SCHOMBURGK’S GUIANA PLANTS. 287 
The form of the seeds in Melastomacee, although it does 
not always occasion perfectly natural separations, and is in 
some few cases ambiguous, yet upon the whole, seems to 
be, as observed by Blume, the most important character that 
can be relied upon for the primary division. It may in the 
gteater number of cases be ascertained with a little pride 
in the ovules even at the time of flowering. 
Blume is of opinion that De Candolle relies too much on 
the two other characters, introduced by him in the distinction 
of the tribes; the capsular or baccate fruit, and the presence 
or absence of the setze on the ovary; the one used to sepa- 
rate Lavoisieree from Miconiee, the other to distinguish 
Osbeckiee from Rhexiew. These are certainly not characters 
of organic importance ; still in other instances in which they 
are relied upon by De Candolle, they appear to be both 
definite and constant, and therefore practically useful. The 
Sete are indeed variable in Miconiew, and so is the consist- 
ence of the fruit in Osbeckiez; but it is a distinguishing fea- 
ture in all natural classifications, that individual characters 
seldom retain the same relative value in the different divisions 
of the same group. 
The five tribes of Melastomacee may therefore be rendily 
distinguished by the following short characters, which are 
not perhaps in all cases, strictly absolute ; but the exceptions 
and ambiguous species will, it is believed, be found to > be but 
very few, 
Tribus I. Osbeckiee. Semina cochlea: Ovarium s seto- 
sum. (Bacca v. capsula.) 
Tribus II. RAeziez. Semina cochleata v. incurva. Ovari- 
um nudum. (Capsula dehiscens). 
Tribus III. Lavoisiereæ. Semina recta v. subrecta. Cap- 
sula dehiscens, : 
Tribus IV. Miconiee. Semina recta. Bacca indehiscens. 
Antherae (ut in precedentibus) 1—2-porose. 
_ Tribus V. Charianthee. Semina recta. Bacca indehiscens. 
Antheræ rima longitudinali dehiscentes. | : 
In the diga of genera I fear that too. LI reliance ze 
