( 306 ) 



Hence in the two first tables the number of concordances is much 

 larger than that of the discrepancies, and, moreover, in tlie favourable 

 cases the mean deviations of number of summerdays and mortality 

 are greater, so that in our opininion no doubt remains concerning 

 the relation between the two phenomena. 



The résumé for the third tabic clearly shows that the parallelism 

 with the deviations of the mean monthly temperatures is not so 

 good as with the number of summerdays. 



We may now try to find the factor, which can express this relation 

 approximately. For this purpose we take the sum of all positive 

 deviations of the number of summerdays, and compute the algebraic 

 sum of the corresponding deviations of the mortality of infants. Likewise 

 for the negative deviations of the summerdays. We thus find : 



Sum pos. dev. T Sum *S' Sum neg. dev. T Sum *S' 

 Groningen + 81 +4(33 - 83 -352 



Utrecht (town) +120 +502 -120 -465 



Without giving all numbers in detail we may add here the result 

 for the province of Utrecht, where 54 concordances occur against 

 1 7 discrepancies : 



Utrecht (prov.) + 133 +543 -118 -433 



The difference between sum S for positive and negative deviations 

 of T, rather large for Groningen, is probably an indication of a 

 non linear relation. If we overlook this and combine the two kinds 

 of deviations, we find for the factor sought for Groningen about 

 5.0, for Utrecht (town) about 4.0, Utrecht (prov.) about 3.9. 



6. What precedes still leaves open to doubt whether the high 

 temperatures themselves cause the increased mortality, or, as Prof. 

 Saltet and Mr. Falkenburg supposed, temperature-fluctuations con- 

 nected with them. It might indeed be imagined that, as a rule, 

 numerous high temperatures would be accompanied by numerous 

 large temperature fluctuations. Though, from a meteorological point 

 of view, this supposition did not look very probable, we have tested 

 it likewise, for which purpose the annals of the Royal Netherlands 

 Meteor. Institute are quite sufficient. For the same periods from the 

 middle of one month to the middle of the next for which the sum- 

 merdays had been counted, we computed for Groningen the sum of 

 the differences of the maximum temperature from one day to another, 

 for simplicity neglecting the tenths of degrees. The results were 

 dealt with in the same way as before: hence the deviations of the 



