OCTOBEB 8, 1915] 



SCIENCE 



503 



are somewhat reversed, the evergreens being 

 mostly in places unattractive to farmers, and 

 the apparent percentages of them probably too 

 high. The percentage of improved land in 

 West Virginia and Arkansas is rather low, 

 perhaps chiefly because these states are off the 

 main routes of travel and have not received 

 as many settlers as their soil would warrant. 

 It is low in Mississippi and Louisiana on ac- 

 count of large areas of alluvial land, which al- 

 though very rich (and originally wooded al- 

 most exclusively with deciduous trees) were 

 very little cultivated in 1880 on account of 

 being subject to overflow. In the northern 

 states improved land includes a large propor- 

 tion of pasture, on which no fertilizer is used, 

 and if the amount of cultivated land could be 

 substituted for improved land the fertilizer 

 figures for these states would average consid- 

 erably more per acre. 



Finally, it can not be doubted that different 

 chemical elements in the soil affect evergreen 

 percentages and other features of vegetation 

 unequally, and it is well known that the com- 

 position of rich soils varies greatly in differ- 

 ent states. The soils of Florida are generally 

 well supplied with calcium and phosphorus, 

 but deficient in potassium, while in Illinois 

 phosphorus is said to be the element most in 

 danger of exhaustion. The average composi- 

 tion of fertilizers used varies from state to 

 state, corresponding more or less with the soils 

 (a larger proportion of potassium is used in 

 Florida than in any other state), but no sta- 

 tistics of fertilizer ingredients are given in 

 census reports; so that matter will not be 

 taken up at the present time. 



Roland M. Haeper 



SPECIAL ARTICLES 

 STANDARD DAIRY SCORE CARDS 



Mr. James D. Drew^ presents data which 



1 Drew, J. D., ' ' Milk Quality as Determined by 

 Present Dairy Score Cards, ' ' Bull. N. Y. Agr. Exp. 

 Sta., 398, Geneva, March, 1915. The work was 

 originally planned by Dr. H. A. Harding, and is 

 now being carried out in cooperation with the 

 Illinois Agricultural Experiment Station. 



should be of very general interest. The pur- 

 pose of this note is to direct attention to these 

 studies in the hope that they may receive the 

 wide attention which they deserve. The most 

 important result of such consideration would 

 be the carrying out of even more comprehen- 

 sive and exact studies of the same kind. 



The problems taken up are essentially two: 



a. What is the correlation between the 

 grades assigned a series of dairies by the same 

 inspectors when different score cards are used ? 



h. What is the correlation between the score 

 assigned a dairy by an inspector and the qual- 

 ity of the milk which it places upon the 

 market ? 



The first of these problems is of technical 

 importance in determining the degree of re- 

 liability of the application of score cards to 

 the grading of dairies. The second is funda- 

 mental to the determination of the utility of 

 the score card in the grading of milk, and thus 

 one of first rate practical significance to the 

 consumer of dairy products. 



For purposes of review it has seemed best 

 to express the detailed observations in the suc- 

 cinct terms of statistical constants. The per- 

 sonal opinion of the reviewer that such sta- 

 tistical constants are better as a means of ex- 

 pressing the results than the mere comparison 

 of individual points of detail, and his convic- 

 tion that the analysis of the data in certain of 

 its more refined essentials can be carried out 

 only by such formulae, must not be taken as a 

 criticism of the data or their discussion in the 

 paper under review. 



With regard to the agreement between the 

 three methods of grading, the authors confine 

 their discussion to the relative positions of the 

 individual dairies on the three score cards. 

 The correlation coefficients^ are illuminating. 

 They are : 



2 Data for 34 barns are given. For 23 of these 

 bacterial counts for morning milk as well as even- 

 ing milk are available. As a precaution against 

 arithmetical slips I have marked out the correla- 

 tions between the results by the three cards for 

 both the total series and the sub-series for which 

 morning milk was available. The latter should 

 be essentially a random sample of the former. 



