RECURRENT SELECTION AND OVERDOMINANCE 461 



bp on P has been calculated. The second order regression function has been 

 solved for the special case bp = 0, to obtain an estimate of Fc the critical 

 value of P where the regression surface is level and heritability is zero. 

 The third summary in Table 28.1 is for average yields in six states of the 



TABLE 28.1 



REGRESSION OF YIELD OF Fi AND F. CORN HYBRIDS ON YIELD OF 

 INBRED PARENTS WITHIN GROUPS HAVING 

 ONE COMMON PARENT 



Yield of parents (P) is recorded in bushels per acre, with the partial regression coefficient 

 (bp) below each one for the group of which it is the tester. The second order regression 62 is 

 regression of bp on F. Critical P (Fc) is estimated value of F for bp = 0. 



Stringfield, G. 11. Unpublished. Ohio Agr. Exp. Sla. and USD A* 



F 14 28 30 46 51 55 



bp(FO .68 .41 .31 .22 .07 .05 



bp{F.;) .55 .45 .33 .24 .26 17 



Mean bp{Fi) .29, (F-,) .33; 6.,(F,) -.014, (Fo) -.008; Mean P 37; P,. 58; Mean F, 97; 



Mean F2 70. 



KinmuH and Spragiie, Agron. Jour. 1Q45* 



F 3 15 20 26 28 28 32 39 40 50 



6p(Fi) .63 .75 .84 .69 .13 .30 .25 .39 .22 .01 



6p(F.,) .26 .36 .42 .69 .24 .29 .37 .58 .54 .47 



Mean ip(F,) .42, (F.) .42; ^,(Fi) -.016, (F2) +.005; Mean F 29; P. 54; Mean Fi 80; 



Mean F. 51. 



USD A and State Regional Tests, Midseason 1943; Iowa, Kans., III., Ind., Ohio, Fenn. F values 

 from Kinman and Sprague above; their Fi's included here* 



bp{Fi) -.05+.11+.08-.13-.20-.11 + .12-.01-.18 

 Mean bp -.01; 62 -.004; Fc 25. 



* Sources of data. 



same Fi's as those of Kinman and Sprague in Iowa. The Iowa test included 

 parent lines and F2's as well as Fi's. The third summary has been made with 

 Iowa records on parent lines. An analysis was made also of the Fi records for 

 each state separately with the same values of P. Regression trend was posi- 

 tive for the Indiana data, thus failing to support any theory of dominance of 

 high yield. Regression trends for the other four states were negative with esti- 

 mates of Pc all lower than the one for Iowa. 



The eighteen other sets of data not summarized in the table are from Min- 

 nesota, Iowa, Illinois, Ohio, New York, and North Carolina. They are be- 

 lieved to be generally independent genetically and ecologically. Regression 

 trends are positive in eight cases. Taking the five cases summarized together 

 in Table 28.1, as five separate ones, we have seventeen with negative regres- 

 sion trend to eight with positive. Estimates of Pc for the seventeen negative 

 trends are near to or within the range of data as in Table 28.1 for each case 

 but one. With one of the least extensive tests the estimate of Pc is roughly 

 12 times the top inbred line, thus agreeing nicely with incomplete dominance. 



Insofar as regression trends are due to heterozygosity they may be expect- 



