In _Purchas his Pilgrimes_ (Lond., 1625), in the Narrative, in vol. iii., of Bylot and Baffin's Voyage of 1616, there is mentioned an island between Whale-Sound and Smith's Sound, where there had been observed a larger variation than in any other part of the world. Purchas, in a marginal note, comments on this as follows: "Variation of the Compass 56° to the West, which may make questionable D. Gilbert's rule, tom. 1., l. 2, c. 1, that where more Earth is more attraction of the Compass happeneth by variation towards it. Now the known Continents of Asia, &c., must be unspeakably more than here there can be, & yet here is more variation then about Jepan, Brasil, or Peru, &c."
Gilbert's view was in truth founded on an incomplete set of facts. At that time, as he tells us, the variation of the compass at London was 11-1/3 degrees eastward. But he did not know of the secular change which would in about fifty-seven years reduce that variation to zero. Still less did he imagine that there would then begin a westward variation which in the year 1816 should reach 24° 30', and which should then steadily diminish so that in the year 1900 it should stand at 16° 16' westward. For an early discussion of the changes of the variation see vol. i. of the _Philosophical Transactions_ (Abridged), p. 188. Still earlier is the classical volume of Henry Gellibrand, _A Discovrse Mathematical on the Variation of the Magneticall Needle_ (Lond., 1635). Gilbert heads chapter iii. of book iiii. (p. 159) with the assertion _Variatio uniuscuiusque loci constans est_, declaring that to change it would require the upheaval of a continent. Gellibrand combats this on p. 7 of the work mentioned. He says:
"Thus hitherto (according to the Tenents of all our _Magneticall_ Philosophers) we have supposed the variations of all particular places to continue one and the same. So that when a Seaman shall happly returne to a place where formerly he found the same variation, he may hence conclude he is in the same former _Longitude_. For it is the Assertion of _Mr. Dr. Gilberts_. _Variatio vnicuiusq; Loci constans est_, that is to say, the same place doth alwayes retaine the same variation. Neither hath this Assertion (for ought I ever heard) been questioned by any man. But most diligent magneticall observations have plainely offred violence to the same, and proved the contrary, namely that the variation is accompanied with a variation."