1953] 



REVISION OF BRACHYOTUM 



347 



of Chaetolepis , with the hypanthium usually not exceeding 3 nun. and the petals 8 

 ram. in length, and the peculiar pubescence of the calycine sinuses of Pterolepis 

 may then serve as further distinctions. The characters noted in Cogniaux's key 

 serve to eliminate all other genera of this group except Svitramia and Tibouchina, 

 In Svitramia and almost all species of Tibouchina examined in the herbarium of 

 the New York Botanical Garden or the descriptions studied, the connective is 

 ''prolonged" below the anther-thecae to the point of filament insertion; the ventral 

 appendages of Tibouchina are thus neither in contact nor fused even partially 

 with the thecae; the prolongation of the exappendiculate connective of Svitramia 

 is obvious because of the "break" at the point of articulation of the filament. All 

 species of Brachyotum with large persistent bracts closely investing the hypan- 

 thium would immediately be unsuitable in Cogniaux's key to the sections of Ti- 

 bouchina because of habit or lack of lepidote pubescence. The 4-merous species 

 of Tibouchina in sect. IX (Pseudopterolepis) all have connectives definitely (al- 

 beit often shortly) prolonged below the anther-thecae. In the 5-nierous species of 

 sect. VIII (Diotanthera) y unequal stamens and/or connective prolongation exclude 

 all species except T. mollis (Bonpl.) Cogn. from Brachyotum. The petals in this 

 species are obviously spreading; the connective however, even in well-developed 

 anthers, may be very slightly prolonged below the thecae to the filament articula- 

 tion or, more frequently, only very shortly prolonged ventrally immediately below 

 and adherent to the thecae. T. mollis is a close approximation of an ancestral 

 Brachyotum. 



Ule (1895) distinguished between the subulate-tipped anthers of Tibouchina 

 and Brachyotum and the truncate-tipped large-pored anthers of Purpurella. This 

 distinction, however, is not valid for B. lycopodioides and its relatives. Ule like- 

 wise noted the campanulate "corolla-tube" of several Brazilian species of Pur- 

 purella An all of these species of Purpurella however, the connective is prolonged 

 below the thecae and points downward, rather than curving along the thecae bases 

 as in Brachyotum. 



Cogniaux's sectional criteria and key characters within his sections did not 

 permit any natural grouping of the species of Brachyotum. Sections based upon 

 the presence [sect. I. Dicentrae (Naud.) Cogn.l or absence [sect. II. Adesmiae 

 (Naud.) Cogn.l of ventral prolongation of the connective at the anther base are ap- 

 parently artificial; indeed, this distinction is often of dubious value even in spe- 

 cific delimitation. For the species presently known, the only natural subgeneric 

 delimitation would appear to be the separation of the wide-pored species, but for- 

 mal change has here been deferred. The species with large persistent bracts 

 closely investing the hypanthium do not seem to form any natural unit. Formal 

 specific changes, such as relegation to subspecific categories, have also been 

 avoided until additional collections establish the best direction and degree of 

 such dispositions (i.e., B. lindenii and B„ alpinum and/or including B. cernuum; 

 B. figueroae , B. rostratum t and B. lutescens; the possible varieties of B. micro- 

 don; and the possible varieties of B. sanguinolentum). The representation of sup- 

 posed relationships between the species of Brachyotum (Fig. 24) has been limited 

 to a drawing portraying only two instead of three dimensions, despite obvious 

 inadequacies. 



GENERAL MORPHOLOGY 



Pubescence. The trichomes vary from perfectly smooth to remarkably shaggy- 

 plumulose or nearly stellate. In those species having hairs roughened, the degree 

 of development of the projection of the surface cells of the hairs varies with the 

 location on the plant; the cauline, lower leaf surface, and hypanthial hairs are the 



