A Letter from the late Reverend Mr. Thomas Bayes , F. R. S. to John Canton , M. A. and F. R. S. SIR , IF the following observations do not seem to you to be too minute , I should esteem it as a favour , if you would please to communicate them to the Royal Society .
It has been asserted by some eminent mathematicians , that the sum of the logarithms of the numbers 1.2 .
3.4 .
&amp;c. to z , is equal to 1/ 2 log .
c+z+ 1/ 2 ~ log .
z lessened by the series z 1/ 12z + 1/ 360z3 1/ 1260z5 + 1/ 1680z7 1/ 1188z9 + &amp;c. if c denote the circumference of a circle whose radius is unity .
And it is true that this expression will very nearly approach to the value of that sum when z is large , and you take in only a proper number of the first terms of the foregoing series : but the whole series can never properly express Read Nov. 24 , press any quantity at all ; because after the 5th term the coefficients begin to increase , and they afterwards increase at a greater rate than what can be compensated by the increase of the powers of z , though z represent a number ever so large ; as will be evident by considering the following manner in which the coefficients of that series may be formed .
Take a= 1/ 12 , 5b= a2 , 7c=2b a , 9d=2ca+ b2 , 11 e= 2da+2c b , 13 f=2ea+2db+ c2 , 15 g=2fa+ 2eb+2d c , and so on ; then take A= a , B=2 b , C= 2X3X 4c , D=2X3X4X5X6 d , E=2X3X4X5 X6X7X8e and so on , and A , B , C , D , E , F , &amp;c. will be the coefficients of the foregoing series : from whence it easily follows , that if any term in the series after the 3 first be called y , and its distance from the first term n , the next term immediately following will be greater than nx 2n-1/ 6n +9X y/ z2 .
Whereforeatlength the subsequent terms of this series are greater than the preceding ones , and increase in infinitum , and therefore the whole series can have no ultimate value whatsoever .
Much less can that series have any ultimate value , which is deduced from it by taking z= 1 , and is supposed to be equal to the logarithm of the square root of the periphery of a circle whose radius is unity ; and what is said concerning the foregoing series is true , and appears to be so , much in the same manner , concerning the series for finding out the sum of the logarithms of the odd numbers 3 .
5 .
7 .
&amp;c. ... z , and those that are given for finding out the sum of the infinite progressions , in which the several terms have the same numerator whilst their denominators are any certain power of numbers increasing in arithmetical proportion .
But it is needless particularly to insist upon these , because one instance is sufficient to show that those methods are not to be depended upon , from which a conclusion follows that is not exact .

