G. L. Stebbins, JR.

Cytogenetic Studies in Paeonia II. The Cytology of the diploid Species and Hybrids

Genetics 23: 83-110 Jan. 1938

University of California, Berkeley, California

Received September 4, 1937


The cytological study of the numerous species and hybrids in the Saunders collection has now been carried on intermittently for five years, during which time a considerable amount of data have been accumulated. The present paper aims to present the most significant findings in the diploid species and hybrids, particularly insofar as they help to explain species relationships in this most complex and interesting genus. The somatic chromosomes of Paeonia have already been studied by Langlet (1928), Dermen (1933), and Dark (1936), and those in meiosis of some species and hybrids by Sax (1932, 1937b, Dark (1936), and the present writer (Hicks and Stebbins 1934). These studies have revealed not only very favorable chromosomes for a study of chromatid relationships, but also a large amount of structural hybridity. The latter subject is, of course, of prime importance in connection with interspecific relationships and species evolution, and has received particular attention in the present paper.


The source of the plants used for this study has been discussed in detail by Saunders and Stebbins (1938). Fortunately, due to the presence of a large number of wild and little cultivated species gathered together by Dr. Saunders, the difficulty expressed by Dark (1936), that hybridization under cultivation has obscured the relationships of many paeony species, has largely been overcome. As previously mentioned (Saunders and Stebbins 1938), this difficulty is confined chiefly to the tetraploid species and varieties of Europe. For this reason they have received less attention than the diploids which, with two exceptions (P. suffruticosa and P. albiflora) represent nearly or quite pure species genetically. As will be seen these two much cultivated species are cytologically indistinguishable from the wild ones.

The somatic chromosomes were obtained mostly from smear preparations, made according to the "Kochmethode" of Heitz (1926) and two different modifications of it: (1) that of Whitaker (1934) and (2) the following method, devised by Dr. A. P. Saunders. A bit of tissue (in Paeonia young anthers and stigmas are used in preference to root tips, since they are more easily obtained) is placed on a slide, flooded with 50 percent acetic acid, covered with a cover glass, and boiled gently on a hot plate for 15-20 seconds; by that time the tissue is translucent and soft. The cover glass is then removed, and the slide is dried as completely as possible with absorbent paper. The tissue is then smeared with the edge of another clean slide, flooded with a fixative (LaCour's 2 BD was most frequently used) and fixed for 5-10 minutes. It is then stained with Newton's iodine-gentian violet in the usual manner. This method, though somewhat erratic, produced excellent results in many cases, success depending largely on the completeness with which the tissue could be smeared so that the smear was only one or two layers of cells in thickness. In many cases the chromosomes were spread out flat, as in Whitaker's method, so that they could easily be measured. Although the boiling in acetic acid damages considerably the outer cells of the tissue, the inner cells are little or not at all affected, so that if the tissue is well smeared these can be stained to show not only the gross morphology but in addition many of the internal details of the chromosomes.

The meiotic chromosomes were studied entirely from smear preparations fixed in LaCour's 2 BD and stained with Newton's iodine-gentian violet.

Drawings were made with a camera lucida at a magnification of 5000, and were reduced to one-third this size in reproduction.


The species of Paeonia show a striking similarity in both the number and the morphology of their chromosomes. All are either diploid with n = 5, or tetraploid with n = 10, Although the numbers of most of them have already been given by Langlet (1928) and Dark (1936), the following counts are reported here for the first time. Diploid (n = 5): P. Brownii, Broteri, obovata. Tetraploid (n = 10): P. corallina, Corsica, tomentosa. Those previously reported as tetraploid are P. tenuifolia "hybrida," P. officinalis and relatives, P. Wiltmanniana, and P. coriacea; all of the others are diploid. P. corallina and perhaps P. Corsica contain both diploid and tetraploid forms, but the tetraploid count reported by Langlet for "P. obovata alba" is probably based on the Chinese form of P. Wittmanniana (= P. Willmottiae Stapf), which is frequently known in cultivation by the above name.


In their morphology, the chromosome sets of all of the diploid species conform to the same karyotype. This is illustrated in figure 1, which shows the haploid complement, drawn from selected chromosomes of the somatic anaphase, of a representative of each of the diploid species. P. Brownii, which is taxonomically and genetically remote from all of the other species, is not illustrated, as it forms a problem by itself, which will be dealt with in a later paper. Its karyotype is, however, quite similar to those of the other species. This karyotype consists of one pair of long chromosomes with median or submedian constrictions, (designated as A), two pairs of shorter chromosomes, also with median or submedian constrictions (B and C), one pair (D) with submedian to subterminal constrictions (arm ratio about 2:1), and often a small satellite on the shorter arm, and the final pair (E) with subterminal constrictions and a satellite on the shorter arm.

