appear to be normally self-fertilizing. Since their flowers have no nectaries for attracting insects suitable for effecting fertilization, we can not say that the presence of ants on the surface of the plant tends to drive away apterous insects that might appropriate the nectar without profit to the plant.

Perhaps the ants might in certain cases assist in transporting the seeds of the plant, which are covered, like those of the mistletoe, with a viscous matter; but such dissemination seems to be effected more commonly by carpophagous birds, who carry them from one tree to another, or by the rain, that washes them from the upper to the lower branches of the same tree. There seems to be no doubt but that the Myrmecodias and the Hydnophtums are derived from Rubiacea that were primitively terrestrial or but feebly epiphytic. Their affinities with Uragoga are very strong. Usually epiphytic plants need for their development a small quantity of vegetable detritus in which their seed can be sheltered while germinating. Normal epiphytes are not provided, as are these Rubiaceae, with fruits having viscous pulp that causes their seed to adhere to the surface of the bark upon which they fall, and would be unable to gain a lodgment on such a surface.

These Rubiacea seem, in fact, to be intermediate forms between the normal epiphytes and the parasites, such as the Loranthaceæ, which are likewise provided with viscous fruits (mistletoe) whose dissemination is effected by fruit-eating birds. The seeds of these Rubiace are, at the time of their germination, peculiarly situated. Subject to desiccation, which is very likely to occur upon the surface of the bark, they can not borrow from the tree on which they rest the moisture necessary for their life, as do the plantlets of parasites. They must, therefore, create for themselves a store of water. This is done by the thickening of their hypocotylous axis, which enlarges into a tubercle that acts as a reservoir.

Beccari supposes that the formation of the flocculent tissue of the tubercle is a consequence of this mode of development combined with alternations of dryness and moisture. But this tissue develops from an internal generating layer, and, since it is composed of dead and dried cells, it seems more logical to suppose that it results from the starvation of such cells because of the formation about them of a corky layer that deprives them of all nutritive material and vascular connection. This would be an example of true parasitism of one tissue in relation to another, the generating layer acting like a parasite as regards the central parenchymatous layer.

Since Treub has not followed the complete evolution of a Myrmecodia from its germination up to its adult state, we may admit the opinion of Beccari until a formal demonstration of its error shall be furnished. According to this author, though the ants may not at the present time be necessary for the formation of the bulbiform enlargement of the hypocotylous axis, they are required for its future growth. To state

this in another way: The plantlets of Myrmecodia, without the help of ants, might, indeed, by reason of their hereditary tendencies, commence to form the tubercle but would be unable to develop to adult dimensions.

The intervention of ants must, then, be considered as indispensable to the life of the plant, since they contribute to the development of the organ that serves as a water reservoir. Were there no ants there would be no internal reserve of water, and the plant would be exposed to all the dangers of drought. It may be remarked here that, according to recent researches, a similar service is rendered to plants that grow in the sand of Sahara by nematode worms, that act on their subterranean organs. These worms (Eterodera, noted for their ravages on certain garden vegetables, particularly the betterave) cause a development of histological elements adapted to serve as water reservoirs.

The biologic relations of the ants with the Myrmecodias seem, then, to be symbiotic. The symbiosis is not, perhaps, as close as some think, but it seems difficult to deny its existence. There seems to be good reason to suppose that if, during several generations, the ants should cease to visit the tubercles, those bodies would undergo a progressive atrophy, or at least be reduced to the state of solid tubercles without internal cavities, such as those of Pentapterygium (Vaccinium) serpens, for example.

The lodging organs of several myrmecophilous orchids have a great resemblance to those of the Rubiacea we have just been studying. We know among the orchids three examples of which there can be but little doubt. One of these has been known for quite a long time, having been already mentioned by Rumphius. It is that of an epiphytous orchid, Grammatophyllum speciosum, whose pseudobulb thickens, even after the fall of the first leaves, and within whose fibrous mass ants establish themselves.

The Lecanopteris deparioides (a fern) has a rhizome similar to the tubercles of Myrmecodia and Hydnophytum, and which, like them, forms a true ants' nest. Within this rhizome there are hollowed-out cavities and galleries that are at first filled with a flocculent matter, analogous, doubtless, to that of the Rubiaceæ above cited. The ants penetrate the interior of this rhizome by an opening situated on the upper projecting part, upon which the fronds are inserted. The same arrangement is found in P. sinuosum, on the surface of whose rhizome are found circular openings, indeterminate as to situation, that seem undoubtedly to be the work of an ant (Iridomyrmex cordata), the same that inhabits Hydnophytum petiolatum.

Certain Melastomacea are likewise epiphytic and myrmecophilous. Such are the Pachycentrias, epiphytic or pseudo parasitic plants, whose branches, interlacing on the surface of tree trunks, give out a great number of adventitious roots. Upon these roots enlargements are found irregularly spherical in shape; and if, as frequently happens, sev

eral such enlargements are contiguous on the same root, a chaplet is thus formed with more or less regular beads.

