BENNETT 



279 



species described above of the genus Viola. We have (1) dimi- 

 nution in size of the calyx, and especially of the auricles at the 

 base of the sepals ; (2) entire or complete abortion of the corolla 

 and consequently of the nectary ; (3) diminution in the number 

 of perfect stamens ; (4) elevation of the anthers on an elongated 

 filament, together with the entire suppression of the nectariferous 

 appendages to the two inferior stamens ; (5) great increase in 

 the relative size of the prolongation of the connective and loss of 

 its brilliant orange colour ; this is accompanied by a diminution 

 in size and a separation of the anther-lobes ; (6) alteration in the 

 mode of dehiscence of the anthers ; (7) diminution in number of 

 the pollen-grains ; (8) diminution in size and complete alteration 

 in texture of the pollen-grains ; (9) great decrease in length, and 

 complete alteration in shape, of the style ; (10) complete alteration 

 in form and position of the stigmatic cavity ; (11) formation of a 

 1 tube-connecteur " through the style. With regard, however, to 

 this last point, the statement that no such tube exists in the elon- 

 gated style of the normal flowers of V. canina (sylvatica) is made on 

 the authority of Mr. Darwin ('Forms of Flowers/ p. 315); but I 

 am inclined to doubt its correctness. At all events, in the few per- 

 fect flowers of V. odorata which I have had the opportunity of 

 examining in reference to this point, there clearly is such an open 

 passage (see fig. 1, h). Moreover, in our ordinary text-books the 

 style of K tricolor is described as being perforated by a similar 

 channel (see, e. g., Sachs's ' Text-book of Botany/ 3rd ed., English 

 translation, fig. 364, p. 499). Even, however, if this be omitted, 

 the number of correlated modifications is sufficiently remarkable. 

 If we look at the number of species of violet in which these flowers 

 occur, and the remarkable similarity of the ten or eleven modifi- 

 cations in them all, we are forced to the conclusion that they 

 must have originated before the differentiation of these species, 

 which are spread widely over the globe. On the other hand, 

 all the species of ffola in which cleistogamic flowers occur belong 

 to the section Nominium ; while in the section Melanium, to 

 which V. tricolor belongs, none have yet been detected; and we 

 may therefore suppose that the first production of these flowers 

 was subsequent to the differentiation of the genus into subgenera. 

 The observation of the actual impregnation in these flowers is, 

 as I have already observed, attended with the greatest difficulties. 

 But the mode of the emission of the pollen-tubes presents the 

 most striking and interesting phenomena. Any one who has made 



