the large percentage of water they usually contain. Thus,
four samples of ordinary plaster analyzed by Landrin have an average
of 90.17 per cent. of CaSO4 and 7.5 per cent. of water, while two
samples of best plaster contained 89.8 per cent. of CaSO4 and 7.93 per
cent. of water. These numbers do not add up to 100, the difference
being due to silica and other impurities of the original gypsum,
amounting altogether to about 3 per cent.

It might be suggested that the reason why these plasters set more
slowly than completely dehydrated plaster is owing simply to the fact
that they contain, apparently, some unaltered gypsum, which serves to
_dilute_ the action. Were this so, a similar result, as far as time of
setting is concerned, should be obtained with a plaster containing a
corresponding quantity of dead-burnt material. This, however, is not
found to be the case. The time of setting appears, then, to be
connected in some special and peculiar manner with the retention of
water by the burnt plaster.

The following explanation of this connection is offered, an
explanation only tentative at present, owing to want of experimental
data.

The following substances are known:

Gypsum, and set plaster, CaSO4 + 2 H2O, containing 20.93
per cent. of water.

Plaster completely burned at moderate temperature, CaSO4,
probably amorphous.

Anhydrite and dead-burned plaster, CaSO4, crystalline.

Selenitic deposit from boilers, 2 CaSO4 + H2O, or CaSO4 +
1/2 H2O, containing 6.2 per cent. of water.

The circumstance that the hot calcium sulphate can crystallize with 1/4
its normal amount of water indicates that for this proportion of water
it has a greater attraction than for the other 3/4. Having a similar
bearing is the fact that when burned at lower temperatures, gypsum
only loses the last portions of water with extreme slowness.

Now, if it be the case that anhydrous calcium sulphate has a greater
attraction for the first half molecule of water, then the operation of
hydration will proceed very rapidly at first, more slowly afterward.
Many such cases are known, e.g., that of copper sulphate. Conversely,
if only 3/4 of the water of hydration be expelled during the baking of
gypsum, the material obtained should hydrate itself more slowly. For
our present purpose it will be convenient to recalculate the numbers
given by Landrin (_vide supra_) so as to make the calcium sulphate and
water add up to 100