it can accumulate impression almost
indefinitely, while from the retina they fade after one-tenth part of a
second, leaving it a continually renewed _tabula rasa_.

It is, accordingly, quite possible to photograph objects so faint as to
be altogether beyond the power of any telescope to reveal--witness the
chemical disclosure of the invisible nebula encircling Nova Persei--and
we may thus eventually learn whether a blank space in the sky truly
represents the end of the stellar universe in that direction, or whether
farther and farther worlds roll and shine beyond, veiled in the
obscurity of immeasurable distance.

Of many ingenious improvements in spectroscopic appliances the most
fundamentally important relate to what are known as "gratings." These
are very finely striated surfaces, by which light-waves are brought to
interfere, and are thus sifted out, strictly according to their
different lengths, into "normal" spectra. Since no universally valid
measures can be made in any others, their production is quite
indispensable to spectroscopic science. Fraunhofer, who initiated the
study of the diffraction spectrum, used a real grating of very fine
wires: but rulings on glass were adopted by his successors, and were by
Nobert executed with such consummate skill that a single square inch of
surface was made to contain 100,000 hand-drawn lines. Such rare and
costly triumphs of art, however, found their way into very few hands,
and practical availability was first given to this kind of instrument by
the inventiveness and mechanical dexterity of two American
investigators. Both Rutherfurd's and Rowland's gratings are
machine-ruled, and reflect instead of transmitting the rays they
analyse; but Rowland's present to them a very much larger diffractive
surface, and consequently possess a higher resolving power. The first
preliminary to his improvements was the production, in 1882, of a
faultless screw, those previously in use having been the inevitable
source of periodical errors in striation, giving, in their turn,
ghost-lines as subjects of spectroscopic study.[1653] Their abolition
was not one of Rowland's least achievements. With his perfected machine
a metallic area of 6-1/4 by 4-1/4 inches can be ruled with exquisite
accuracy to almost any degree of fineness; he considered, however,
43,000 lines to the inch to be the limit of usefulness.[1654] The ruled
surface is, moreover, concave, and hence brings the spectrum to a focu