270 II. II Robinson — Chemical Analyses of Igneous Hocks. 



(3) would seein to be that an analyst naturally works toward 

 the general average rather than away from it. 



Distribution curves with a single maximum, as illustrated at 

 their best in fig. 3, are the common type. A less common 

 type is a double distribution curve — one with two maxima. 

 Triple distribution curves, due to pure chance, probably do not 

 occur. 



Several examples of double distribution curves are shown in 

 fig. 4. There is no doubt as to the reality of such distribu- 

 tions, the data are ample. In fact, the best curve of the four 

 (D) rests on the smallest number of analyses. Just what con- 

 ditions influence an analyst to get summations thus distributed 



Fig. 4. 



72 <}$37 9962 *?*?■& 7 WO./ 2 /0037 /OOiZ /0087 /O/ Z2. /0/.37 JO/ 62. /0/S7 



is difficult to say. They may be entirely personal, and the 

 problem is then one for the psychologist, or they may result 

 from such causes as changes in analytic methods, technic or 

 quality of reagents, or to differences in the composition of the 

 rocks analyzed. 



Referring to fig. 4, it is seen that analyst D has larger pro- 

 portions of summations in the two groups lying either side of 

 the common maximum group than in that group itself. His 

 tendency, also, is to get more summations in the lower group 

 (99-75-99-99) than in the higher (100-25-100-49). The dis- 

 tribution of the summations under the circumstances, as shown 

 by the form of the curve, is excellent. It must be concluded 

 that such results are not in any sense haphazard. The ten- 

 dency of analyst E toward double maxima is less marked than 

 that of D. This analyst's upper maximum is the group 

 100-50-100-74 per cent, 'his lower is strictly in 99-75-99*99 per 

 cent, but practically covers that group and the next lower. 

 His maxima are thus farther apart than those of D and his 



