30 



MEMOIES OF THE NATIONAL ACADEMY OF SCIENCES. 



the light intensity at a given division of the lamp-scale by the same factor that it is diminished 

 when the wheel is interposed between the lamp and the slit. 



Two more such complete sets of observations were made under favorable circumstances, one 

 on the sun on November 1, and one on the moon on October 31. Other observations made under 

 disadvantageous circumstances, such as a hazy or smoky sky, were rejected. 



The diflereuces, which are sometimes considerable, between those results of these observations 

 which should be identical, are duej to errors of observation, as well as to different conditions of 

 the atmosphere at the times of observations, differences in altitude of the heavenly bodies observed 

 and variations in the intensity and quality of the light from the photometer lamp. Since the 

 effects of these sources of error, with, perhaps, the exception of that due to difference in altitude, do 

 not allow of computation, the best we can do is to regard them as made under perfectly similar 

 circumstances and combine them accordingly. We shall then at least know, from a consideration 

 of the general effect of the actual differences in circumstances, in which direction the error of the 

 combination lies. 



The following table exhibits the mean values resulting from snch a combination. In the last 

 three columns are given the intensities of the three kinds of light in terms of lamp light, cdl being 

 supposed equal in the yellow. The fifth and sixth columns are obtained by multiplying the second 

 and third columns throughout by a proper factor : 



Sunlight. 



.26 



.46 



.79 



1.00 



3.33 



5. 22 



9.03 



16.92 



Plate 4 is a graphical reprsentation of this table. The intensity of lamp-light is repre- 

 sented by a straight line everywhere at the distance 1 from the axis of A. The sunlight and 

 moonlight curves intersect this line at the point A=0''. 586. They rise rapidly towards the violet 

 end, but the sunlight ordinates increase faster than the moonlight ones. These curves show that 

 the proportion of violet in sunlight is much greater than in moonlight, although as a quantitative 

 determination the observations are not entirely satisfactory. The principal cause of error is, as 

 already mentioned, the deficiency of violet rays in the light from the comparison lamp. The errors 

 of observation become more apparent on eliminating this intermediate term, and conqjariug directly 

 the light from the sun with that from the moon. From the curve in the figure we obtain the first 

 part of the following table, and by a graphical construction of this part we get the last two col- 

 umns from a smooth curve. This curve, as given by the table, is concave towards the axis 

 of k It is quite certain, however, that if the observations had been perfect and made under 



