1924] 
HARRIS—THE INFLORESCENCE OF MANFREDA VIRGINICA 441 
the number of flowers and fruits which it produces and the fertility 
of its fruits. 
This seems a question of very real physiological interest On 
a priori grounds one might be inclined to suggest that the size of 
the inflorescence is а measure of vigor, and that, as another ex- 
pression of the greater vigor, larger numbers of seeds would be 
expected to be associated with larger numbers of flowers per in- 
florescence. Оп the other hand, it may be urged that since the 
inflorescences with larger numbers of flowers also produce larger 
numbers of fruits, the demands for plastie materials due to greater 
numbers of fruits would result in а reduction rather than in an 
increase in the number of seeds per locule. 
The product moments for the relationship between the number 
of flowers per inflorescence and numbers of seeds per fruit may be 
calculated from table 1.! 
Table хуш shows the correlation between the number of 
flowers per inflorescence and the number of seeds per locule and 
between the number of fruits per inflorescence and the number of 
seeds per locule in the 4 series in which the numbers of seeds were 
determined. 
Three of the constants measuring the relationship between the 
number of flowers and the number of seeds are positive, while one 
is negative in sign. АП are small, however, ranging from — 0.020 
to + 0.095. In general the coefficients are not as large as their 
probable errors.? 
Correlations between the number of fruits per inflorescence and 
. Ж. ға А ті har 
= 2.4 oi 
! Note that in det i th t 
of seeds the means and standard deviations for number of flowers per infloresce 
must be redetermined by weighting with the number of fruits or number of seed 
per inflorescence. The constants thus weighted may be used in the determination 
of correlations for number of flowers and number of seeds per locule or seeds per 
fruit, since all of the fruits are triloc 
? The question of the number to be used 0 caleulating the probable error of these 
constants has presented considerable difficulty. Тһе number of fruits in which 
the seeds were counted has been very large. If this N were used in the determination 
of the probable error it would be very small indeed. It шау be questioned, “sid 
whether the probable error of the correlation between a weighted variable z and 
another variable y is any lower than that obtained when the unweighted е of 
the 2 characters is used. We have, therefore, in determining these probable errors 
taken N to be the actual number of inflorescences. 
