He states that the peduncles of the cleistogamic flowers curve downwards and bury the ovaries beneath the soil. (8/8. These statements are taken from Professor Oliver's excellent article in the 'Natural History Review' July 1862 page 238. With respect to the supposed sterility of the perfect flowers in this genus see also Timbal-Lagrave 'Botanische Zeitung' 1854 page 772.) I may here add that Fritz Muller, as I hear from his brother, has found in the highlands of Southern Brazil a white-flowered species of violet which bears subterranean cleistogamic flowers.
Viola nana.
Mr. Scott sent me seeds of this Indian species from the Sikkim Terai, from which I raised many plants, and from these other seedlings during several successive generations. They produced an abundance of cleistogamic flowers during the whole of each summer, but never a perfect one. When Mr. Scott wrote to me his plants in Calcutta were behaving similarly, though his collector saw the species in flower in its native site. This case is valuable as showing that we ought not to infer, as has sometimes been done, that a species does not bear perfect flowers when growing naturally, because it produces only cleistogamic flowers under culture. The calyx of these flowers is sometimes formed of only three sepals; two being actually suppressed and not merely coherent with the others; this occurred with five out of thirty flowers which were examined for this purpose. The petals are represented by extremely minute scales. Of the stamens, two bear anthers which are in the same state as in the previous species, but, as far as I could judge, each of the two cells contained only from 20 to 25 delicate transparent pollen-grains. These emitted their tubes in the usual manner. The three other stamens bore very minute rudimentary anthers, one of which was generally larger than the other two, but none of them contained any pollen. In one instance, however, a single cell of the larger rudimentary anther included a little pollen. The style consists of a short flattened tube, somewhat expanded at its upper end, and this forms an open channel leading into the ovarium, as described under V. canina. It is slightly bent towards the two fertile anthers.
Viola Roxburghiana.
This species bore in my hothouse during two years a multitude of cleistogamic flowers, which resembled in all respects those of the last species; but no perfect ones were produced. Mr. Scott informs me that in India it bears perfect flowers only during the cold season, and that these are quite fertile. During the hot, and more especially during the rainy season, it bears an abundance of cleistogamic flowers.
Many other species, besides the five now described, produce cleistogamic flowers; this is the case, according to D. Muller, Michalet, Von Mohl, and Hermann Muller, with V. elatior, lancifolia, sylvatica, palustris, mirabilis, bicolor, ionodium, and biflora. But V. tricolor does not produce them.
Michalet asserts that V. palustris produces near Paris only perfect flowers, which are quite fertile; but that when the plant grows on mountains cleistogamic flowers are produced; and so it is with V. biflora. The same author states that he has seen in the case of V. alba flowers intermediate in structure between the perfect and cleistogamic ones. According to M. Boisduval, an Italian species, V. Ruppii, never bears in France "des fleurs bien apparentes, ce qui ne l'empeche pas de fructifier."
It is interesting to observe the gradation in the abortion of the parts in the cleistogamic flowers of the several foregoing species. It appears from the statements by D. Muller and Von Mohl that in V. mirabilis the calyx does not remain quite closed; all five stamens are provided with anthers, and some pollen-grains probably fall out of the cells on the stigma, instead of protruding their tubes whilst still enclosed, as in the other species. In V. hirta all five stamens are likewise antheriferous; the petals are not so much reduced and the pistil not so much modified as in the following species.