The size of the chromosomes of all the species is nearly the same, as is evidenced by measurements made at meiotic anaphase. Such size differences as appear in the diagrams are due to unavoidable differences in the cells used for drawing, and comparable differences can be found in the different tissues of any plant of Paeonia. The same may be said of the size differences in the figures of Dark (1936). The two forms, P. Beresowskii (P. anomala Beresowskii), and P. Smouthi (P. albiflora × tenuifolia) whose somatic chromosomes appear smaller than the others (his figures 1d and 1f) have meiotic chromosomes of exactly the same size as those of the other species. This is illustrated for P. Smouthi in his figures nb and c, and in figure 5 of the present paper, while figure 4 of this paper represents a variety of P. anomala taxonomically very close to Beresowskii, and with chromosomes of just the same size. Hence differences between the complements of the various species, as noted also by Emsweller and Jones (1935) in Allium, cannot be very well expressed by a recording of the actual length of the chromosome arms. In Paeonia, however, each species is characterized by a definite relation of the lengths of the different arms to each other, and this relation can be expressed and compared in the different species if the length of a particular arm in any given set is used as a unit, and the lengths of the other arms of that set are given in terms of that unit. This was done and results are recorded in table 1. These emphasize the general similarity between the complements of the different species, but show also that, even with the slight differences that exist, there is more similarity between subspecies of the same species and closely related species, than between species widely separated taxonomically. For instance, the correspondence between the figures for the three forms of P. anomala is very close, and the three species with entire leaflets and without strong midribs on their innermost sepals, namely, P. triternata Mlokosewitschii, P. Broteri, and P. obovata all resemble each other in the relatively smaller size of their D and particularly their A chromosomes. On the other hand, the wide taxonomic difference between P. Delavayi, P. suffruticosa, and the herbaceous species is not reflected by any comparable difference in their chromosomes, the greatest difference being the marked inequality in the arms of the A chromosome in the shrubby species, particularly P. Delavayi. On the basis of these ratios, the "mean arm deficiency," as defined by Levitzky (1931), could be calculated, and is recorded in table 1. As expected, there is little difference between the various species in this figure, and here again similar values are correlated with taxonomic similarity of the varieties and species.



The complements of the tetraploid species are more difficult to measure, on account of the greater number of the chromosomes, and their crowding on the cell plate, but such observations as were made indicate that they are quite similar to those of the diploids. The haploid complement of P. Corsica (fig. 1K) contains the same types as P. triternata Mlokosewitschii and P. obovata, its near relatives among the diploid species, but each type is represented by two chromosomes. The inequality of the two representatives of each type, particularly the D and E chromosomes, indicate that the form of P. Corsica in the Saunders collection is an allotetraploid. This indication is supported by the configurations found at meiosis, as will be described in a later paper.

Unfortunately, no satisfactory preparations to illustrate the somatic chromosomes of the other tetraploid species were obtained. The tetraploid varieties of the P. corallina complex resemble somewhat in their external morphology P. Corsica, and their chromosome complement is also similar. That of P. tomentosa, another species of this general relationship, is also similar, all these tetraploid species resembling their diploid relatives in the relatively small size of their A chromosomes, which are hardly distinguishable from the B and C pairs. No cytological preparations have been made of P. coriacea, Wittmanniana, and Willmottiae, which are not as yet well established in the Saunder’s garden. Langlet's report of the tetraploid number for these species is supported by observations of meiosis in hybrids of them with P. albiflora and tomentosa, and by measurements of their stomata.

The complex of P. officinalis, the largest and most polymorphic of the species complexes in the genus, deserves special attention on account of other evidence concerning its origin, and will be treated in a later paper.


Hicks and Stebbins (1934) described some features of meiosis in P. albiflora, tenuifolia, triternata Mlokosewitschii, and suffruticosa, in which emphasis was placed on certain abnormalities: asynapsis, abnormal separation of bivalents, fragmentation, and "chromatid fusion." The latter phenomenon is in most cases associated with fragmentation (Hicks and Stebbins 1934, figs. i and 2), and is therefore the result of chiasma formation in an inverted segment (McClintock 1933; Dark 1936). Nevertheless, "chromatid fusion" may occasionally occur without any evident fragment (Sax 1932, fig. 12), and in these cases has probably resulted from some other cause. Fragmentation may not infrequently occur from causes other than chiasma formation in an inverted segment, as is clearly shown by Hicks and Stebbins (1934 fig. 19) and by figures 6A and 6D presented here (see below, under description of hybrids). Of the diploid species which were not studied previously, two, P. anomala, particularly the forms Veitchii and Woodwardii, and P. Delavayi are important as parents of the hybrids to be considered here. These resemble most P. triternata Mlokosewitschii in their relatively high percentages of fragmentation and of abnormal separation. These are recorded in table 2, which includes also another plant of Mlokosewitschii, whose behavior was somewhat different from that recorded in the previous paper, and three additional strains of P. albiflora. The plant called "Vilmorin" is a white single, received from Vilmorin et Cie., Paris, as the wild P. albiflora, although it differs considerably from any herbarium specimens of this form seen by the writer. It has larger relatively narrower leaves and larger flowers, with broader, less strongly notched petals. No. 1505 is a white single seedling from an ordinary horticultural variety.



From table 2 it is evident that the species fall into two groups with respect to the frequency of abnormalities. P. albiflora, Emodi, tenuifolia, and suffruticosa have a relatively low percentage of all types, while in P. anomala, triternata, and Delavayi this is relatively high. Fresh preparations of the only clone of P. tenuifolia available, and of P. albiflora "Silvia Saunders" showed conditions essentially similar to those already recorded. as did also those of another plant of P. suffruticosa.



