1100 



The scale of the diagram — in which tl)e smallest subdivision, 

 somewhat reduced in the reproduction, corresponds to two milli- 

 metres — is the following : 



For curve A, 89-y. period (full line: 1,5 centimetre for 1°0 C, 

 average deviation per meteor, winter from the mean — 0,68. (This 

 average is negative, since the historical data concerning cold winters 

 are much more numerous than those concerning mild winters). The 

 zero-line for the curve A has been derived, however, for better 

 comparison, from the period-years 1852 — 1916, it is — 0,56 (assumed : 

 — 0,60) and so lies a little above the true average of the 89-jear 

 period. Since the sum of all positive' and negative deviations of the 

 B-curve (Paris — de Hilt) is 127, while that of the corresponding 

 deviations of the A-curve over these 62 years amounts to 49, the 

 scale-ratio 3 : J has been adopted. 



For the curves a and b the zero-line of the b-curve (deviations 

 from monthly meajis) has of course been taken as the common normal 

 line. The scale is: J cm. for 1,5^ 0. sum of deviations in the meteor, 

 winter. Obviously the values of the curve represent the "historical" 

 deviations very incomplete!,) ; besides the unreduced "temp.-coefïicients" 

 were used here; 1 t.c. =r about 0,5 mm. 



For the curves a and ^ (strongly smoothed means) the scale is 

 2 cms. for 1°0 C. with /i; 1cm. for 0,5 unit with «. The first dot 

 of the curve erroneously lies one scale-division too low. 



The agreement between the two curves, built up from absolutely 

 different materials, certainly does not extend to details, as a preli- 

 minary comparison of the separate historical periods had already 

 revealed. Nevertheless this agreement is, I believe, so good as to 

 be somewhat surprising. In both curves not only the dip between 

 the period-years 25 and 31 is the same, also the wave crests in the 

 following years, but during the last 40 years the full-drawn line 

 and the dotted line show the same undulations, allowance being 

 made for a small difference of phase. This similarity is striking where 

 the details have been eliminated, as fn the historical and thermo- 

 metric curves « and ^. At any rate it seems to me that the curves 

 are essentially of the same character, so as entirely to preclude 

 an accidental conformity, especially if it is borne in mind that this 

 similarity is lost as soon as the 89-year period is lengthened or 

 shortened by one year. 



It is only natural that the agreement between the curves a and 

 b is less complete than between the others ; still I think this agreement 

 is also remarkable enough to confirm the conclusion that the resulting 



