The spindle fibers are formed from the ends of the paradesmose or the centroblepharoplasts, pass through the nuclear membrane, which remains intact, and the chromosomes become arranged upon the spindle in the equatorial plate by a process analogous to that found in Trichonympha. The paradesmose is found outside the nuclear membrane (figs. 8, 9) and fades out in the late telophase precisely as does that of the other species.

It is thus evident that these two organisms, though differing widely in their general morphological characters, yet follow precisely similar modes of mitosis. These similarities consist in the following; Division in both is fundamentally longitudinal, including the splitting of the chromosomes, the centroblepharoplast and the entire body. The chromosomes are precocious in their splitting, doubling the number which later goes on the spindle. The

FIGS. 6-10. 4 Trichomonas angusta Alexeiefft, after Kofoid and Swezey. × 1500. 6 Active motile form prior to division. 7. Late prophase with five pairs of split chromosomes centroblepharoplasts connected by paradesmose. 8. Late metaphase; paradesmose outside nuclear membrane. 9. Telophase with centroblepharoplasts still connected by paradesmose. 10. Mitosis completed, cytoplastic division approaching. Ax., axostyle; c., centroblepharoplasts; m., myonemes; par., paradesmose.

centroblepharoplast divides with a paradesmose formed to connect the two parts as they separate. The motor organelles are divided with new outgrowths to complete the full complement of each daughter cell. These remain attached to the centroblepharoplasts throughout the entire process of division. The nuclear membrane persists, with the spindle fibers passing through from the ends of the paradesmose, which remains outside the membrane. Pseudosynapsis occurs as part of the chromosome cycle in both forms. The paradesmose persists through the late telophase stage after the rounding up of the daughter nuclei, fading out before plasmotomy occurs.

Two phases of these processes deserve particular attention. These are the formation of the paradesmose and the attachment of the motor organelles to the centrosomes. So far as present records go both of these phenomena are

typically and solely flagellate in their occurrence. Nothing comparable is known to occur in the ciliates in which at mitosis the cilia bear no morphological relation to the centrosomes or to any other part of the mitotic figure. In the group of flagellates comprised in the order Polymastigina, the paradesmose is a characteristic feature of mitosis in so far as these processes have been studied. In the trichomonad flagellates, as pointed out above, its occurrence is constant and characteristic. In the nearly related form, Hexamitus, the paradesmose is also present. A peculiar development of this structure is found in the so-called sphere of Noctiluca. This is produced by the elongation and division of a mass of differentiated archoplasm outside the nucleus. The exact relation of this to the tentacle and flagellum of Noctiluca has not been determined, but its position would suggest that some close connection between them exists.

It has been pointed out in our earlier paper on trichomonad flagellates that the paradesmose is not a precise homologue of the centrodesmose or central spindle of the metazoan type of mitosis. It is the result of two peculiar specializations in trichomonad division, the continuity of the nuclear membrane which excludes it from the typically axial position of the central spindle, and the connection of the centrosome-blepharoplast with the entire motor organelle complex. The latter feature is probably the most important one in its bearing on the development of the paradesmose, the strain on the spindle resulting from the constant activity of a group of flagella attached to each pole, requiring the development of a stronger support than is found in the mitotic structures where this condition does not prevail. As a result of this necessity we find a paradesmose developed in those flagellates where the flagella form part of the mitotic figure, but it is usually, if not always, absent where this does not occur.

In no other group of Protozoa do we find a structure approximating the paradesmose of the flagellates. Among the ciliates division of the nucleus occurs as distinct phenomena separated from, though synchronous with, the division of the motor organelles and other structures of the body.

Another point in the division process of Trichonympha which is equally important as those mentioned above, is the fact that division is longitudinal. This is shown in the longitudinal splitting of the centroblepharoplast and the ectoplasmic structures. This fact also serves to separate it from the ciliates where transverse division is general, and allies it with the flagellates.

