114 THE AMERICAN MONTHLY [July 



will be difficult to find anything sufficiently contrasted in 

 detail to see at all, though the very same illumination 

 of a black- and-white photograph of small scale, or of the 

 same diatom in a medium of 2 "4 index, might show it 

 easily. But plane-wave illumination might very easily 

 bring about phase relations more or less approximating 

 to half -wane discordance, which we know well would be 

 more effective than black-and-white itself by direct light ; 

 in any case these phase-relations will produce conspicuous 

 effect in a Fresnel-fringe image. Thus the Abbe method 

 has a most important function in enabling us to see 

 contrast in the details of a large class of objects — especi- 

 ally hyaline or transparent objects — which do not present 

 contrast or opacity sufficient to be seen in any other way. 

 The error has been in giving to it the sole or all-im- 

 portant place, not recognizing that there is quite another 

 kind of image also available, depending upon Airy's 

 theory; and that this latter, while in the the case of 

 transparent details often giving images insufficient, or at 

 least far inferior, in black-and-white contrast (what 

 microscopists call " resolution " ) , is free from the 

 contour errors of the Abbe image, and must be used to 

 correct it so far as is possible in the individual cases. 



The errors of the " spectrum " image are well known : 

 Prof. Abbe himself has sufficiently insisted upon them. 

 Its very contrast, or " resolution," is in most cases a 

 glaring departure from truth, to which (when we can get 

 resolution at all) the more indistinct self-luminous image 

 is in reality a far nearer approach. It tends to make 

 details which should be only geometrically symmetrical 

 to a limited extent, perfectly so. In extreme forms it 

 makes rows of spots into lines, and these lines straight 

 when not really so.. It is always liable to false resolu- 

 tions of double fineness. It fails to give even a toler- 

 able image of the larger features of the object, thereby 

 showing its failure to be a real "image" at all. All 



