9214 MR. A. 8. HORNE ON THE STRUCTURE AND 
ovarian rudiments. The central group of stamens, together with the ovarian rudiments, 
represents the median flower with which the lateral, partially aborted flowers have been 
united so as to form one floral organ. In Corylopsis spicata all stages may be found 
in the replacement of petals by stamens, so that a ring of twice the normal number of 
stamens may surround the ovary. A process of the latter kind would explain both 
the origin of the ring of stamens and the absence of a perianth in Davidia. 
It might be argued that some or all of the stamens are really toral stamens which 
have been caught up in the growth changes inducing epigyny through successive 
generations in just such a way as the epigynous tendency asserts itself in the genus 
Garrya, with the result that bracts are inserted superiorly on the ovary wall. There is 
no direct evidence in favour of this view. At the same time the irregularities in the 
number of stamens in the staminal zone come within the limits of reasonable variation. 
In such forms as Cornus mas and C, sanguinea there is extremely little variation in the 
flower. In Sambucus nigra sepals vary in number independently of the number of 
the loculi of the ovary. The proportion of normal to exceptional flowers in these respects 
is 77 out of 316 examined. In Hedera Helix var. arborea the stamens vary independently 
of the loculi. Only 18 flowers were observed, and in these the loculi varied from 4 to 7 
and the stamens from 5 to 9 (Pl. 32. fig. 12). The table for Davidia, which, it must 
be remembered, refers to a very limited quantity of material, resembles that of Hedera 
Helix var. arborea in the degree of variation. The most frequent correlation is that 
of seven loculi with twenty-one stamens (Pl. 32. fig. 10). The tables also indicate 
a tendency towards two means, both multiples of five. The number of stamens in 
the epigynous zone of Davidia has been described as five by Hemsley *, while the 
loculi are stated to vary from 5-10. If, as Wilson states, there be more thanone 
form of Davidia, the range of variation may be considerably greater than that shown 
by the table. 
The idea of a serial arrangement of stamens suggested by the correlation table is 
strongly reinforced by the anatomical evidence, which points to the existence of à 
definite whorl of oppositilocular stamens, the peculiar orientation itself suggesting an 
arrangement of stamens in series. The stamens appear to vary independently of the 
number of loculi. In a few cases they are three times as many as the mean number or 
loculi. Greater stress should be laid upon this correlation than it apparently deserves, 
since variation itself would account for considerable irregularity in the correlation 
table. 
Taking into account flowers of the whole inflorescence, the following points are 
significant :—(1) The existence of a whorl of oppositilocular stamens, which points to an 
arrangement of stamens in series.  ( 2) The number of loculi, which is generally six 0r 
seven, seven being the mean. (3) The emergence of the toral stamens in groups 
principally of fives and sixes. From this point of view, the hermaphrodite flower loses 
its anomalous aspect and becomes more definitely related to the toral stamens, which, a$ 
I have shown, emerge in groups in numbers whose mean is just below the mean number 
of ovarian loculi. 
* Hemsley in Journ. Linn. Soc., Bot. xxxv. (1903) p. 556. 
