SCIENCE. 



[Vol. XVI. No. 388 



SCIENCE: 



A WEEKLY NEWSPAPER OF ALL THE ARTS AND SCIENCES. 



PUBLISHED BY 



N. D. C. HODGES. 



47 Lafayette Place, New York. 



of tornadoes occurred in those years. It is to be noted that 

 the reports of tornadoes must necessarily increase as houses 

 increase in the tornado States, but this will be balanced by 

 the fact that fire losses will also increase. It is perfectly safe 

 for us to compare fire and tornado losses, and to determine 

 approximately what the comparative premium should be. 



The accompanying table shows for the years 1876-84, in 

 the seventeen tornado States, the total loss by tornadoes of 

 scale (3), (2), (1); the loss by fire during the same years; 

 and the relative loss by the two. 



Tornado and Fire Losses in Seventeen States, 1876-84. 



Subscriptions.— United States and Canada $3.50 a year. 



Great Britain and Europe 4.50 a year. 



Communications will be welcomed from any quarter. Abstracts of scientific 

 papers are solicited, and twenty copies of the issue containing sucb will be 

 mailed the author on request in advance. Rejected manuscripts will be 

 returned to the authors only when the requisite amount of postage accom- 

 panies the manuscript. Whatever is intended for insertion must be authenti- 

 cated by the name and address of the writer; not necessarily for publication, 

 but as a guaranty of good faith. We do not hold ourselves responsible for 

 any view or opinions expressed in the communications of our correspondents. 



Attention is called to the "Wants" column. All are invited to use it in 

 soliciting information or seeking new positions. The name and address of 

 applicants should be given in full, so that answers will go direct to thenj. The 

 *' Exchange " column is likewise open. 



NEW YOEK. July 11, 



No. 388. 



CONTENTS: 

 Tornado Losses and Insurance 



H. A. Hazen 15 



Notes and News 19 



M. Pasteur and Hydrophobia 



Thomas M. Dolan 2.3 



Book-Reviews. 



Advanced Physiography 25 



Graphical Statics 25 



Cycling Art, Energy and Loco- 

 motion 25 



Steam 26 



Among the Publishers 96 



TORNADO LOSSES AND INSURANCE. 



[Continued from p. 19.] 

 Constancy of Tornadoes. 

 It is an exceedingly important question to determine, if 

 possible, whether tornadoes are on the increase, or whether 

 we may reason from their occurrence in the past as to their 

 probable action in the future. The tendency is to take 

 serious alarm when a tornado has happened near a com- 

 munity, and to conclude at once that the risk has suddenly 

 become very much greater than before, and is likely to con- 

 tinue. We may cite as an instance the alarm in a good 

 many minds after the occurrence of the recent Louisville 

 tornado. Notwithstanding the fact that this was the first 

 serious tornado in Kentucky, and that in all probability 

 another such would never occur there, certainly not in 

 Louisville, yet many seem to have thought that a change 

 had come in the climate, and we were now to have more of 

 these outbursts. A careful study has shown that such fears 

 are groundless, and that our climate is practically constant. 

 We may strike an average of tornadoes in the past, and 

 reason that there will be no more than that average in the 

 future. In fact, if we include 1883 and 1884 in our average, 

 we know that it will be too great, for a very large number 



The tornado loss in this table has been rigidly computed 

 from the actual estimated loss for scale (3), and allowing 

 $20,000 each for (2), and $3,000 each for (1). There is no 

 doubt that this loss is more than ten per cent too great. It 

 should also be remembered that during the two years 1883 

 and 1884 there were as many tornadoes as in the remaining 

 seven, and in the sixteen years thus far studied this is very 

 nearly the relation. There are several most astonishing 

 facts brought out in this table, and such as are very difficult 

 to understand. Perhaps the most surprising result is that 

 in Kansas the tornado loss is one-third of that by fire. I 

 have gone over the data, and can find no flaw. All the tor- 

 nado reports from the difi^erent States were treated exactly 

 alike, and the result for any one State may be compared di- 

 rectly with that for any other. In Nebraska, a neighboring 

 State, the tornado loss is insignificant, not more than one- 

 twentieth of that by fire; and in Missouri it is one twelfth 

 of the fire loss. Of course, where no estimate of the loss is 

 given, as is the case in more than half the occurrences, the 

 position in the scale is dependent on the reported violence of 

 the tornado. I am inclined to think that Kansas has had 



