86 Proceedings of Royal Society of Edinlurgh. [sess. 
volume of the iron tube as to compensate accurately for the degree 
or two of temperature variation that might occur in the course of a 
single experiment. But it is obvious that the iron tube will be first 
affected by a change of temperature coming in from without, and 
the glass last of all. Consequently, it seemed altogether a vain quest 
to seek for anything like a perfect compensation. 
The liquid finally chosen to fill the volume of the tube was 
water. Mercury and coloured alcohol were both tried, simply to be 
discarded ; the former because of its tendency to go by fits and starts 
in the capillary tube, the latter because of its high expansibility. 
Only one iron tube, No. I., has so far been satisfactorily 
experimented with. Rough trials with Tubes II. and III. show 
that similar effects are produced with them. The present note, 
however, refers only to No. I., which has the widest bore or 
thinnest wall of the five. The very interesting results obtained with 
it call for a brief description. 
On reference to the paper already cited, it will be seen that up to 
the magnetic fields used the internal capacities of the iron tubes all 
diminished. It was pointed out, however, that a maximum change 
seemed to have been about to be reached ; and, reasoning from the 
remarkable behaviour of the nickel tube, we might expect not only 
the existence of a maximum contraction for a moderate field, but 
even a change from contraction to dilatation at higher fields. Such 
a result would not be surprising in the light of Mr Bidwell’s 
experiments on the changes of length of magnetised rods. 
But the extraordinary fact to be noted is that, with Tube I., we 
get not only a single maximum and a change of sign in the dilatation, 
but three maxima (or two maxima and one minimum) and two 
changes of sign. It will suffice to give one set of measurements. 
Very similar results were obtained in all the really good experi- 
ments. 
The bore-space filled by the water during the experiment had a 
volume of 343 "75 cub. cm., with a cross-section of approximately 
8 sq. cm. and a length of nearly 42 ‘7. The thickness of wall was 
*33 cm., the external and internal diameters being respectively 3*84 
and 3T9 cm. 
The tube when being magnetised was well included in the 
magnetising coil. 
