39.2 42.0 Alph.Gruris, 2d Sept....

27.4 42.2 what are the chances that, by one error all the latitudes in the first column should be brought so nearly to an agreement as they are in the second column?
The circumstance of the number of divisions of the level being almost arbitrary within limits, might perhaps be alleged as diminishing this extraordinary improbability: but let any one consider, if he choose the error of each set, as independent of the others, still he will find the odds against it enormous.
When it is considered that an error, almost arbitrary in its law, has thus had the effect of bringing discordant observations into an almost unprecedented accordance, as at Maranham; and not merely so, but that at eight of the nine stations it has uniformly tended to diminish the differences between the partial results, and that at the ninth station it only increased it by a small fraction of a second, I cannot help feeling that it is more probable even that Captain Kater, with all his admitted skill, and that Captain Sabine himself, should have been both mistaken in their measures of the divisions of the level, than that so singular an effect should have been produced by one error; and I cannot bring myself to believe that such an anticipation is entirely without foundation.
Whatever may be the result of a re-examination, it was a singular oversight NOT TO MEASURE the divisions of a level intended to be used for determining so important a question; more particularly as, in the very work to which reference was made by Captain Sabine for the purpose of comparing the observations, it was the very first circumstance which occupied the French philosophers, and several pages [See pages 265 to 275 of the RECUEIL D'OBSERVATIONS GEODESIQUES, &c.

PAR MM.

BIOT ET ARAGO, which forms the fourth volume of the BASE DU SYSTEME METRIQUE.] are filled with the details relative to the determination of the value of the divisions of the level.

It would also have been satisfactory, with such an important object in view, to have read off some of the sets after each pair of observations, in order to see how far the system of repetition made the results gradually converge to a limit, and in order to know how many repetitions were sufficient.