112 CONDUCTION AND RADIATION [CH. x. 



to until, as hinted above, Zeeman of Amsterdam, in 

 1897, with a good grating and a strong electromagnet, 

 skilfully observed a minute effect, consisting in a 

 broadening of the lines emitted by a sodium flame 

 placed between its poles. On seeing a two-line 

 notice of this in Nature in December, 1896, Dr. 

 Larmor wrote to me, saying that this must be the 

 effect which he had thought of, but concluded must 

 be too small to see. On receiving this intimation 

 I immediately, with a little trouble, repeated and 

 verified the experiment,* and exhibited it at the 

 Royal Society soire'e in May that same year. 



From this simple but important beginning the 

 large subject of the influence of a magnetic field on 

 the radiation from different substances has been 

 laboriously worked at ; not only by the original 

 discoverer, but by Preston in Dublin, Michelson in 

 America, Runge, and others ; and a whole series of 

 important facts have been made out. Every line has 

 been studied separately ; some lines are tripled, some 

 quadrupled, some sextupled, and so on as said above. 

 One mercury line is resolved into nine or perhaps eleven 

 components. The effect is therefore not too small to 

 see, though it needs excessively high power and per- 

 fect appliances to display it ; and so it became evident 



* See Proc. Roy. Soc. t vol. 60, pp. 466, 513, and vol. 61, p. 413, or Nature, 

 vol. 56, p. 237 ; also several articles by Lodge in The Electrician, for 1897, 

 vol. 38. The whole matter is elucidated by Zeeman, aided by Lorentz, 

 on the basis of theory illustrated by a picture or model of an orbitally 

 revolving electron, which, though crude, was adequate as a guide : the 

 small mass of the revolving particle being thereby deduced, and being 

 in general conformity with J. J. Thomson's direct determinations of 

 the mass of an electron some months previously. With higher ex- 

 perimental power greater precision was reached, and an unexpected 

 development appeared in the tripling of each line, a result which was 

 suggested by the model, but could not have been predicted from it 

 alone. Other lines were found to divide into more than three com- 

 ponents, in a very suggestive but still imperfectly understood manner, 