The occurrence of these peculiar flagellate specializations in Trichonympha would therefore preclude the possibility of any ciliate affinities for that genus. This conclusion receives confirmation from a careful analysis of its morphological features which, though superficially ciliate in appearance, yet are fundamentally flagellate in their character and relationships.

The relations of other members of the group of curious and peculiar organisms which, in company with Trichonympha, are parasitic in the intestinal

tract of the termites, have also been obscured by the high degree of specialization shown in their motor organelles. In one of these, Joenia, Grassi and Foa, have figured longitudinal division with the formation of a prominent paradesmose which persists until the daughter cells are ready for plasmotomy. The paradesmose here, as in other flagellates, is intimately connected with the flagella and their related neuromotor structures.

In Lophomonas the old motor organelles disappear, according to Janicki who has described division in this form. An entire new motor organelle system is developed from the ends of the paradesmose after completion of division of the nucleus. The exact origin of the paradesmose in this case cannot be determined from the figures of Janicki, but it arises from the nuclearneuromotor complex. It remains outside the nuclear membrane, hence is evidently a true paradesmose.

In the remainder of this group mitotic phenomena have not yet been described and it is in their morphology only that we must look for relationships. These are found in the relations of flagella, internal myonemes and centroblepharoplasts with their various modifications, the neuromotor sys

In Spirotrichonympha Grassi' has figured, in these structures, relations which are comparable with those found in Trichonympha, although much. simpler. The flagella are arranged in spiral courses around the body starting from the anterior tip. Each series is accompanied by a slender band or myoneme extending along the line of basal granules beneath the surface of the body. The number of lines of flagella and myonemes varies slightly in the different species. At the anterior tip of the body these are joined in a small granular mass, the centroblepharoplast.

Zulueta has figured the same spiral myonemes in Dinenympha without the series of flagella outlining their course. A single flagellum arises from the end of each myoneme. At the time of division the small granular mass at the anterior tip of the body divides, each moiety taking four of the myonemes, and forming a paradesmose between them as they separate. This centroblepharoplast acts as the centrosome in the formation of the spindle for the division of the nucleus, with its attached myonemes taking the place of the astral rays, as in Trichonympha (fig. 2).

The same relations of motor organelles and internal myonemes, by means of which an integrated neuromotor system is formed, may be found throughout all the members of this group of organisms. The range in complexity extends from the trichomonad type of structure to that shown in Trichonympha, which exhibits a higher degree of specialization and development. than do many of the lower Metazoa. This specialization is confined almost exclusively to the motor organelles and the accessory structures connected with them, the neuromotor system. At the time of division this acts as a unit, dividing and half going to each daughter cell. An apparent exception to this is found in Lophomonas, but here a part of the old neuromotor system

is evidently retained in the paradesmose and from it the new motor organelles of each daughter cell are formed.

In these structures and more particularly in the processes by means of which they are passed from one generation to the next, we have a clue to relationships that seems to be more fundamental and significant in its importance than are external features, which are constantly subject to modifications through environmental changes. The amount of specialization has in no way changed the basic facts of these relations and processes. Thus in the highly complex Trichonympha there are precise homologues with the simpler features in Trichomonas with its simple centroblepharoplast granule and few flagella.

The superficial resemblance between the trichonymph parasites of the termites and the ciliates is the result of the high degree of specialization and evolutionary development to which the former have attained, and is not indicative of a derivative relationship of even the most remote kind between the two. The Trichonymphidae are fundamentally and characteristically flagellate in their type of structure as well as in their methods of division. We may therefore dismiss completely the early conception of Leidy, Kent and their followers that the Trichonymphs are ciliates and revise our wide-spread conception that flagella are universally or even characteristically few in number. These protoplasmic processes are flagella primarily because of their relation to the nucleus and the mitotic figure. Flagella are attached directly or indirectly to the centrosome and share in mitotic processes. Cilia are not thus attached and have no correlated part in mitosis. Numbers contribute no necessary part of this definition of flagella, they apparently do of cilia.

