In some double flowers, in which the stamens assume more or less completely the appearance of petals, a singular appearance is afforded by the presence of four wing-like processes emanating from the central filaments, two on each side, so that the arrangement may be compared to two sheets of paper folded in the centre and adherent in that situation, though perfectly separate elsewhere, except sometimes at the top, where they form a sort of hood. This change results from an imperfect petalody of the anther; the two wings on each side of the central vascular cord represent the front and back walls of an anther lobe, or rather of that portion of the anther which, under ordinary circumstances, produces pollen. In the malformed flowers no pollen is formed, at least in the more complete states of the malformation, but the walls of the anther lobe become preternaturally enlarged, and petaloid in texture and appearance. This change occurs in some semi-double rhododendrons and azaleas, in crocuses, and in a species of violet found at Mentone by Mr. J. T. Moggridge. There are numerous intermediate forms wherein the wing-like processes may be traced all the way along the filament till they ultimately lose themselves in the anther-lobes, with which they become continuous. In some cases, as in _Crocus_ and _Rhododendron_, this is shown even more clearly by the existence of two perfect pollen-sacs or quarter-anthers, the remaining portions being petaloid and continuous with the dilated filament. Not unfrequently these semi-petaloid stamens adhere to the fronts of the petals, and then it appears, at a first glance, as if three organs were stuck together, one in front of another, while in reality there are but two.[305] (See _ante_, p. 35, fig. 12.) [Illustration: FIG. 155.--Four-winged filaments of _Rhododendron_.] The change in the anther, above alluded to, must not be mistaken for that far more common one in which only a small portion of the anther becomes petaloid, forming a sort of lateral wing or appendage to the polliniferous portion, as happens normally in _Pterandra_, and is common in some double fuchsias. In this latter instance there is but a single wing, and the nature of the case is obvious.