In addition, the chiasma frequency at metaphase of the various species was computed, and is shown in table 3. The most striking fact evident from a study of these figures is the 'great similarity between the species, and the lack of any correlation between chiasma frequency and percentage of asynapsis. This is due to the higher percentage of bivalents with three and four chiasmata in P. anomala, triternata Mlokosewitschii, and Delavayi, as is illustrated in figure 2, showing the five bivalents from a nucleus of P. anomala Veitchii with a relatively high number (12) of chiasmata.



There is a consequently greater coefficient of variability of the mean chiasma frequency in these species, though this difference is only in some cases significant, for example, as between P. triternata Mlokosewitschii, C.V.=-45.7 ±2.5, and P. tenuifolia C.V. =37.2 ± 2.1 (D/S.E. =2.6). This difference is illustrated in the diagram, figure 3.



The terminalization coefficients at metaphase are essentially similar in all of the species, and here again there is no correlation with the meiotic abnormalities.

In addition to the abnormalities described above, multivalents have been found in two diploid species. In P. Delavayi these are as described and illustrated by Dark (1936), and were interpreted by him as the result of the pairing of reduplicated segments of non-homologous chromosomes.

Their occurrence in this species, however, is very rare, as in the material investigated by the writer only two cases were seen in about 200 cells, and with the exception of a doubtful case in P. suffruticosa, multivalents of this type have not been found in any other diploid species of Paeonia.

In one form of P. anomala, including all of the plants received from dr. Hesse (cf. Saunders and Stebbins 1938) a different type of multivalent was observed in 60-80 percent of the metaphase nuclei. There were either trivalents or quadrivalents, and they were in the form of rings or chains, with or without terminalization of chiasmata (fig. 4A, C, D).



They always included the submedianally constricted (D) pair and one of the smaller medianally constricted (B or C) pairs of chromosomes. In most of the nuclei of this form which did not contain multivalents, there were three pairs of equal homologues and two unequal pairs, each involving one B or C and one D chromosome (fig. 4B). This condition apparently resulted from the failure of chiasma formation in two of the arms of an x-shaped multivalent pachytene configuration such as that illustrated by Cooper and Brink (1931) for Zea, and, along with the lack of triple chiasmata in the multivalents, is strong evidence that this race of P. anomala is heterozygous for an interchange of non-homologous segments, involving a large portion of the long arm of a D, and of one arm of a B or C chromosome. Two different plants were studied, one from .the group received as P. anomala, and one from those designated P. anomala alba, and since both showed the same type of multivalent and unequal bivalent configurations, it is very likely that these occur in all of the plants. An extraordinary fact is that their pollen is about 75-80 percent viable, as determined by germination tests. Since cytological observation indicates that in three-fourths of the cells the separation is "disjunctional," that is, homologous chromosomes pass to opposite poles, this fertility is not unexpected.

An explanation for the preponderance of disjunctional separations in organisms with terminalized chiasmata is given by Darlington (1937, pp. 151-152), that is, that such rings are flexible enough that the "forces of repulsion associated with the centromere are able to effect regular disjunction. . . ." This explanation holds at least to some extent in P. anomala even though the chiasmata are not completely terminalized, since they are mostly near the distal ends of the chromosome in this configuration, and are never numerous as in Pisum and Hyacinthus. Furthermore, the chromosomes of Paeonia are unusually large for their nucleus, giving a "crowded" cell plate (Darlington 1936a), in which the free distal ends of the longer bivalents are actually near the end wall of the cells. Hence it seems likely that pressure from the cell walls would in most cases prevent the formation of the open non-disjunctional type of configuration illustrated in figure 4D, and would absolutely prevent the linear orientation of a trivalent. This latter configuration has not been found.


The group of hybrids which are the subject of this study includes all possible crosses between four of the herbaceous species of the subgenus Paeon, P. albiflora, P. anomala Veitchii and Woodwardvi, P. tenuifolia, and P. triternata Mlokosewitschii, and between the two shrubby species of the subgenus Moutan, P. Delavayi and P. suffruticosa. Members of different subgenera cannot be crossed, and within Paeon the cross P. albiflora × triternata Mlokosewitschii has failed in spite of repeated efforts (Saunders and Stebbins 1938).

The characteristics of meiosis in the various hybrids are summarized in table 4. Since in these hybrids many nuclei contain only 3, 2, 1, or even 0 bivalents, the mere listing of the percentage of cells with univalents is not significant as it is in the species, where practically all of the nuclei contain either 5 or 4 bivalents. The average number of bivalents per nucleus, moreover, cannot be accurately determined in many of the hybrids, on account of the frequent presence of trivalent and quadrivalent associations. Hence the amount of chromosome pairing characteristic of each hybrid is best expressed by giving the average number of univalents per nucleus. The percentage of cells with fragments also cannot be satisfactorily listed, since in many cases the univalents split at first anaphase, producing single chromatids which are in many anaphases impossible to distinguish from large single chromatid fragments. On the other hand, the chiasma frequency is here low enough so that two chiasmata occur only rarely in an arm bearing an inverted segment, and hence the percentage of bivalents showing a fragment and a continuous chromatid is some indication of the amount of pairing of inverted segments in these hybrids.