We may also dismiss the later conception of Hartmann that the Trichonymphs are an intermediate group between flagellates and ciliates. In the first place a transition type between these primary groups can not be expected as parasites of a highly organized group of social insects. The appearance of transition is illusory, depending on superficial structures and numbers merely, while the deeply significant mitotic process and its structures are unequivocally flagellate in nature. We therefore reject Hartmann's transitional conception and with it his Trichonymphida and retain Grassi's Hypermastigina as the fitting as well as the legitimate designation for this most highly specialized group of the flagellates.

1 Leidy, J., Proc. Acad. Nat. Sci., Philadelphia, (Ser. 2), 7, 1877, (146–149).

Kofoid, C. A., and Christiansen, E. B., Univ. Calif. Publ. Zool., Berkeley, 16, 1915, (30– 54), pls. 5-8, 1 fig. in text.

3

Swezy, O., Ibid., 16, 1915, (71-88), pls. 9-11.

4 Kofoid, C. A., and Swezy, O., Proc. Amer. Acad. Arts Sci., Boston, 51, 1915, (289–374), pls. 1-8, 7 figs. in text.

'Grassi, B., and Foa, A., R. C. R. Acad. Lincei, Rome, (See. 5), 13, 1904, (241–253), 17 figs. in text.

6 Janicki, C., Zs. wiss. Zool., Leipzig, 95, 1910 (243–315), pls. 6–9.

7 Grassi, B., and Sandias, A., Galatola, Catania, 1893, (1-151), pls. 1–5.

8 Zulueta, A. de., Trab. Nac. Cienc. Nat., Madrid, 23, 1915, 25 pp., 1 pl. 6 figs. in text.

9 Hartmann, M. Hertwig's Festsch., 1, 1910 (349–392), pls. 27–30, 3 figs, in text.

THE TERNARY SYSTEM CaO-MgO-SiO2

BY J. B. FERGUSON AND H. E. MERWIN

GEOPHYSICAL LABORATORY, CARNEGIE INSTITUTION OF WASHINGTON

Communicated by A. L. Day, November 25, 1918

A number of investigations dealing with one or more of the four oxides, lime, alumina, magnesia and silica, have in recent years been carried out at the Geophysical Laboratory of the Carnegie Institution of Washington, as a preliminary step in the study of the rocks and minerals of the earth's crust. These studies1 include all of the possible binary systems and three of the four possible ternary systems which may be constructed from these oxides. The fourth ternary system has been studied only in part. We desire in this paper to present a summary of the results we have obtained in a study of the remainder of this fourth ternary system, and also to correlate these results with the results previously obtained.

The experimental methods employed are similar to the methods used in the previous investigations at this laboratory. Samples of known compositions were prepared by fusing together in platinum crucibles weighed amounts of chemically pure calcium carbonate, magnesia and silica and subsequently reducing the samples to fine powders. The production of homogeneous samples usually necessitated the repeating of this process once or twice. The investigation of a sample was conducted as follows: a small quantity of the sample was wrapped in a piece of platinum foil which was about 1 sq. cm. in area and this charge was tied by a fine platinum wire to a small ring of marquardt porcelain. The porcelain ring was hung on a fine platinum wire, the ends of which were connected to two stout platinum leads. A marquardt porcelain tube carried these leads and also a platinum-platinrhodium thermoelement the junction of which was not more than a few millimeters from the charge. The charge, thermoelement, etc., were inserted into a hot platinum resistance furnace and the furnace temperature was regulated in such a manner as to maintain the charge at a predetermined temperature for a sufficient time to allow an equilibrium condition to be attained (usually from fifteen to thirty minutes but sometimes as long as forty-eight hours). The wire supporting the ring and the charge was then fused by passing an electric cur