82 EDWARD M. NELSON ON DIATOM STRUCTURE. 



This is a difficult image, not only on account of its minuteness, 

 but because of its liability to be merely an interference image. 

 There are certain minute microscopic images which, although 

 difficult to see, are nevertheless undoubtedly true, for by no 

 known method could they have been caused by interference. 

 There are, on the other hand, some comparatively large micro- 

 scopic images, which are so likely to have been caused by inter- 

 ference, that, apart from some special proof, they cannot be 

 regarded as true. The Syste2)hania diadema = Stephanopyxis 

 corona, Fig. 4, PI. 20, Yol. 4, Ser. 2, p. 316, may be cited as an 

 example of this latter class of object. There the structure, 

 comparatively coarse, is one that is in all respects very similar to 

 an interference image — so much so that for loDg I regarded it as 

 such ; but further search revealed a broken piece with some of 

 the supposed ghost structures sticking over the edge, which 

 proved the image to be unquestionably a true one. In the 

 case now before us, however, its very minuteness prevents this 

 test being applied, for it would be quite impossible to see at a 

 fractured edge such a structure, even though it were several 

 degrees coarser. 



In micrpscopical images all fractures and edges are bounded 

 by an umbra, coma, or undefined margin. This umbra tends to 

 blot out all finer structures in its immediate vicinity. Thus the 

 umbra of a primary blots out (either partially or wholly) the image 

 of the secondary; much more, therefore, would it blot out a tertiary. 

 For example, the postage stamp fracture in a Triceratium favus 

 is not so easily seen as it might be on account of the umbra of 

 the large primary structure. On the other hand there is no similar 

 difficulty found with the postage stamp fracture of a P. angidatmn 

 or a N. rhomboides, for there is no coarser primary present. 



It is obvious, therefore, that we must resort to some test of 

 truth other than that of a fractured edge. Let us, therefore, try 

 counting the number of spots in each aperture, for if they are 

 ghosts they will be replicas of some pattern — e.g. six spots sur- 

 rounding a central one. This, though a difficult operation, has 

 been done, and both a five and a four have been seen, and some- 

 times the central one has been found w^anting ; further, these 

 observations have been confirmed by an independent observer ; 

 therefore we may conclude that the image is not that of a ghost, 

 but an indication of an actual structure. 



