considered in Relation to the Phyllode Theory. 325 
that of Crocus. So it may be well to begin our consideration of the 
Crocoideae by trying to make out the nature of the leaf of Romulea , and then 
to consider the relation borne to it by the highly peculiar leaf of Crocus. 
The leaf structure of Romidea is illustrated in Figs. 61-4. R. Bulbo - 
codium, Sebast. et Mauri, is a typical example. Here there is a sheath, 
closed at the base (Fig. 61 a), whose most striking feature is the relative 
unimportance of the median bundle \m.b.) as compared with the main 
laterals (m.l). Higher up, the sheath opens out (Fig. 61 B) and the passage 
into the limb (Figs. 61 c and D) shows that the leaf is essentially of the 
ensiform type, not unlike that of Gladiolus ornatus (Fig. 45 B, p. 318), but 
modified by the presence of two pairs of grooves or invaginations, one pair 
{g. 2 ) lying between the median bundle and the main laterals, and the other 
pair (gf between the main laterals and the ventral bundle formed by the 
fusion of the external lateral bundles (e.l.). Romulea Colmnnae , Sebast. et 
Mauri (Fig. 62), and R. rosea , Eckl. (Fig. 63), do not differ essentially from 
the type of R. Bulbocodium. One member of the genus, however, R. crocifolia , 
Vis. (Figs. 64 A and B), diverges from the R. Bulbocodium type in being dorsi- 
ventral to the extreme apex of the leaf. 1 I interpret the leaf of this species 
as a reduced form, equivalent merely to the sheathing leaf-base of the 
other members of the genus, the petiolar limb having been lost. R. crocifolia 
thus bears the same relation to R. Bulbocodium that the Juno Irises bear to 
the Irises with ensiform leaves. 
The leaf of Crocus is described by Ross 2 as belonging to the dorsi- 
ventral type, and is compared by him to the aberrant Romulea crocifolia . 
Celakovsky 3 and Velenovsky 4 take the same view, while Chodat and 
Balicka-Iwanowska 5 expressly state that the leaf of Crocus is in no way 
comparable with that of a normal Romidea , and Balicka-Iwanowska, 6 in her 
later paper, describes it as ‘ absolutely isolated among the Irideae ’. My 
study of the anatomy and mode of origin has, however, led me to a con- 
clusion wholly different from that of these five writers ; I hope to show that 
the limb of the Crocus leaf is a petiolar phyllode, strictly homologous with the 
leaf of such a Romulea as R. Bulbocodium, though it is true that transverse 
sections of the mature leaves of the two genera (cf. Figs. 56 1 and 6 id) 
would scarcely suggest the possibility of any relationship between them. 
We may take Crocus Tomasinianus , Herb., as a case for description. 
Figs. 56 A-D represent sections from a microtome series through the base 
1 Ross, H. (1892-3), draws attention to the dorsi ventral character of R. crocifolia, Vis. Ac- 
cording to Beguinot, A. (1907-9), this specific name is merely a synonym of R. nivalis , Klatt, but 
judging from a leaf from Visiani’s type specimen at Padua, generously sent to me by Professor 
Beguinot, R. crocifolia is more completely dorsiventral in structure than R. nivalis , and the leaf 
anatomy certainly suggests that a separate specific name might usefully be retained. 
2 Ross, H. (1892-3). 3 Celakovsky, L. J. (1903). 
4 Velenovsky, J. (1907). 5 Chodat, R., and Balicka-Iwanowska, G. (1892V 
Balicka-Iwanowska, G. (1892-3). 
