126 



SCIENCE. 



[Vol. IV., No. 80. 



extension of such foundations as the latter in 

 this generous country, does there not lurk a 

 possible danger, — a danger that their bestowal 

 will fall into hands incapable of proper ad- 

 ministration? If an} 7 one think this danger 

 remote, let him reflect on the ill-judged selec- 

 tion of recipients for honorary degrees in man} 7 

 of our best universities and colleges. Let such 

 foundations remain, as now, in the hands of 

 those whose position has been gained solely 

 by research, and the danger vanishes. 



The standard of light adopted by the Paris 

 electrical conference last April is the amount 

 of light emitted by a square centimetre of 

 melted surface of platinum at the point of 

 solidification. It was believed that advantage 

 could thus be taken of a physical constant 

 (namely, the melting-point of platinum) upon 

 which could be based all our present changing 

 and unsatisfactory photometric standards. 

 The adoption of this standard has been much 

 criticised, for it does not seem to lend itself 

 easily to actual photometric tests. Werner 

 Siemens proposes that a piece of platinum foil 

 should be enclosed in a cavit}' provided with a 

 conical opening 0.1 of a square centimetre ; 

 this piece of platinum to make part of an elec- 

 trical circuit, the current in which can be so 

 regulated that a comparison with an} T light can 

 be made at the moment of fusion. The tem- 

 peratures of solidification and fusion of plati- 

 num do not differ sensibly from each other, and 

 Siemens believes that the error introduced b} 7 

 taking the temperature of fusion instead of that 

 of solidification would be small. The use of an 

 electrical current to produce fusion has certain 

 advantages, for the time of fusion can evi- 

 dently be deferred until the proper momejit. 

 Preliminary experiments have shown that the 

 light emitted from an opening 0.1 of a square 

 centimetre in section by Siemens 's method is 

 equivalent to nearly one and a half standard 

 English candles. 



Although the standard adopted by the Paris 

 conference seems to be based upon the unalter- 

 able laws of matter, it does not seem as if 

 it would ever be practically adopted. Some 



form of the modern incandescent electric light, 

 it seems to us, would afford a much better 

 prospect of a standard light. It is difficult 

 to maintain the steadiness of such a light 

 for photometric purposes ; but this does not 

 seem impossible to accomplish. It is evident, 

 that, if we could maintain an electrical current 

 constant through a platinum wire or carbon 

 filament in a suitable medium, we should have 

 the means of reproducing the same amount of 

 heat, and therefore light, from the same area. 

 Unfortunately, carbon changes in resistance 

 at the point of incandescence ; and the resist- 

 ance of platinum is not invariable under re- 

 peated heating and cooling in a comparative 

 vacuum. An exhaustive investigation of the 

 peculiarities of platinum or of iridium, under 

 the effect of incandescence produced b} 7 the elec- 

 trical current, would seem to be desirable before 

 the French standard is accepted as a finalit} T . 



LETTER TO THE EDITOR. 

 Tornado predictions. 



In an article on ' Tornado predictions,' published 

 in the July number of the American meteorological 

 journal, a table of verifications is given, in which the 

 average of successful predictions for several months 

 is from ninety-six to ninety-eight per cent. 



An examination of the table shows that this re- 

 markably high percentage of verification is largely 

 made up, not of successful predictions Of tornadoes, 

 but of successful predictions of no tornadoes. In 

 justification of this method of verification, the writer 

 says, "It requires as much and often more study to 

 say that no tornadoes will occur, as to make the 

 prediction that conditions are favorable for their 

 development." If this explanation be accepted as 

 satisfactory, what do the verifications signify ? 



A little consideration will show that the absolute 

 value of these figures gives no basis from which to 

 judge of the real success of the tornado predictions. 

 The averages of ninety-six and ninety-eight per cent 

 are mainly functions of the non-tornado days, with 

 but slight modifications for the success or failure of 

 the prediction of actual tornadoes. An ignoramus in 

 tornado studies can predict no tornadoes for a whole 

 season, and obtain an average of fully ninety-five per 

 cent. The value of the expert work must, therefore, 

 be measured by the excess which is obtained over 

 the man who knows nothing of the subject. This is 

 the only way to determine any significance in the 

 method of verification above described. The excess 

 is but one or two per cent, and poorly exhibits the 

 present stage of progress in tornado studies. The in- 

 justice which is done is to be found in the method of 

 verification adopted. In ascertaining the value of tor- 

 nado or any other special storm predictions, the con- 

 sideration of days on which no storms occur, and 

 none are predicted, is entirely beside the question. 



If the writer of ' Tornado predictions' will give the 

 verifications obtained from positive predictions, and 



