582 Proceedings of the Royal Society 
from error — much more accurate than any neiu table could possibly be.” 
In opposition to this gigantic absurdity, I had pointed out the well 
recognised danger arising from the use of moveable types : M. 
Lefort denies and yet admits this danger in one sentence : — 
'“J’admets bien qu’a la suite des tirages successives on ait pu 
rendre plus corrects les exemplaires de Ylacq, mais je n’admets 
pas qu’on les ait rendus moins corrects.” It is enough for my 
argument that two copies may differ. In the supplementary table 
to the Errata of Briggs, M. Lefort supplies a strong corroboration 
of what I advance. The logarithm of 2087 is therein stated as 
9 . . instead of 952. The two figures 52 had been drawn out or 
been broken while the copy “ Sainte Genevieve” was being printed; 
in my own copy they are correctly given. 
In order to sustain this dictum of u Nature,” we have to suppose 
the Tables du Cadastre, which were the basis of Lefort’s com- 
parison, to be absolutely correct. Now, while composing the re- 
marks made at the first meeting of the Session, I had only access 
to M. Lefort’s papers in the Comptes Bendus; but through his 
great kindness I am now in possession of a copy of his most 
valuable paper inserted in the Annales de TObservatoire de Paris, 
and am thereby enabled much more satisfactorily to explain the 
defects of Prony's mode of procedure because an example of the 
actual work is therein given. 
The design was to compute the successive differences of the 
logarithms, carrying the decimals two places further at each step ; 
and by the summation of these to obtain 200 terms of the logarith- 
mic progression. An error of unit in the last place of each of 
these differences will produce an effect on the final term, according 
to the following scale : — 
Is*, 2-00 
2 d, 1-99 00 
3d, 1-31 34 00 
4:th, -64 68 49 50 
5th, *25 35 65 00 40 
6th, *08 24 08 62 63 00 
making a total, if all the errors should happen to be in one way, of 
