According to Dr. A. W. Roberts, the components of the following "Algol variables" "revolve in contact": V Puppis, X Carinæ, β Lyræ, and υ Pegasi. Of those V Puppis and β Lyræ are known spectroscopic binaries. The others are beyond the reach of the spectroscope, owing to their faintness. A very curious variable star of the Algol type is that known as R R Draconis. Its normal magnitude is 10, but at minimum it becomes invisible in a 7½-inch refracting telescope. The variation must, therefore, be over 3 magnitudes, that is, at minimum its light must be reduced to about one-sixteenth of its normal brightness. The period of variation from maximum to minimum is about 2·83 days. The variation of light near minimum is extraordinarily rapid, the light decreasing by about 1 magnitude in half an hour.[323] A very remarkable variable star has been recently discovered in the constellation Auriga. Prof. Hartwig found it of the 9th magnitude on March 6, 1908, the star "having increased four magnitudes in one day, whilst within eight days it was less than the 14th magnitude."[324] In other words its light increased at least one-hundredfold in eight days! The period of the well-known variable star β Lyræ seems to be slowly increasing. This Dr. Roberts (of South Africa) considers to be due to the component stars slowly receding from each other. He finds that "a very slight increase of one-thousandth part of the radius of the orbit would account for the augmentation in time, 30{m} in a century." According to the theory of stellar evolution the lengthening of the period of revolution of a binary star would be due to the "drag" caused by the tides formed by each component on the other.[325] M. Sebastian Albrecht finds that in the short-period variable star known as T Vulpeculæ (and other variables of this class, such as Y Ophiuchi), there can be no eclipse to explain the variation of light (as in the case of Algol). The star is a spectroscopic binary, it is true, but the maximum of light coincides with the greatest velocity of _approach_ in the line of sight, and the minimum with the greatest velocity of _recession_. Thus the light curve and the spectroscopic velocity curve are very similar in shape, but one is like the other turned upside down. "That is, the two curves have a very close correspondence in phase in addition to correspondence of shape and period."[326] The star now known as W Ursæ Majoris (the variability of which was discovered by Müller and Kempf in 1902), and which lies between the stars θ and υ of that constellation, has the marvellously short period of 4 hours (from maximum to maximum). Messrs. Jordan and Parkhurst (U.S.A.), find from photographic plates that the star varies from 7·24 to 8·17 magnitude.[327] The light at maximum is, therefore, more than double the light at minimum. A sun which loses more than half its light and recovers it again in the short period of 4 hours is certainly a curious and wonderful object.