473 
in the Stock {Mat thiol a incana). 
As Buchenau truly remarks, the stigmas do not appear to be tied to any 
particular morphological position, but he offers no explanation of this sur¬ 
prising fact, although in this genus they are so distant from one another that 
on question of fusion between neighbouring half-stigmas can arise. Another 
interesting fact from our present view-point is that in some species each 
placentiferous cord remains intact throughout its length, whilst in others it 
bifurcates above, a character of some importance in determining relation¬ 
ships. Both these features, viz. the variable position of the stigma and the 
behaviour of the placenta, hitherto unaccountable and apparently without 
import, now become significant and easy of explanation, for here too we 
have distinct evidence of consolidation resulting in dimorphism of the 
carpels. In such forms as R. luteola, L., R. glauca , L., R. complicata , Bory, 
R. virgata , Boiss. and Reut., the carpels are all of the valve type. Separa¬ 
tion of the conjoined edges of the carpels above is accompanied by a prepara¬ 
tory separation of the two ovuliferous strands—the double placenta 
bifurcates. In such types the stigmas occupy a position over the midribs of 
the valves. In other species in which the double placenta remains whole, 
as e. g. in R. lutea , L., R. alba , L., R. phytewna, L., the sutures present 
a double contour line—they are in fact not merely placentae but whole 
(solid) carpels and hence remain intact. In the light of this fact it is com¬ 
prehensible why in those species which have a line of hairs or protuberances 
on the midrib of the valve, a line of these structures should also be present 
on the suture (the midrib of the solid carpel) as in R. arabica , Boiss. It is 
now also clear why even the young syncarpous gynoecium in Reseda is open 
at the top. The valve carpels are joined to the intervening placentiferous 
solid carpels, but the latter do not extend to the same height as the valves ; 
consequently where the solid carpels cease, the valves are disjoined and the 
ovarian cavity is not closed in. That the same condition obtains in the 
species with the few carpels present all hollow may be due to the form of 
the valve having become fixed before the disappearance of the solid 
members. We see in the R. odorata , L., type with its stigmas on the solid 
carpels, and the R. luteola , L., type with valve stigmas, an example of the 
same interchange of functional activity as has been shown to occur in the 
other families treated above. Finally in Randonia , where we have in 
reality not G 2 but G 4 (two solid and two hollow), we find a construction 
which is the counterpart of the typical siliqua, with this difference, that here 
it is the lateral carpels which become solid and the median which remain 
hollow. The general plan is precisely the same as in the Crucifer, but in 
Randonia there is great development of the flower in the median plane on 
the posterior side, and the plane in which consolidation takes effect is 
(? consequently) shifted to that in which the lateral carpels lie. 
With the clue furnished by the dimorphic carpel yet another mystery 
is dispelled, for we are now able to interpret the extremely interesting and 
