476 THE LIFE OF JAMES J). FORCES. [CHAP. 



'Of the latter I have read a few pages respecting 

 Mitchell. You hold as I do that he makes the atrocious 

 blunder of saying that the antecedent improbability of 

 an event is the consequent probability of its having a 

 physical cause, e.g. that if the odds against throwing 

 aces with two dice be 1 to 35, then (the aces having 

 turned up) the odds for the dice being loaded or some 

 equivalent cause is 35 to 1. 



' But then this blunder is so gross that I can scarcely 

 believe it could be swallowed by Herschel or any of the 

 men of science who have endorsed Mitchell's argument. 

 This leaves a sort of vague suspicion upon my mind that 

 we do not do Mitchell justice. 



' I am gratified by your quotation from Laplace as to 

 regular distributions for I had written down some- 

 thing to the same purpose, with the addition that it is 

 regularity which induces us to believe in design, and 

 that as there is no unit of perfect regularity there can be 

 no definite expression of imperfect regularity by means 

 of such fractions as Mitchell gives. 



' I am, my dear Sir, yours faithfully, 



' C. H. TERKOT, Bp.' 



From SIR GEORGE AIRY. 



'KoYAL OBSERVATORY, GREENWICH, October llth, 1850. 

 ' . . . "Article 23, beginning/' This is the only place 

 in which you have adverted to the long run of many 

 trials as entering into a chance-problem : and I think 

 the want of more energetic reference to it is a defect 

 in your paper. I think that that consideration is the 

 foundation of all calculations of probabilities. 



* Article 28 : " An axiom which cannot be admitted as 

 certainly true/' I think in discussing Mitchell on chances 

 you must admit it as certainly true. It is the mathe- 

 matical expression of chance. The denying it amounts 

 simply to saying that there cannot be any chance, 

 in a shape for mathematical consideration, in the whole 

 matter ; and it might as well be said in so many words. 



* Article 30 : " Tending ultimately to uniformity of 



