158 ON MAGNETISM. 



short bars in its place, just touching or merely separated 

 by bits of thin paper : and the quadrantal deviation 

 will be reduced to four-fifths of its previous value, or 

 less. Or, apply the long bar sideways to the compass 

 with its advanced end abreast of the center, and note 

 the deviation; substitute the four short bars for it, 

 and the deviation will be reduced to about three-fifths. 

 The author has repeatedly tried these experiments, 

 with bars of different lengths, and has always arrived 

 at the same result. 



The following observation is precisely similar, but 

 on a much larger scale. It has been found by Capt. 

 Evans, R.N., that when a ship is built of plates of iron 

 very closely riveted together in every part, the quad- 

 rantal deviation of the compass is considerable. But, 

 when a wood-built ship is covered with heavy iron 

 armour, in plates which (though thick, and screwed to 

 the wood, and perhaps lightly touching each other) are 

 not riveted together, the quadrantal deviation is small. 

 In both these classes of experiment it is evident 

 that Poisson's fundamental suppositions are at fault. 

 It would seem that magnetism can flow through the 

 unbroken bar or the closely-riveted iron nearly as 

 through the steel of a magnet, but is not permanently 

 retained as in a steel magnet. And it appears that ; 

 instead of a thin skin of magnetism on each side of the 

 mass of iron, as in Article 65, Figure 53, where the 

 thickness of the skin is determinate without reference 

 to the depth of the iron ; there will be a dense collection 

 of magnetism of one or the other kind, brought from 



