554 Proceedings of the Royal Society 
no red visible, as a little time was taken getting any adhering metal 
knocked off. 
(3) A piece of pig iron 8-F inches long floated in about six 
seconds, and appeared to have gained about T \ of an inch in 
length. 
(4) A piece of test-bar 2 in. x 1 in. x 9 in. floated in about four 
seconds ; showed from Jg- in. to in. increase in length, and was a 
dull red ; broken through, and a piece of lead laid upon the surface, 
the lead did not melt, but did so afterwards when pressed between 
the two broken parts of the bar. As lead melts about 630° F., the 
temperature of the iron could not have been much above this. 
(5) A piece of a casting 18J- inches long and of triangular section 
was found to measure 18 T 7 F inches after floating, equal to -- of an 
inch of increase, was dull red ; afterwards, when quite black, had 
in. to - 3 3 2 in. of expansion. A piece of lead laid upon it did not 
melt for some time. 
(6) Another piece of a test-bar 2 in. x 1 in. x 12 in. was tried ; 
after floating, it was taken out and found to measure 12 J inches; 
here again an increase of -J of an inch was obtained, and the bar 
was barely a dull red. This piece was again put in ladle, floated 
easily, and was taken out somewhat red, callipered, and found to be 
12^ inches in length.* 
(7) Other experiments have been made, but sufficient instances have 
been given to prove the large increase of bulk which so rapidly takes 
place in all these floating pieces of cast iron. The author therefore 
believes that the cause of the floatation is in the increased bidk thus 
obtained, which from its extent is quite sufficient in itself to explain 
the phenomena satisfactorily, as it will be observed that the incre- 
ments found to exist were practically equal to the decrement which 
occurs from the molten to the finally cooled state, and that, there- 
fore, the floating pieces had become of equal bulk to an equal weight 
of the liquid metal. It may be mentioned that great care was taken 
to get the real surface of the pieces after being taken out of the 
ladles, and in some cases the parts to be measured were oiled or 
coated with blacking to keep the surface free of scum or loose metal. 
* From measuring the pattern of these test-bars and the test-bars themselves 
there appears a difference of § of an inch in a length of 42 inches, showing a 
shrinkage of rather less than § of an inch to the foot. 
