OF ARTS AND SCIENCES. 157 



taken of the tangent and the sine galvanometers ; this last being very 

 sensitive, and the instrument used for measuring the small shunted 

 current. This process was necessary every time a new wire was 

 put in the frame, in order to obtain a constant that entered into the 

 formula for determining the temperature. It was found necessary to 

 change the radiating wire a great many times during the experiment, 

 because of fusion by too strong a current. 



The above process being completed, the vulcanite frame was set in 

 position in the box, and a strong current sent through the wire so as 

 to make a brilliant luminous spectrum. The pile was then raised so 

 as to be replaced by the slit, the small spectroscope placed behind 

 this, and an exploration of the luminous spectrum made. 



The pile was then lowered into position, the box carefully closed 

 and covered with cloths, and left for several hours with the current 

 cut off, so that everything in the interior might come to the same 

 temperature. An even temperature being secured, sufficient current 

 was sent through the wire to bring it to a low red heat, and the ther- 

 mopile was moved step by step through the entire spectrum, and then 

 back again ; readings of the galvanometer being taken on both sides 

 of the at each step. The order of exploring the spectrum was some- 

 times reversed, sometimes the observations were begun at the middle 

 and alternated, first to the right and then to the left, until the extremes 

 were reached, and sometimes the order was from the extremes towards 

 the middle. The method was thus varied in order to eliminate any 

 possible error. The exciting current was kept constant during these 

 measurements by means of the mercury resistance, and readings taken 

 of both the tangent and sine galvanometers on both sides of the 0. 



This series being completed, the temperature was varied and the 

 experiment repeated ; and so on for any number of temperatures, each 

 series occupying from two to four hours' time. The results of the 

 ten series are given in Table I., and are graphically represented by the 

 curves of Plate II. These curves are each the mean of several 

 observations, and are sometimes filled out in part from different series, 

 all being reduced, of course, to the same ordinates and abscissas. 

 Each curve has been constructed by plotting all the observations upon 

 which it is founded, and drawing a mean curve through them. The 

 number of series of measurements that entered into these curves was 

 fifty-two. 



The table shows the co-ordinates of the points of these mean curves. 

 "With each is given the temperature, and probable error of determina- 

 tion of a point near the B line. The points near the maximum were, 

 of course, determined with far greater accuracy. 



