangle. 

 The further side of the rectangle affords the scale for reckoning 

 distances from c along that line when produced ; similarly the 

 nearer side affords the scale for distances from c measured along 

 and beyond itself. The cross distance between those points is 

 known to be 20 inches, so the obliquity is easily found. The 

 accompanying table may be found convenient. It applies with 

 strictness only to objects viewed from a great distance, but is 

 practically correct for much smaller ones. 



The mean scale for the slightly oblique median plane is the 

 perspective length of the rectangle at the point where a 

 line drawn through the middle of ST cuts the perspectively 

 viewed ends of the strip. It is unnecessary to attempt greater 



NO. 1471, VOL. 57] 



minuteness, as by determining the vanishing point (the 

 position of which is given by prolonging two or more of the 

 cross-lines upon the ground to their points of common inter- 

 section in the photograph), and then employing the further 

 methods known to draughtsmen in perspective. Much could 

 be written of which it is unnecessary to speak here, because it 

 is a condition that the obliquity shall never be great. A strict 

 attention to the elementary requirements laid down above, 

 makes the problem of measurement extremely simple ; otherwise 

 it becomes complex and troublesome. 



The next point to be considered is the method of measuring 

 between points situated on the side of the horse, such as from 

 the haunch bone to the shoulder. I shall speak of these in 

 general terms only, because the most suitable points for measure- 

 ment have yet to be determined. Whatever they be, it is a 

 great assistance, before photographing the horse, to mark the 

 points to be measured either by chalk, or more neatly by a disc 

 of gummed paper, the size of a shilling, wetted and stuck on. 

 Veterinary- Captain F. Smith has used both these plans. It is 

 also an excellent plan to prick through these points in the 

 photograph, and through a piece of paper laid below, and to 

 measure between the prick holes. The general principle of 

 dealing with these measurements is to find a mean correc- 

 tion suitable to each distance, when those distances have 

 been calculated as if they were situated on the median 

 plane. The lateral deviation from that plane of each one 

 of these points, ranges within narrow limits, when the height 

 of the horse is taken as unity. The mean deviation even 

 of either protuberant haunch bone from the median plane 

 between them, is much under 20 inches in a horse of 60 

 inches (15 hands in height). The mean range of this deviation 

 in different horses of that height, judging from what occurs 

 in anthropometric measurements, is probably very much 

 under an inch, and its extreme range in ordinary cases would 

 be under 2 inches. Extraordinary cases of massive or slender 

 build would be betrayed by the photograph itself, and could be 

 allowed for. It seems, then, that after the desirable points had 

 been determined, between which measurements might be 

 wanted, it would be a straightforward piece of work to make 

 numerous measurements between them in different horses, and 

 to draw up the suggested table of corrections for 2 or 3 different 

 positions in the rectangle. 



The head and neck can hardly be measured on the above 

 principles, as it is very difficult to ensure that their median plane 

 should be the same as that of the body. A strip of card chequered 

 with inches, alternately white and black, and fastened to the 

 head stall, affords a serviceable scale, and is by no means un- 

 sightly. 



From measurements obligingly procured for me by Dr. 

 MacFadyean, the Principal of the Royal Veterinary College at 

 HoUoway, the measurements being repeatedly made of the sam^ 

 horses by different pupils, I learnt two things. One was that 

 horses of the same class vary among themselves as much as men. 

 In short, they could be identified by a Bertillon method. The 

 other was that the fallibility of a measurer was considerable. 

 I think that measurements made on a half- plate photograph, 

 under the conditions I have described, would on the whole 

 be more trustworthy than direct measurements made with a 

 tape or callipers, especially on fidgety horses. 



Francis Galton. 



THE MAGNETIC PROPERTIES AND ELEC- 

 TRICAL RESISTANCE OF IRON AT HIGH 

 TEMPERA TURES. 



npHE magnetic properties of iron and, to a lesser extent, of 

 the associated metals, nickel and cobalt, have always been 

 a fascinating subject of study. Possessed by these three metals 

 alone, these properties, so peculiar and so different from any of 

 the other known properties of matter, have imparted to the 

 study of these so-called magnetic metals a special charm and in- 

 terest, apart from that excited by the vast industrial importance 

 of at least one of them. 



Among the very early inquiries into the nature of magnetism 

 there were not neglected experiments on the effect produced by 

 change of temperature. Three centuries ago, Gilbert recorded 

 the observation that a piece of iron or steel, if heated more 

 strongly than up to a full red heat, ceased to be attracted by a 

 magnet, though it regained its previous magnetic qualities on 

 cooling below that temperature. 



