good sight), but if it had been seen and catalogued as a
star it would have moved from its place, and the catalogue would by that
entry be wrong. The thing to detect, therefore, was errors in the
catalogues: to examine all entries, and see if the stars entered
actually existed, or were any of them missing. If a wrong entry were
discovered, it might of course have been due to some clerical error,
though that is hardly probable considering the care taken over these
things, or it might have been some tailless comet or other, or it might
have been the newly found planet.

So the next thing was to calculate backwards, and see if by any
possibility the planet could have been in that place at that time.
Examined in this way the tabulated observations of Flamsteed showed that
he had unwittingly observed Uranus five distinct times, the first time
in 1690, nearly a century before Herschel discovered its true nature.
But more remarkable still, Le Monnier, of Paris, had observed it eight
times in one month, cataloguing it each time as a different star. If
only he had reduced and compared his observations, he would have
anticipated Herschel by twelve years. As it was, he missed it
altogether. It was seen once by Bradley also. Altogether it had been
seen twenty times.

These old observations of Flamsteed and those of Le Monnier, combined
with those made after Herschel's discovery, were very useful in
determining an exact orbit for the new planet, and its motion was
considered thoroughly known. It was not an _exact_ ellipse, of course:
none of the planets describe _exact_ ellipses--each perturbs all the
rest, and these small perturbations must be taken into account, those of
Jupiter and Saturn being by far the most important.

For a time Uranus seemed to travel regularly and as expected, in the
orbit which had been calculated for it; but early in the present century
it began to be slightly refractory, and by 1820 its actual place showed
quite a distinct discrepancy from its position as calculated with the
aid of the old observations. It was at first thought that this
discrepancy must be due to inaccuracies in the older observations, and
they were accordingly rejected, and tables prepared for the planet based
on the newer and more accurate observations only. But by 1830 it became
apparent that it would not accurately obey even these. The error
amounted to some 20". By 1840 it was as much as 90', or a minute and a
half. This discrep