This table shows that the hybrids are of two types. In the first type, represented by P. albiflora × tenuifolia and P. albiflora × anomala Veitchii, the chromosome pairing is nearly as complete as in some of the species, and the reduction in chiasma frequency is relatively slight. Furthermore, the percentage of nuclei bearing fragments, though greater than that in the parents of these two hybrids (Hicks and Stebbins 1934, table 1) is about the same as in several of the diploid species. There is a low percentage of configurations indicating chiasma formation in an inverted segment, and multivalent configurations, as in the species, are absent or rare. The terminalization coefficient in all of the hybrids shows the same range of variation as in the species, and since the prophase stages could not be studied in this material, this figure is of relatively little significance in Paeonia. Figure 5 shows a first metaphase nucleus of P. albiflora × tenuifolia no. 8276, in which there is complete pairing, a relatively high number of chiasmata (9), and a case of distal interlocking, which is not uncommon in Paeonia. It is interesting to note that in their cytological as well as their morphological characteristics the seedlings of this cross produced recently in the Saunders garden, represented in this study by nos. 8276 and 8280, are essentially the same as the clone, "P. Smouthi" of the same hybrid which was produced almost a hundred years ago, and is represented in this work by three different plants, no. 6556 recorded here, and nos. 6550 and 6582, recorded by Hicks and Stebbins (1934). Since this is true in spite of the fact that the slides on which the studies are based were made during three different seasons, 1932, 1933, and 1934, there is good evidence that normal fluctuations in the external environment have little effect on meiosis in this hybrid.

In regard to pollen sterility nos. 8276 and 8280 also compare rather closely with P. Smouthi. Although there are about 40 percent of normal appearing grains, only a small percentage of these are viable under artificial conditions of germination. In the other hybrid of this group, P. albiflora × anomala Veitchii, nearly all of the pollen is shrunken and obviously sterile.

The members of the second group, which constitute the majority of the diploid hybrids, show a much higher degree of meiotic abnormality in every respect. First metaphase nuclei which show complete pairing are relatively rare, occurring in only 6-13 percent of the total while the most frequent number of bivalents per nucleus is correspondingly lower, and in every case the number of chiasmata per (potential) bivalent is less than one. The hybrids are all essentially the same in this respect, and such significant differences as exist occur between hybrids whose parents were different plants of the same species, that is, between P. tenuifolia × P. anomala Woodwardii, as well as between the two numbers of P. anomala Veitchii × P. triternata Mlokosewitschii, nos. 3469 and 5360, of which the Mlokosewitschii parents were different plants.



Fragmentation is extremely common in these hybrids and may result either from chiasma formation in an inverted segment, or from other unknown causes. Both types are shown for P. Delavayi var. lutea × P. suffruticosa, "Argosy," in figure 6A in which the left hand bivalent shows a fragment still paired with the normal chromatids, while one of the univalents has split, and of the separate chromatids one has broken in two, and the other, situated near the upper pole of the spindle, shows a partial break. In figures 6B-D, also from "Argosy," fragmentation of bivalents is seen. Figure 6B shows a stage preliminary to that illustrated by Hicks and Stebbins (1934 fig. 19), while figures 6C and 6D show two stages of the same type of fragmentation involving two chromatids.



An abnormality not encountered in any of the species or in the hybrid P. albiflora × tenuifolia, but occurring in all of the other diploid hybrids is the formation of "restitution nuclei" at the end of the first division of meiosis, and the consequent presence of a single metaphase plate at the second division, resulting in the formation of dyads of diploid gametes. A second metaphase plate of this type is illustrated in figure 7 of P. anomala Veitchii × P. triternata Mlokosewitschii. While generally most frequent in the hybrids in which chromatin bridges are also the most commonly found, and due partly to the persistence of these bridges, as is evidenced by the frequent presence of dumbbell shaped nuclei, the formation of restitution nuclei at first telophase occurs in one hybrid in which chromatin bridges are relatively uncommon, P. albiflora × anomala Veitchii. This hybrid, although it forms only two percent more chromatin bridges than P. albiflora × tenuifolia, nevertheless produces restitution nuclei in 12 percent of its sporocytes, while no such nuclei have been seen among about 300 sporocytes at second metaphase in P. albiflora × tenuifolia. On the other hand, P. anomala Veitchii × tenuifolia, although it has almost three times as high a percentage of chromatin bridges (11.9 percent) as P. albiflora × anomala Veitchii nevertheless forms only 14 percent of restitution nuclei. That persistent chromatin bridges do not in most cases produce by themselves restitution nuclei is made evident by the frequent presence in such species as P. suffruticosa and P. triternata Mlokosewitschii (Hicks and Stebbins 1934, figs. 2 and 8) of quite distinct homoeotypic metaphase plates connected by such a bridge.

The most characteristic and unusual meiotic abnormality in the hybrids of this group except for P. anomala Beresowskii × Emodi is the formation of trivalent, quadrivalent, and quinquevalent associations. These configurations often show at late metaphase or early anaphase triple or quadruple chiasmata, like those illustrated by Dark (1936, fig. 7) for P. Delavayi. This suggests that they are probably produced by the pairing of reduplicated segments, as suggested by Dark, but another explanation is possible, that many small interstitial translocated segments are present in these hybrids. This explanation however, is made less probable than Dark's by the following facts:



1. Study of a single chromosome pair (that with subterminal constrictions) in different cells of P. Delavayi var. lutea × P. suffruticosa indicates that chiasmata may form at almost any point along the length of the chromosomes of this pair (fig. 8, A-G). The separation of a completely terminal chiasma is not included in this figure, but is illustrated in figure 11A, and many such chiasmata were observed at metaphase in other cells. Nevertheless, at least three different types of multivalent configurations occur which involve one or both members of this pair and those of a different pair (figs. 10A, E, F, G, H). If these are due to translocations, therefore, the total length of each translocated segment cannot be greater than 1/10 the length of the chromosome.

