192 H. S. H. WARDLAW. 



These figures show also that the mean values of the 

 results do not differ much, as a rule, from the most probable 

 values. This is due to the fact that the results lying out- 

 side the range of most probable values are fairly evenly 

 distributed on either side, and so balance when the arith- 

 metic mean of the whole is taken. The arithmetic mean 

 of series of analyses such as these, although of itself it has 

 no meaning, thus lies close to a value which has a definite 

 significance, and the mean value may, therefore, be used 

 in many cases instead of the better defined, but not so 

 readily accessible, most probable value. It should be 

 remembered, however, that the occurrence of a number of 

 very high or low values in a series of results would con- 

 siderably alter the mean of the series, but not the most 

 probable value. 



5. Variation of composition with, time since parturition. 

 The composition of a sample of human milk is known to 

 depend on the stage of lactation of the woman from whom 

 it is obtained (Schloss, 1912 ; Oamerer and Soldner, Engel, 

 Adriance and Adriance, loc. cit.). The main features of 

 this dependence are that the ash and protein of the milk 

 tend to decrease as the lactation continues, while the 

 amount of milk sugar slightly increases. These statements 

 apply to the very gradual alteration in the composition 

 of human milk during the course of months of lactation. 

 During the first few days of lactation, however, the alter- 

 ation in the composition must be very much more rapid, 

 since the milk first secreted after the birth of the child, the 

 colostrum, has a composition which differs considerably 

 from that of the milk secreted later, the principal differ- 

 ence being the high percentage of protein in the former. 

 Sample No. 29, Table II, is a typical example of colostrum. 

 The variation of the composition of human milk obtained 

 in the first few days after the colostral period, which 



