390 DISTRIBUTION OF HEAT IN THE SPECTRUM. [MEMOIR XXVIII. 



that centre and H 2 into another focus, and by the thermo- 

 pile, or any other suitable means, measuring the heat of 

 these foci. 



Such is the method I have followed in obtaining the 

 measures now to be presented : but before giving them, 

 there are certain preparatory facts which I wish to sub- 

 mit to the consideration of the reader. 



(1.) In the mode of experiment hitherto adopted, no 

 special care has been taken to ascertain with accuracy the 

 position of the " extreme red," yet that is held to be the 

 point from which, on one side, we are to estimate the in- 

 visible, and, on the other, the visible spectrum. Differ- 

 ent persons', perhaps because of a different sensitiveness 

 of their eyes, will estimate that position differently. The 

 red light shades off gradually it is almost impossible 

 to tell when it really comes to an end. A linear thermo- 

 pile, such as is commonly used, is liable, under these cir- 

 cumstances, to give deceptive results, and any error in its 

 indications counts in a double manner. It not only di- 

 minishes the value of one spectrum, but it adds that dim- 

 inution to the value of the. other. The force of this re- 

 mark will be understood by considering the best experi- 

 ments hitherto made on this subject, those of Dr. Tyndall, 

 as related in his "Heat a Mode of Motion" (London edi- 

 tion, 1870, p. 420, etc.). In the case of the electric light, 

 the result yielded by those experiments was that the 

 heat in the invisible is eight times that of the visible 

 region. But had there been an error in estimating the 

 position of the extreme red by only two millimeters, so 

 much would have been taken from the invisible and add- 

 ed to the visible that they would have been brought to 

 equality, and then the slightest turn of the screw that 

 carried the pile towards the dark space would have 

 given a preponderance to the visible. It is obvious, 

 therefore, that there cannot be certainty in such meas- 