2. The low chiasma frequency characteristic of all of the hybrids in which these multivalents are abundant indicates a lack of homology between the parental chromosomes, and a relatively loose association between them at pachytene. The formation of triple and quadruple chiasmata, however, requires a rather intimate type of pairing in certain regions, with frequent exchanges of partners as in polyploids (Darlington 1937, pp. 110-122), which could be expected to occur only rarely in such hybrids. On the other hand, reduplicated segments occurring in two different chromosomes derived from the same parent would be completely homologous, and hence might be expected to pair rather frequently, if the corresponding segments derived from the other parent were less homologous. Further evidence in this direction is presented below.


3. In one case (fig. 9, A-C), two different cells of the hybrid P. tenuifolia. × P. anomala Veitchii each contained a trivalent apparently involving the same three chromosomes, one with a submedian and two with median constrictions. In both figures two chiasmata are present, one normal, involving the two median chromosomes, and the other, formed in an inverted segment, involving one median and one submedian (D) chromosome. The former type of chiasma is in figure 9A in a different arm from the latter, while in figure 9B the two chiasmata are in the same arm. The inverted chiasma is in figure 9A. nearer the distal end of the median and the proximal end of the submedian chromosome, while in figure 9b it is nearer to the distal end of the submedian chromosome. If, as is suggested by the similar size of the two fragments, two chiasmata were formed at different points along the same inverted pairing segment, then the two members of the submedian (D) pair should not form chiasmata in this segment, unless the segment occurs in all three chromosomes. The fact that in another cell of the same hybrid the two D chromosomes were found paired in this fashion (fig. 9C) indicates that the two trivalents illustrated result from chiasma formation in an inverted reduplication.

ANALYSIS OF STRUCTURAL HYBRIDITY IN P. Delavayi var. lutea × P. suffruticosa

On account of their variety and complexity, configurations of this type have been studied intensively in one hybrid, P. Delavayi × suffruticosa, "Argosy," of which exceptionally good preparations were obtained during the seasons of both 1933 and 1934. The two sets of preparations show no observable differences from each other, suggesting that even this very irregular type of hybrid meiosis is in Paeonia little affected by normal fluctuations in the external environment.

The simplest type of configuration found is illustrated in figure 10A, a quadrivalent resulting from pairing of a reduplicated segment involving the distal end of the long arm of the subterminally constructed (E) pair and of one of the medianally constricted (A, B, or C) pairs. The reduplication in figure ioB involves the shorter arm of a submedianally constricted (D) chromosome and a proximal segment of a median pair.



In all of the other multivalents illustrated chiasma formation has taken place in inverted as well as reduplicated segments. In figure l0C two submedian (D) homologues are held together by a chiasma in an inverted proximal segment of their long arms, and the distal end of each is paired with a medially constricted chromosome, in one case by a reduplicated segment which is apparently at the very end of one, but not of the other chromosome. In all of the other configurations illustrated a chiasma has formed in an inverted reduplication of otherwise non-homologous chromosomes, while one or both of these have in addition formed a normal chiasma with their true homologue. In figures 10D and 10E the inverted reduplication is distal to the normal chiasma, so that the fragment is paired with chromatids representing only two of the three members of the trivalent, while in figures 10F, 10G, and 10H, it is proximal, and the chromatid fragment is paired with chromatids of three different chromosomes. Figure ioH shows at the left evidence of a deficiency in one of two homologues. In figure 10B a single chromatid has broken, while figure 10C shows a type of fragmentation similar to that in figure 6D. Figure 11A shows a whole nucleus in which half of the ten chromosomes have formed a quinquevalent which involves the pairing of three different nonhomologous chromosomes by means of reduplicated segments. Figure 11B is a very complex trivalent containing two fragments, one larger and one smaller. It has apparently resulted from chiasma formation in two different inverted segments, one belonging to homologous chromosomes, and the other an inverted reduplication.

Figure 11C is a remarkable type of bivalent found only occasionally, for which two possible explanations exist. Either it is the result of a reduplication involving the two ends of the same chromosome, or of a long inversion involving the attachment constriction. The latter, suggested to the author by Dr. D. G. Catcheside, is the more likely.


The significance of interspecific differences in the character of meiosis

The diploid species of the Old World subgenera of Paeonia are all essentially alike in the size and external morphology of their somatic chromosomes, and the present work demonstrates also their slight variability in two characteristics of meiosis, the frequency and distribution of the chiasmata at metaphase. Nevertheless, there are definite differences between the species in one cytological characteristic—the percentage of meiotic abnormalities, particularly of fragmentation. In this respect the species can be roughly divided into two groups: P. albiflora, Emodi, tenuifolia, and suffruticosa with a low percentage of these abnormalities, and P. anomala, triternata, and Delavayi in which they are relatively frequent. As emphasized previously (Hicks and Stebbins 1934) there is no correlation between the percentage of these abnormalities and that of asynapsis or of pollen sterility, and from the results of this study it may be said that there is also none between it and the chiasma frequency. In fact the chiasma frequencies of those species with frequent abnormalities have slightly, though not significantly higher averages than those in which they are less common.

