494 PROCEEDINGS OF THE AMERICAN ACADEMY 



portions of the same mass. The most we know, in any case, is the 

 composition of one or more fragments selected for analysis, and, as 

 we have shown, even these data are so discordant as to be wholly 

 untrustworthy. It is impossible to infer with certainty from the 

 descriptions whether the disagreement arises from faulty methods of 

 analysis, or from actual difference of composition in the fragments 

 selected, and no satisfactory conclusions can be reached in regard to 

 the influence of impurities on the crystalline structure, until an ex- 

 tended series of systematic analyses has been made of the iron mete- 

 orites, by processes which have been well considered, and the limits 

 of whose accuracy have been carefully determined. In such a chemi- 

 cal investigation, regard should be paid to the probable variation in 

 composition of different parts of the mass. This is a subject to which 

 we hope to return at a future time. 



The action of the process of crystallization in eliminating impurities 

 produces effects with many minerals not unlike those of the Widman- 

 stiittian figures. Very striking exhibitions of such effects may be 

 seen with the microscope in rock-sections containing crystals of 

 leucite, nosean, nepheline, and other minerals. Illustrations of these 

 phenomena have been given in works on lithology, and it is only 

 necessary to allude to them here, in order to make clear their analogy 

 with the phenomena we are studying. But a far more striking illus- 

 tration of this similarity is shown by a large specimen of fluorite in 

 the Harvard collection, which has been cut through a mass of com- 

 pacted crystals. The polished surface imitates very closely the 

 features of the Widmanstattian figures. There are distinct bands, 

 marked by difference of color instead of difference of lustre, separated 

 from the ground mass of the mineral by definite layers of less pure 

 material. The bands are about the width of those in the Glorietta 

 meteorite, and they all appear alike, bordered in every case by the 

 same layers of impurities arranged in a definite order, and these 

 bands, crossing at various angles, with a predominance of right angles, 

 bear a most striking resemblance to the Widmanstattian figures. 

 Such examples as these indicate clearly that the Widmanstattian 

 figures are not a peculiar phenomenon, but that such an alternation 

 of plates is often a characteristic of crystalline structure, when the 

 process of crystallization is attended by the elimination of foreign 

 ma'^erial. 



There is another feature of the Widmanstattian figures which often 

 appears, and which is best explained by the assumption that the 

 process of crystallization was extremely slow. Figure 10 shows two 



