of the whole world, which I shall not object to your publishing as abridged characters of the genera of the tribe Cucurbiteae of Schrader."
At the conclusion of the conspectus he continues, — " These seem to be all the genera known, that truly belong to Cucurbiteae, they have all unisexual flowers. Gronovia has them bisexual, but is otherwise very closely allied. Allacia cannot be of this order, unless we suppose the description quite erroneous ; and if so, Loureiro may have had before him, in part at least, the Telfairia pedata ; Myrianthus cannot belong to Cucurbitaceae. Thladianthus Runge is imperfectly described as to the stamens, but may perhaps form a 7th tribe.
" I have laboured under great difficulty in making out these characters, partly because the published descriptions were very imperfect, partly because I had not several of the genera, and partly, from the extreme difficulty of examining the anthers after being dried and pressed. I would therefore suggest to you and other Indian botanists, to re-examine all the Indian ones on living plants, and have drawings made, paying particular attention to the representation of the anthers.
" At first you may, as I did, confuse section 7 with section 9, but if you will compare the flower of Citrullus, Momordica or Lagenarea with Cucurbita or Coccinia, you will readily see the difference. In section 7, the connectivum is lobed, and the anther cells are placed along the edge of the lobes— in sections 8 and 9 the connectivum is not itself lobed, but the anther cell is bent," (it winds upwards and downwards along the back of the connectivum).
Warned by the concluding paragraph of the difficulty attending the description of these plants from dried specimens, I took occasion, as opportunity offered, to compare some of the sectional characters with recent specimens, and feel disposed to think the sections too numerous, and not sufficiently distinguished. I have not yet succeeded in comparing the whole, but would suggest the following alterations, which 1 think would improve the arrangement.
Section 6 might with advantage be suppressed, and its only genus referred to section 5. Sections 7 and 9 would be better united, the anthers being the same in both ; transferring, however, Cucurbita to section 8, on account of the anthers, which are similar to those of Trichosanthes, making the insertion of the filaments a matter of secondary consideration, a generic not a sectional distinction.
The difference between sections 7 and 8 would then be — not that in the former the anther is lobed and in the other entire, but that in section 7 the back is traversed by an elevated gyrose ridge, on the top of which the long gyrose anther cell is placed, while in section 8 there is no such elevation, the anther cell being sunk into the substance of the connectivum, not elevated on a ridge with a deep furrow between each bend. To this may be added that the oonnectiva in section 8 is elongated ; hence, from the union of the three, a cylinder results, while in the other their union produces a sort of capitulum.
Bryonia Garcini, doubtfully referred to Bryonia, is a new species of Pylogyne : Bryonia leiosperma, I find, from the examination of dried specimens, is a second species of Mukia, with which it agrees well in habit.
Notwithstanding these differences of opinion, it is not my intention to alter the conspectus, but print it simply as it reached me, the few additions I have to make, being included within brackets — thus [ ]. Before proceeding further it may be well to explain what is meant by the term tri-adelphous, as applied to this family, which is of such frequent occurrence in the following characters. The normal structure of Cucurbitaceae is to have five stamens, in place of which we usually find only three ; but when these are carefully examined it appears that two of them are twice the size of the third, and are actually made up of two united : each set is then called an adelphi or brotherhood, and the three together tri-adelphous. This structure is readily seen in the Pekunkie ( Cucumis acutangulus, Ainslie) where the anthers do not cohere. In those where they do cohere it is not so clearly seen, as they then require to be separated artificially before it can be made out.
In some genera the anthers are described as being one or two-celled; these characters require to be used with caution, as being generally of very difficult application in practice. Theoretically every anther is two-celled, and herein examining a number of instances with particular care, under a high magnifier, I have found most of them actually two-celled, though on less careful examination they appeared only one celled. If such is the case when examined with fresh specimens, how much more liable to error must we be when working with dried ones.