Since the fragmentation is due chiefly to chiasma formation in inverted segments, and since the chiasma frequency of all of the species is approximately the same, the percentage of fragmentation in a species probably is roughly proportional to that of the inverted segments for which it is heterozygous. Therefore P. triternata, anomala, and Delavayi, or at least the clones of these species here studied, are heterozygous for a considerable number of inversions. The following points favor the view that in these species there is a relatively large number of small interstitial inverted segments, rather than that the segments are of the same number as, but larger than those in the species which show few chiasmata in inverted segments:

1. Occasional configurations resulting from chiasma formation proximal to an inversion, as well as in it, were found (figs. 10D, E).

2. In some cells both arms of all of the medianally and submedianally constricted chromosomes were united by chiasmata, but in none of these was there a chromatin bridge and fragment. In such cases chiasma formation must have been in every case distal to the inverted segments.

3. No configurations were found corresponding to those illustrated in richardson's (1936) Diagram 1 as resulting from the formation of two chiasmata within an inverted segment.

4. In one hybrid, P. Delavayi var. lulea × P. suffruticosa, the length of one inversion was determined as not exceeding 1/10 the length of the metaphase chromosome (fig. 8). Probably similar short inversions are responsible for the rare occurrences of the bridge fragment configurations in such species as P. albiflora.

It will be noted that the three species with more numerous inversions not only represent both of the Eurasian subgenera (P. Delavayi in subgenus Moutan, P. anomala and P. triternata in subgenus Paeon), but that the last two species are, respectively, nearly the most primitive and nearly the most advanced of their subgenera (Stebbins 1938). Furthermore, they represent the geographical extremes of the range of the genus, since the range of P. triternata, in the region about the Black Sea, is near the southern limit of the range of at least the diploid species, while P. anomala ranges much farther north than any other species of Paeonia. There is, however, a significant taxonomic characteristic which these species have in common. All three of them are morphologically very variable in nature, while the other species, P. albiflora, Emodi, suffruticosa, and tenuifolia are much more constant in the wild state, although in albiflora and suffruticosa there is a multitude of horticultural forms. This suggests that, although the production of new varieties under cultivation has taken place chiefly through gene mutation, the variations found in nature are due at least partly to the formation of inversions. The suggestion of Hicks and Stebbins (1934), that the production of varying forms, or ecotypes, might be effected by the accumulation of the smaller, non-lethal deficiencies produced by fragmentation is not substantiated by a study of the somatic chromosomes of the different forms, since no significant difference is found between the different forms of a species. On the other hand, Muntzing (1934) has demonstrated that new types of chromosomes could arise through breakage of chromatin bridges, some of which may have translocations added, so that by alternation of translocation and fragmentation considerable structural change could be produced without visible alteration of the karyotype. However, one crucial test of this hypothesis as applied to these species of Paeonia does not favor it. Although one would expect, on this basis, that hybrids between two of the races of any of these species would show more structural differences than the races themselves, this has not proved to be the case. A cursory examination of meiosis in P. anomala Veitchii × Woodwardii was sufficient to determine that it is as regular, if not more regular than in the parent species, and the hybrid is as fertile as its parents. For this reason, furthermore, the role of structural change in producing isolation, and thus favoring independent mutation, as mentioned by Stern (1936), cannot have been important in the evolution of the races (ecotypes) of P. anomala.

Darlington (1936b, 1937, p. 273) has given an explanation which exactly fits the case of these species of Paeonia. He points out that the presence of many small inversions, by inhibiting effective crossing over in the regions in which they occur, .and thereby "holding groups of genes together as units" may establish a "group-discontinuity, a fission in the species." This effect is exactly what is seen in the taxonomic characteristics of P. Delavayi, anomala, and triternata. In Paeonia, with its low number of chromosomes, crossover suppression would be unusually effective in producing discontinuity.

Differences in the frequency and distribution of chiasmata

Although the chiasma frequencies and terminalization coefficients at metaphase do not show any differences between the species that are correlated with any other characteristics, there are nevertheless some interesting points to be noted. In the first place, the chiasma frequencies of both species and hybrids are all about 0.3-0.4 higher than those recorded for the same forms by Dark (1936). Dr. Dark has very kindly examined reproductions of figures 3 and 5, and assures me that this difference is actual. Two possible explanations for this occur to me:

1. The strains in the Saunders garden are different from those examined by Dark. This seems hardly likely in view of the consistency of the differences. Furthermore one form, P. Smouthi, is a sterile hybrid which is very difficult to produce, and therefore every probability exists that the clone of this species in Mr. stern's garden is identical with that of Dr. Saunders.

