ANALYSES OF ROMAN GLASS FROM SILCHESTER. Tn 
very fair results and is specially applicable to samples in 
which the manganese percentage is very low. 
The Alkalies were determined by the Lawrence Smith 
method,’ i.e., by converting to chlorides by heating with 
calcium carbonate and ammonium chloride and then sepa- 
rating the two alkali chlorides with hydroplatinie acid. 
The remaining constituents, were determined by the 
ordinary methods and call for no special comment. 
SUMMARY OF ANALYSES. 
I. II. Mean 
Hygroscopic H.O ... °40 °42 °41 
*Combined H.O ... °00 00 °00 
SiO. ace O9SSS 69°67 69°76 
AleOs. 3.6 2) 1:89 1°94 1°92 
"FeO so a1 ia 74 
Fe;,O3  ... 16 18 vals 
Min@ ec 20 61 °65 
CaO cover eh OO 7°08 7°08 
MgO case Ile 1°07 1°09 
Na,O ... 17°34 WI 17°53 
2K,O aa 39 38 D9 
Totals wis w. 99°84 100°04 99°94 
1 Penfield’s method. *% Cooke’s method. * Lawrence Smith’s method. 
Since the specimen (A) had a greenish tinge, due to iron, 
the obvious inference was that if manganese had been 
added purposely as a decolouriser then not quite enough had 
been used. In the nearly colourless variety (B), it was ex- 
pected that the manganese had neutralised the effect of the 
iron and therefore that the ratio of manganese to iron would 
be greater than in (A). In the bluish-green variety (C), 
similarly a large excess of iron was expected. The analyses 
‘J. L. Smith—Am. Journ. Sci., 1., p. 269, 1871. 
