December 12, 1902.J 



SCIENCE. 



949 



logarithmically, an error of 1/20 in s (which 

 need not be made) will be equivalent to an 

 error of 34 per cent, in n. Hence the un- 

 certainty is about 300 particles in 1,000. The 

 Tariation of n is, however, somewhere be- 

 tween and 5,000, and hence the large relative 

 error is for the present at least without signi- 

 £eance. The error of 300 particles is smaller, 

 moreover, than the usual values of the 

 minimum. TTsing the new long ' receiver,' 



-//■6 'i-i' 



(upper figure), where the number of nuclei 

 is laid off vertically and the day of the month 

 horizontally. The weather is specified as in 

 the maps of the U. S. Weather Bureau, open 

 circles denoting clear weather, usually with 

 sun, black circles cloudy weather, and partially 

 filled circles the intermediate case in which 

 the heavens are not quite covered. R denotes 

 rain. The change of temperature in degrees 

 centigrade, during the last twelve hours, the 



«■«• . CO' -T8' 



iSOekfe if 1% i<) io ii 'S8.' ' sj' ' ' a ' ' ■'jy 



ftfar «8 



« was measurable with an error less than 1/50, 

 reducing the error of n to about 100 par- 

 ticles in 1,000. With properly constructed 

 glass apparatus the conditions will be more 

 favorable and thus fully adequate. 



Finally, it should be noted for future con- 

 sideration that for coronas of a high order 

 (». e., with colored centers), the proportion- 

 ality of n and s" is not at once assured. 



5. Treliminary Data. — During the earlier 

 observations made from October 2 to 14, the 

 water-bath was not at hand. Though uncer- 

 tain to this extent they are, nevertheless, very 

 striking in their general bearing. I have, 

 therefore, given them graphically in the chart 



change of the barometer in centimeters and 

 the prevailing winds are given above the 

 chart, for each day. Finally, h denotes hazy, 

 n refers to observation much after sundown, 

 etc. For convenience the observations are 

 connected by straight lines, but where these 

 supply the night hours they are naturally 

 much in error. 



The nucleation begins low on the second of 

 October with the rain, but thereafter in- 

 creases nearly fourfold with the bright 

 weather of the succeeding days. Note the pre- 

 vailing winds from the north. On the fifth 

 and sixth, clouds and rain usher in a second 

 minimiun of nucleation. On the fair days 



