OF ARTS AND SCIENCES. 237 



incorrect, it is clear that the value so obtained must be somewhat 

 increased. 



Several of the artificial lights were next tried. The first one to 

 suggest itself was naturally the standard candle ; but a very few ex- 

 periments sufficed to show that it would not do; and I have since 

 found that of all the lights examined, including the Sun, Lime, Mag- 

 nesium, and Electric, none was so uncertain in color as the standard 

 candle. After experimenting with several other lights, the standard 

 finally adopted was the gas flame from an Argand burner, using 

 about 5 cu. ft. per hour. A diaphragm .5G8 cm. in diameter, and 

 having an area of .253 cm. 2 w;is placed over the most brilliant por- 

 tion of the flame. A standard was thus obtained which would be 

 almost absolutely constant, in both light and color, during any one 

 set of experiments (usually occupying about an hour), and which I 

 judge from subsequent experiments would vary very little even in the 

 course of a month. The candle-power of the whole flame when 

 burning 5 ft. 3 per hour is about 1G.0, that of my standard, .67. 



Having obtained a satisfactory light, the next step was to get an 

 instrument by means of which the various lights to be measured 

 could conveniently be compared. For this purpose I use an ordinary 

 double-slit spectroscope, furnished with a grating, having the lines 

 (6480 to the inch) photographed on glass. In front of the slits are 

 placed two right-angled prisms, arranged to reflect the light from 

 ' opposite directions into the collimator. On looking through the 

 instrument, the two spectra will be seen one above the other, and by 

 means of two sliding metal plates, placed at the focus of the telescope, 

 the spectra may be cut down so that only a narrow vertical strip of 

 each shall be visible. 



The standard light is fastened upon a little car, rolling upon a track 

 over a fixed scale, by means of which its distance from the slit is 

 measured. The light to be compared is placed at a known distance 

 on the other side of the slit ; the telescope is pointed to some particu- 

 lar color and the standard moved backwards or forwards till the two 

 spectra are of the same brilliancy. The distance is then read off on 

 the scale. In measuring the red and violet ends, it was usually found 

 necessary to place the light to be measured nearer to the slit than for 

 the other colors. The " standard " slit was kept at a constant 

 breadth of .056 mm. through all the experiments, and the light could 

 be moved from it through a distance of from 10 to 60 cms. I found, 

 however, that it was generally better not to place the standard nearer 

 than 15 cms. It will be noticed that the slits are generally kept 



