15 
enter the micropyle, and subsequently the ovule Jevelopes so as to 
fill the ovarian cavity. The aborted ovule is retained at the sum- 
mit of the cavity, its funiculus never developing, and its micropyle 
out of reach of the pollen-tubes, even if its nucellar structures 
were developed to receive tliem. 
To recur now to the mature seed, it will be found to consist 
mainly of a much developed embryo-sac, filled with endosperm 
and a small embryo. The embryo-sac never entirely replaces the 
nucellus, more or less of the nucellar tissues being found on the 
commissural side, either pressed against the flat or concave face of 
the embryo-sac or embraced in its infolding. Important charac- 
ters have been obtained from the commissural face of the seed, 
based upon the fact that It may be convex, plane, concave, or miore 
or less involute. Great care should be exercised, however, to 
obtain seed of perfect maturity, or a comparison on this basis will 
amount to nothing. All seeds at first have a convex or plane face, 
and the amount of concavity or infolding will depend upon the de- 
velopment of the embryo-sac. In some cases the embryo-sac, in- 
stead of developing' uniformly, developes strongly towards the 
commissure on the two sides, resulting in a concave or involute 
seed-face. This variation may occur in the mature fruits of a single 
species (as in Eulophus)^ so that there may be found plane or con- 
cave seed-faces in one and the same plant. 
Characters used in classification. — By far the most im- 
portant characters are obtained from the fruit. The three series of 
Bentham and Hooker, based upon simple and compound umbels 
and oil-tubes, we have not found tenable. The compounding of 
umbels is too irregular in some genera to determine their proper 
position, and in the first series, with ad valleculas o,” San- 
icula and Eryngium both have oil-tubes. We have therefore \ 
divided our Umbelliferce into two series, based upon the develop- 
ment of secondary ribs or not, which grouping practically merges 
Bentham. and Hooker’s first and secotid series. Our second series, 
which is by far the larger, vve have broken into groups based upon 
the flattening of the fruit, which seems to be a veiy reliable char- 
acter, and one which best groups together related genera. Char- 
acters which are used in further subdivisions, enumerated in the 
Older of their general importance, are as follows: flattening of the 
/ 
