240 PROCEEDINGS OF THE AMERICAN ACADEMY 



tometer. As the arrangement of the scale in this instance was some- 

 what complicated, and the form of the observations is well known, I 

 will merely state the results in the two instances as 90 and 84 candle- 

 power. I have since measured the light, and obtained a maximum 

 brilliancy of 231 c. p. And from this, by varying the supply of gas, 

 and the distance of the lime from the burner, the light could be 

 diminished gradually to any extent. 



The intrinsic brilliancy is then obtained by placing a diaphragm of 

 known size over the light and remeasuring. As no good standard of 

 intrinsic brilliancy exists, I adopt for the present purpose the light 

 given off by my " standard." This is about .67 of a candle-power 

 at the same distance. When at a maximum, the intrinsic brilliancy 

 of the lime was 121 st. when the total light was 90 c. p. The in- 

 trinsic was 54 st. 



Probable Error. 



Using a perfectly invariable light, the mean probable error of six 



observations for the different colors was found to be : — red, 6.7 per 



cent; yellow, 3.4; green, 2.4; violet, 6.1. These figures may 



seem rather large, but when we consider that in most of the lights 



the chief discrepancies are caused not by instrumental errors, but by 



differences of color, and brilliancy in the lights themselves, we see 



that it would not be much advantage to have the instrument more 



accurate than it is ; and that if we are to measure the lights at all, we 



must allow some pretty large variations. Moreover the different 



lights vary from each other frequently by more than 100 per cent, 



which leaves room for quite large differences. In fact, we find this to 



be a subject where the magnitudes are of great range ; and accuracy 



such as we are in the habit of obtaining in other branches is out of the 



question. The mean probable error of six observations with the 



Bunsen photometer, on a constant source, varies from .5 per cent 



under the most favorable circumstances, up to 2 or 3 per cent when 



less favorable. 



Description of Plate I. 



On this plate each broken line represents some particular light. 

 The abscissae denote wave lengths expressed in .00001 of a mm. 

 The ordinates represent the brilliancies of each color, the unit being 

 the brightness of the " standard " for that particular wave length. 

 The standard light is therefore represented by the horizontal line 

 St. As observations were taken only at four particular points, we 

 have no means of knowing the shape of the curves outside of these ; 