2. The differences in climatic conditions between England and central New York are responsible. During the early part of the growing season at Clinton, from late April until mid-May, sudden changes of temperature are the rule. The nightly temperature may drop to 30°F (—1°) or even 20° F (—7°C) as on May 15,1936, while the daytime temperatures usually reach 75-80° F (24-2 7° C) for a few days during that period. The extremes over the same period in England are much less. Although Sax (1937a) has shown that sudden changes of temperature produce asynapsis in Tradescantia and Rhoeo, this need not be true in all genera, particularly those whose natural distribution and time of flowering is that of Paeonia. All of the species whose chiasma frequency is here reported are natives of regions where extremes of temperature are the rule (Siberia, Manchuria, western China, the Himalaya and Caucasus Mountains) and are undoubtedly adjusted to them. Hence the climate of Central New York is normal for them, while that of England is not. This abnormality of the external environment may be responsible for the unusually low chiasma frequency and high percentage of asynapsis found by Dark. A crucial test of this explanation would, of course be obtained by keeping a plant of Paeonia in a constant temperature chamber during the beginning of its spring growth.

Distribution of abnormalities in the hybrids

Although no type of meiotic abnormality occurs in the hybrids which is not found also in at least some of the species, all of the abnormalities are, as would be expected, more common in the hybrids than in their parents. An unexpected result, however, is that the amount of abnormality in the hybrids is not at all proportional to the difficulty with which their parents can be crossed, or with the degree of taxonomic difference between them. The easiest cross to make between diploid species is P. tenuifolia × triternata Mlokosewitschii, but the resulting hybrid has a very abnormal meiosis. On the other hand, the two most difficult crosses to make are P. albiflora × tenuifolia and P. albiflora× anomala, and yet these two hybrids are in their meiotic behavior the least abnormal of the series. The taxonomic differences will be discussed below in relation to the problem of chromosome pairing, but it may be mentioned here that, if one considers the sum total of vegetative and floral characteristics, P. albiflora is farther from P. tenuifolia than is any other species of the subgenus Paeon; yet their hybrid has much the most regular meiosis of any. We may conclude, therefore, that the meiotic abnormalities in the hybrid are not produced by the same causes which produce incompatibility between the species, nor are they the direct result of the evolutionary changes which have separated the species from each other. A fuller understanding of their significance depends, therefore, on an analysis of each of the more important abnormalities in turn.

Inversions.—The significance of inversions in producing variability within the species has already been discussed. The frequency with which inversion heterozygotes occur in both animals and plants is amply exemplified by the numerous cases now on record (Darlington 1937, p. 274) even though the method of identification of this type of structural hybridity is a relatively new one; while the studies of Sturtevant and Dobzhansky (1936) on Drosophila pseudoobscura give direct evidence of the correlation of inversions with geographic distribution. Hence inversions must be considered one of the important causes of variation within the species. That they are much less important in producing complete isolation and discontinuity in morphological characteristics is certainly true in Paeonia although in Drosophila they seem to be important in this respect also (dobzhansky and tan 1936).

Multivalent configurations. The most striking feature of all but two of these diploid hybrids is the frequent appearance of multivalents of the type resulting from the pairing of reduplicated segments or from small translocations. Although several points have already been advanced in favor of reduplications rather than translocations as the cause of these multivalents, one point may be considered further here, the correlation between the amount of asynapsis and the frequency of multivalents. This is particularly striking in the triangle of hybrids involving P. albiflora, tenuifolia, and anomala Veitchii. The hybrid P. tenuifolia × Veitchii is the only one of this triangle in which the chiasma frequency is less than one and the average number of univalents per nucleus is more than two, and also the only one in which multivalents occur with any frequency. These multivalents are of three types:

1. Trivalents or quadrivalents of which all the components have median constrictions; no inversions evident (fig. 9D).

2. Trivalents involving two median and one submedian (D) chromosome, the latter always paired inversely with one of the other two (fig. 9A; 9B).

3. Trivalents involving one subterminal (E) and two median chromosomes; no inversions evident (fig. 9E).

If these multivalents are the result of translocations, then we might say that the genom of P. anomala Veitchii differs from that of P. tenuifolia by three translocations.

On the same basis P. anomala Veitchii differs from P. albiflora by only one small translocation. Hence P. albiflora should differ from P. tenuifolia by at least two translocations. But four different plants of P. albiflora × tenuifolia, in spite of their high chiasma frequency, form no multivalents whatever, indicating that P. albiflora differs from P. tenuifolia by no translocations at all. Hence the assumption of interchange clearly does not explain the observed phenomena. On the other hand, if we assume that in P. anomala Veitchii a segment AB, occurs in both an A, B, or C and a D chromosome pair; that another, CD occurs in two different median pairs; and that a third, EF, occurs in the E as well as in A, B, or C chromosome pair, we may explain the phenomena as follows. In P. anomala Veitchii itself AB, CD, and EF pair only with those homologous segments located in chromosomes which are otherwise completely homologous, and hence multivalents never occur. Nevertheless, every gamete of Veitchii possesses two segments each of AB, CD, and EF, which therefore have entered into the hybrid with P. tenuifolia. In the latter species, however, none of the chromosomes are completely homologous to those of Veitchii, and therefore the gamete of P. tenuifolia contributes to the hybrid corresponding segments that may be denoted A'B', C'D', and E'F', which have a weakened affinity for AB, CD, and EF, and which are probably not reduplicated. Hence in this hybrid the two AB segments derived from Veitchii have a greater affinity for each other than either has for A'B' and they can be expected to pair occasionally, as can also the two CD and EF segments. However, since the majority of the chromosomes in which these segments are located still have a weak affinity for their homologues derived from the opposite parent, pairing between them will occur, and multivalents will therefore be produced. A similar explanation, reduction of competition, was used by Catcheside (1932) to explain bivalent formation in a haploid Oenothera. On this basis, multivalent formation due to reduplication should be increased by a lowering of the chiasma frequency, and this is exactly the case in Paeonia.

