82 CELESTIAL SPECTROSCOPY. 
reversion which is manifested by dark lines is to beregarded as taking 
place. 
Passing backward in the stav’s life, we should find a gradual weak- 
ening of gravity at the surface, a reduction of the temperature-gradi- 
et so far as it was determined by expansion, and convection currents 
of less violence producing less interference with the proportional quan- 
tities of gases due to their vapor densities, while the effects of erup- 
tions would be more extensive. 
At last we might come to a state of things in which, if the star were 
hot enough, only hydrogen might be sufficiently cool relatively to 
the radiation behind to produce a strong absorption. The lower vapors 
would be protected, and might continue to be relatively too hot for 
their lines to appear very dark upon the continuous spectrum; besides, 
their lines might be possibly to some extent effaced by the coming in 
under such conditions in tho vapors themselves of a continuous spec- 
trum. 
In such a star the light radiated towards the upper part ef the atmos- 
phere may have come from portions lower down of the atmosphere 
itself, or at least from parts not greatly hotter. There may be no such 
great difference of temperature of the low and less low portions of the 
star’s atmosphere as to make the darkening effect of absorption of the 
protected metallic vapors to prevail over the illuminating effect of 
their emission. 
It is only by a vibratory motion corresponding to a very high tem- 
perature that the bright lines of the first spectrum of hydrogen can be 
brought out, and by the equivalence of absorbing and emitting power 
that the corresponding spectrum of absorption should be produced; 
yet for a strong absorption to show itself, the hydrogen must be cool 
relatively to the source of radiation behind it, whether this be con- 
densed particles or gas. Such conditions, it seems to me, should oc- 
cur in the earlier rather than in the more advanced stages of conden- 
sation. 
The subject is obscure, and we may go wrong in our mode of con- 
ceiving of the probable progress of events, but there can be no doubt 
that in one remarkable instance the white-star spectrum is associated 
with an early stage of condensation. 
Sirius is one of the most conspicuous examples of one type of this 
class of stars. Photometric observations combined with its ascertained 
parallax show that this star emits from forty to sixty times the light 
of our sun, even to the eye, which is insensible to ultra-violet light, in 
which Sirius is very rich, while we learn from the motion of its com- 
panion that its mass is not much more than double that of our sun. 
It follows that, unless we attribute to this star an improbably great 
emissive power, if must be of immense size, and in a much more diffuse 
and therefore an earlier condition than our sun; though probably at 
a later stage than those white stars in which the hydrogen lines are 
bright, 
