494 



SCIENCE 



[N. S. Vol. XXXVII, No. 952 



ham' has arranged the data of the trials of 

 peas made at Wisley in 1911' in a convenient 

 form for statistical analysis. 



Comparison with the wheat data already 

 discussed is interesting. In wheat, selection 

 has been primarily in the same direction — 

 largeness of yield in bushels per acre — for all 

 varieties. In peas, it has been in diverse di- 

 rections. For the character here dealt with — 

 time required for development — the tendency 

 has been, consciously or unconsciously, to dif- 

 ferentiate widely the varieties. Hence in the 

 pea data, it is idle to lump all the materials to- 

 gether, for any constant thus obtained would 

 be largely spurious and insignificant.* In 

 practical tests one must decide between a 

 series of closely similar strains — not between 

 those which are widely and obviously differen- 

 tiated ; hence, we split the material up into the 

 four classes recognized by Durham, that is, 

 into 81-90, 91-100, 101-110, 111-120 day peas 

 and calculate the inter-period correlation" for 

 each class separately for the only economically 

 important characte/ for which data are avail- 



^ Durham, F. E., "An Analysis of the Pea 

 Trials at Wisley, 191]," Journ. Eoy. Hort. Soc, 

 38: 67-72, 1912. 



2"Pea Trial at Wisley, 1911," Journ. Soy. 

 Hort. Soc., 37: 403-424, 1911. The two original 

 papers must be consulted for the details that are 

 here excluded for lack of space. 



^ As a matter of fact these constants have been 

 calculated for another purpose, but they need not 

 be published here. 



" To be of value in determining the relative 

 merit of varieties the performances of a given 

 strain in a test should be a good basis for predic- 

 tion as to the results of a subsequent experiment. 

 To what extent this is true may be determined for 

 any two or more series of trials of a number of 

 varieties by determining the coefficient of correla- 

 tion between their performances, correction being 

 made when symmetrical tables are involved — i. e., 

 when any determination is used both as a first and 

 as a second member of a pair — for environmental 

 heterogeneity from experiment to experiment. 

 This does not apply to the pea data. 



' The data for the individual growth periods 

 give, on the basis of the total data, such irregular 

 correlations that it is not worth while to consider 

 them for the subclasses. 



able — total days required for the formation of 

 usable pods. Hence, designating by the sub- 

 scripts 1, 2, 3, the three successive cultures of 

 1909' we have the following relationships : 

 For 91-100 day peas, m = 30, 



r^ = .16± .12, r,3 = .19± .12, ra = .12± .12 

 For 101-110 day peas, mr=40, 



r^ = .43 ± .09, r^3 = .37 ± .09, r^ = .78 ± .04 

 For 111-120 day peas, n = 18, 



r^ = .27 ± .15, r^ = .50 ± .12, r,, = .45 ± .13 



All the values are positive. Their wide 

 fluctuation and the magnitude of the prob- 

 able errors is probably largely attributable to 

 the necessary smallness of the number of va- 

 rieties in each class. 



It is clear that a single test when carried 

 out in the manner of those of the Royal Horti- 

 cultural Society has little decisive value con- 

 cerning the merit of a variety. This is not 

 intended as a criticism of these tests, for they 

 are in comparison with many others appar- 

 ently of a very high order of merit. But cer- 

 tainly they lend their emphasis to the point' 

 made in the preceding paper. 



Is it not time for a concerted and systematic 

 effort on the part of those interested in agricul- 

 tural science to put this important problem on a 

 sound basis, biologically and statistically? 



J. Arthur Harris 



Cold Spring Hakbok, N. Y., 

 December 21, 1912 



ON THE metamorphosis OP AN AMCEBA, VAHL- 



KAMPFIA SP., INTO FLAGELLATES AND 



VICE versa' 



An amceba of the limax group isolated, in 

 1909, from tap-water in Oakland, California, 



' Had the cultures been made at the same season 

 in three succeeding years, the test would have 

 furnished data of more value to the practical 

 grower. Data for such tests are, as far as I am 

 aware, not available. 



° This has already been emphasized on general 

 grounds by various students of agronomy. See 

 especially C. V. Piper and W. H. Stevenson, 

 ' ' Standardization of Field Experimental Methods 

 in Agronomy," Proc. Amer. Soc. Agron., 2: 70- 

 76, 1910. 



^ Presented to the Cincinnati Eesearch Society, 

 January 9, 1913. 



