Copper and Iron for Heat at different Temperatures. 163 



much as these baths admitted of changes of position about the 

 bar and alterations of distance from one another, it was possible 

 thereby to heat the bar to different mean temperatures, and so 

 to arrive at its conducting-power at different temperatures. In 

 the part of the apparatus for generating vapour, this alteration 

 was made : the vapour, after passing through the cooling appa- 

 ratus, was ^brought back to the vessel for boiling ; and so the 

 generation of steam could continue for any length of time with- 

 out a fresh supply of water. 



In the reduction of the observations, I have followed the same 

 method as in my former paper. In order to eliminate from the 

 periodical variations the error which is introduced by the varia- 

 tion in the mean temperature of the bar, and which may be called 

 the secular variation, I have combined in the calculation 



1 + 1' 2 -f 2' 



^—^ with 12, -^- with 13 &c, 



rwith^, 2' with ^^ 



where the observations taken in periods consisting of twenty-four 

 divisions are denoted by 



(1. Period) . 1, 2, 3, . . . 12, 13, 14, . . . 24. 



(2. Period) : 1', 2', 3', . . . 12', 13', 14', . . . 24'. 



The necessity of this was particularly manifest in the case of the 

 iron bar, which, after the lapse of three hours occupied in the 

 experiments, had not acquired a constant value for its mean 

 temperature ; moreover, in experiments which last several hours, 

 the mean temperature must always vary somewhat on account of 

 the alteration in the temperature of the room, a circumstance 

 which cannot be avoided. 



The thirty-two observations making up the periods for the 

 iron bar have not been taken together, but have been divided 

 into two series of sixteen ea*ch. Out of these two series, values 

 have been calculated, and the mean of these calculated values 

 has been ultimately taken. 



With three exceptions, the calculated values in the folio wing- 

 Table are the results of a double series of observations in which 

 the thermometers had their positions changed, whereby the error 

 in the value of the scale due to the unequal sensitiveness of the 

 thermometer was eliminated. Inasmuch as each double series 

 of this kind consisted of at least four full periods, it follows that 

 the values deduced in the case of the periods divided into twenty- 

 four divisions must be the results of at least 192 readings, and 

 the values deduced in the case of the iron bar must be the results 

 of 256 readings. 



M2 



