the Boundary of a Wave of Conducted Heat. 259 



plate and allowing the water spontaneously to evaporate. Thus 

 we obtain a smooth evenly distributed coating, giving a sharp 

 outline to the elliptical figure of the conducted heat. The plate 

 was screened from radiations of the flame which heated the wire, 

 but was not shielded from currents of air, nor was unequal ra- 

 diation of the iodide specially prevented. The method of mea- 

 surement was as follows : after the ellipse was well formed and 

 of permanent dimensions, the extremities of its longer and of its 

 shorter axes were marked by scratching through the iodide with. 

 a very slender steel point ; the plate was then removed and the 

 lengths of the axes determined by means of dividers and a scale, 

 divided into half millimetres. 



kp. 



Major axis. Minor axis. 



Ratios.- 



1. , 



. . 12-5 9-25 



1-35 



2. . 



. . 14-0 10-5 



1-33 



3. , 



. . 17-75 13-5 



1-31 



4. , 



. . 18-25 14-0 



1-30 



5. 



. . 12-75 9-5 



1-34 



6. , 



. . 12-8 9-5 



1-34 



7. , 



. . 12-8 9-5 



1-34 



8. , 



. . 16-4 11-8 



1-38 





Mean ratio . . . 



1-33 



)i 



iJ ii 



j> 



i\x, a 



3) 



3) 



in the mean ratio 



}} 



S.'s 



}} 



}) 



a » 



it 



M.^s 



}> 



An opinion on the relative values of the two modes of experi- 

 menting can only be formed from a discussion of the two series 

 of observations by the method of least squares. It is true that 

 the series are not as extended as one would wish for the applica- 

 tion of this process ; yet its results are equally fair for both. We 

 thus have found that the 



Probable error of a si n gle determination of ratios in S .^s series is*0267 



•0170 

 •0094 

 •0060 



From these figures we infer that Senarmont^s ratio is barely 

 true to a hundredth, while my result can be relied on to that 

 figure ; and if my measures had been made with a micrometer 

 microscope, on a plate protected from unequal radiation and 

 shielded from currents of air, I should have obtained a ratio 

 reliable to the third decimal place. 



To the higher ratio of my determination I attach no import- 

 ance j 1 attribute it to the peculiarity of this particular crystal ; 

 for several measures on this plate, with a waxed surface, gave 

 even a higher ratio than when the iodide was used. It hence 

 appears that, to obtain the correct ratio for a given crystal, the 

 mean ratio derived from several plates should be adopted. 



S2 



