E. R. Saundkrs 315 



proportion of single to double, but both plants were presumably 

 producing an excess of doubles as was also apparently K, probably J, 

 and certainly I ; L on the other hand yielded a proportion of about 

 3 s.: 1 d., i.e., the proportion we should expect from a cross-bred rather 

 than a pure-bred. In the next generation a single descendant from 

 each of the two plants / and / was selfed, and both like their parents 

 gave doubles in excess ; both in short behaved like eversporting indi- 

 viduals as we should naturally expect. In the case of K and L however 

 the results obtained in F^ are not so easily comprehended, for in neither 

 case did all the Fj singles yield doubles in F^. In fact the same 

 diversity of behaviour exhibited by the haphazard collection of singles 

 (plants C — if) is here found among the sister plants of a self-bred 

 family derived from one of these singles (K). 48 ^i descendants of K 

 were tested, 1 (plant U) by cross-fertilisation only, 47 by self-fertili.sation 

 either alone or in addition to cross-fertilisation. [Where self-fertilisation 

 shows that an individual was throwing doubles it is unnecessary for the 

 present purpose to complicate the pedigree further by introducing into 

 it the results of cross-fertilisation, and these results have therefore been 

 omitted where the evidence from self-fertilisation was sufficient.] The 

 former plant (U) and 40 of the latter again produced doubles in the 

 next generation, but the remaining 7 yielded only singles, the numbei-s 

 in these 7 families ranging from 8 to 68. How many among the 

 40 mixed families can be regarded as showing the true proportion of 

 singles and doubles is uncertain, since in many the totals are very 

 small ; moreover the seed was not sown until two years after it was 

 harvested, and in some cases germinated badly. (See later, p. 361, where 

 the probability that seeds giving rise to singles and doubles respectively 

 diflfer in viability is discussed.) 



To sum up the foregoing results : 



Plant iT as a matter of fact gave a very slight excess of doubles, but 

 among the Fi singles derived from K some were evidently giving doubles 

 in the proportion of only 1 d. : 3 s. while others were apparently breeding 

 true to singleness. Some of the F^ singles similarly yielded the pro- 

 portion 1 d. : 3 s. in F,. (See Table II.) 



In the case of plant L, 31 Fi descendants were tested by self- 

 fertilisation ; 18 of the resulting jPj families included some doubles, 

 13 were composed entirely of singles, the numbers in the latter class of 

 families ranging from 7 to 34. If we review these 18 families we find 

 that in 13 the numbers agree well with the ratio 3 s. : 1 d., and that in 

 the remaining 5, none of which included more than 6 individuals, there 