It is admitted that these tubercular roots may offer an asylum for ants. But the study of these tropical plants is as yet very incomplete. We can not even affirm that the tubercles are hollow. Many appear, when dry, to be full of spongy tissue, loosely arranged toward the center. We might be inclined to believe, with Beccari, that later this tissue is destroyed by ants, who thus hollow out a regular cavity within these tuberculous roots. But it should be noted that certain species of this genus have tubercular moniliform roots that are entirely solid, and only in certain specimens, even in species with hollow tubercles, do we find perforations allowing a communication between the inside and the exterior, which might, indeed, have easily been the work of ants.

The Pachycentrias do not seem to be provided with extra floral nectaries capable of attracting auts. The insects are, then, attracted to the plants only by the chance that they may be able to install themselves in the tubercle. But it should be noted that a type closely related to the Pachycentrias, Pogonanthera robusta, has a limbus prolonged at its base into two auricles, decurrent upon the petiole, that appear to be nectariferous. These nectaries attract ants, but the roots of these epiphytic plants are not, like those of the Pachycentrias, tuberculiform, but normal and incapable of affording lodgment to the ants attracted by the nectaries. This fact may, perhaps, give us a clew to the way in which biological relations were first established between the ants and the Pachycentrias. The ancestors of these latter plants were, without doubt, like the Pogonantheras, provided with extra-nuptial nectaries frequented by ants. These ancestors gave rise on the one hand to types that preserved the primitive features, as in Pogonanthera, and on the other to types better adapted to myrmecophilism, as in Pachycentria. The ants, impelled by their hereditary habits to visit. those plants provided with foliary nectaries, continued to visit them even when those nectaries were undergoing atrophy. Profiting by the tendency of these plants to form tuberculous roots, they have progressively transformed these tubercles into ant nests, causing, by the irritation of their presence, a more marked hypertrophy of those orgaus. In a word, the Pogonantheras, utilizing the protection the ants afford against plant-eating animals, may have found it a real advantage to give those insects a mere asylum instead of offering them nutritive matters in the form of nectar. It is evidently an economy to the plant to offer simply a lodging to its defenders instead of both food and lodg ing, as does the Acacia cornigera, for example, and other plants that both feed and shelter the ants. If we accept this interpretation, which has only the value of an hypothesis, we would be led to regard the Pogonantheras as having economical myrmecophilous features. In the Pogonantheras, as in the myrmecophilous Rubiaceae, the ants take up

their abode in swollen organs that act as water reservoirs, and consequently aid the plant against drought, which may become suddenly serious for any epiphyte.

The lodging organs of these plants appear to be purely of physiological origin. In other types they appear to have had primitively a parasitic origin. We have already spoken of the analogy between the tubercles of the myrmecophilous Rubiacea and certain galls. But it is more than doubtful whether the origin of these tubercles was primitively parasitic and traumatic. In the types of which we have yet to speak, parasitism, of animal origin, seems to have played an important part, even in certain cases a primordial one, in the formation of myrmecophilous organs.

We will first fix our attention on the myrmecophilous features of Dischidia.

The Dischidias are Asclepiadacea of the farthest Orient. With flexible stems and branches they twine upon trees, and are especially noted for possessing appendages in the form of pitchers. These are generally pendent from the branches, and into them plunge adventitious roots that spring from the supporting peduncle (Pl. XXI). The resemblance of these pitchers to the galls produced on the leaves of various trees by aphides of the genus Pemphigus is such that a number of the early observers of these plants considered them as abnormal organs caused by the punctures of parasitic insects.

The morphology of these curious organs has been fully elucidated by the researches of Treub. They are modified leaves. The normal leaves of Dischidia are orbicular, thick, fleshy, and opposite. A pitcher is merely the blade of a leaf whose lower surface corresponds to the inner surface of the pitcher, and whose petiole is thicker than that of normal leaves. We can get a perfectly good idea of the formation of these organs by imagining the blade of a normal leaf to be folded toward the ground, then turned over and the borders brought together. There is, besides, a change of growth in the young developing pitcher, its increase being almost wholly along its middle, so that it takes the form of a hood, with its opening first turned downward, then becoming gradually set more or less upright.

The Dischidias have opposite leaves, but the normal leaf opposite the pitcher usually aborts. When the young urn takes on the form of an elongated flask, there are produced upon its petiole some adventitious rcots, of which those arising near the mouth of the pitcher enter its cavity. A full-grown pitcher usually contains one or two long adventitious roots provided with a well developed system of radicles (fig. 1, Pl. XXII). The internal surface of these pitchers is purple, while their external surface is a grayish, glaucous green, like that of the surface of the stems and leaves.

The direction assumed by the pitchers is variable and merits some attention. The greater number are hung vertically with the mouth