We now come to the long-period variables of which Mira Ceti, χ Cygni, and U Orionis are examples. Although varying in regular periods, generally of about a year, they are subject to remarkable irregularities, so that an exact period cannot be assigned even to Mira Ceti, of which the maxima are at times retarded and at others accelerated with no apparent law. The spectroscopic investigations of Campbell in 1898 have shown that Mira Ceti is a solitary star, while bright lines of hydrogen appear in its spectrum at maximum, showing that the variations are due to periodical conflagrations in its atmospheres. In many other long-period variables bright lines have been observed. A remarkable fact regarding these stars is the amount of their light change. Mira Ceti, for instance, varies from the first to the ninth magnitude, and U Orionis from the sixth to the twelfth. As M. Flammarion says, “the longer the period the greater the variation.” Another remarkable fact is that their light curves show a curious resemblance to the curves of the solar spots, only on a vastly greater scale, which indicates that, relatively, these long-period variables are much older than our Sun, the small variations in the light of which are imperceptible. “Here, if anywhere,” says Miss Clerke, “will be found the secret of stellar variability.” To the irregular variables no period can be assigned. Betelgeux, in Orion, the variation of which was noted by Sir John Herschel in 1840, is a typically irregular variable. But the most extraordinary of all variables is η Argus, in the southern hemisphere, which is probably a connecting link between variable and temporary stars. The traveller Burchell, from 1811 to 1815, observed the star as of the second magnitude, but in 1827 he noted it to be of the first magnitude. In the following year it fell to the second magnitude. In 1834 Sir John Herschel noted the star to be between the first and second magnitude, and in 1838 it rose to the first, being equal to α Centauri. After a decline, it became in 1843 equal to Canopus, and not much inferior to Sirius. Then it began to fade, and in 1868 it was only of the sixth magnitude. In 1899 Innes estimated it as 7·71. Rudolf Wolf suggested a period of 46 years, and Loomis 67 years; but astronomers generally agree with Schönfeld that the star has no regular period. The first temporary star of the nineteenth century was discovered by Hind, in London, April 28, 1848. It was of the fifth magnitude at maximum, and soon after began to fade, falling to the tenth magnitude. In 1860 a new star appeared in the cluster Messier 80 in Scorpio, and was discovered by Auwers at Königsberg. It reached only the seventh magnitude.