There are apparently differences between the species in the amount of reduplication. P.Delavayi does not have a significantly lower chiasma frequency than the other diploid species, yet pairing of reduplicated segments occurs in both the plant studied by Dark (1936) and those in the Saunders garden. Since P. Delavayi has been in cultivation for only a short while, and the plants are all derived from seeds obtained by a few expeditions to Western China, there is considerable possibility that these two plants are fairly closely related genetically. Therefore it is likely that both have a similar set of reduplications, which are more numerous than those in the other species. In this connection may be noted the large number of different reduplications which can be detected in P. Delavayi var. lutea × P. suffruticosa, "Argosy." In this hybrid configurations have been drawn representing the following reduplications:

1. The distal ends of the E and of an A, B, or C chromosome, not inverted (fig. 10A).

2. The distal ends of the long arms of an E and of a D chromosome, inverted (fig. 10E).

3. The proximal portion of the long arm of an E and the distal of a D, inverted (figs. 10F, 10H, and probably 10G).

4. The distal ends of the long arms of the D and of an A, B, or C pair of chromosomes, not inverted (fig. 10C, and fig. 11A, rightmost chiasma). That this is actually a reduplication and not a case of segmental interchange is made evident by the occasional association of the two D chromosomes by terminal or nearly terminal chiasmata.

5. The short arm of a D and a proximal segment of an A, B, or C chromosome, not inverted (fig. 10B).

6. The distal, or nearly distal segment of the long arm of a D and an interstitial segment of an A, B, or C chromosome, inverted (fig. 11A, at left, and one unpublished figure).

7. Interstitial segments of two median chromosomes, inverted (figs. 11D and 11B).

There are probably other reduplications whose presence has not been detected. That the multivalent formation involving so many different segments is not due to more or less random non-homologous pairing is made clear by the frequent recurrence of the same type of configuration, in spite of the large number of different ones found.

Multivalent configurations similar to those found in these diploid species and hybrids of Paeonia have been found in diploids of Tradescantia bracteata, (Darlington 1929) and Matthiola incana (Philp and Huskins 1931), while in trisomic forms of Datura (Belling and Blakeslee 1924) and in Drosophila (Bridges 1935) there is evidence for the presence of reduplicated segments within the same chromosome.

They are therefore probably of occasional occurrence throughout the plant and animal kingdoms, but in most organisms can be detected only under very unusual conditions. Two explanations can be given for the unusually abundant evidence for their existence in these hybrids of Paeonia.

1.Reduplications are actually more frequent in Paeonia than in most plants. This could explain in part the remarkable length of the chromosomes in this genus.

2. Due to the great length of the chromosomes and the relatively small size of the nucleus when zygotene pairing begins, the chromosomes are all closely crowded together, and hence, with reduced competition, homologous segments of non-homologous chromosomes have an unusually good chance of coming together. There is of course no way of deciding which of these is the true cause of the frequency of multivalents, and both may be partly responsible for it. Since bridges (1936) has shown that a "mutation" in Drosophila is the result of reduplication, this may be an important factor in the evolution of new forms in Paeonia.

An interesting fact is that the species and hybrids with a great amount of structural hybridity do not have a lower proportion of terminalized chiasmata at metaphase than those in which there is less hybridity. This supports the opinion of Dark (1936), that chiasma movement in Paeonia is very slight. With an equal amount of movement of chiasmata, one should expect a larger proportion to be terminal in hybrids with a genetically induced reduction in pairing. Since as McClintock (1933) has shown pairing begins at the ends of the chromosomes both where the homology is strong and where it is weak, a failure of the chromosomes to complete synapsis would reduce chiasma frequency in their middle parts more than at the ends. This would result in the formation of a relatively high proportion of chiasmata distal to the non-homologous segments, and thus could compensate for the reduction in tenninalization due to change of homology. Although there are no good observations of prophase stages to support this hypothesis, partial examination of three nuclei of P. tenuifolia × triternata at early to mid-diplotene showed the presence of terminal chiasmata already at this stage, even though the total number of chiasmata was very small.

Chromosome pairing and interspecific relationships

Two common criteria for determining the distinctness of species from each other are the amount of sterility in the hybrids between them, and the extent to which their chromosomes fail to pair. In Paeonia, these two criteria are useful to only a limited extent, that is, to distinguish between the varietal or subspecific and the specific status of a. form. There is a sharp distinction between the complete fertility of such hybrids as P. Delavayi× Delavayi var. lutea and P. anomala Veitchii × Woodwardii and the almost complete sterility (at least so far as seed setting is concerned) of all of the true interspecific hybrids. This is undoubtedly a valuable asset in the delimitation of species.

Beyond establishing the distinctness of the species from each other, however, the amount of pollen sterility and of asynapsis have little or no correlation with the degree of morphological difference between them. The albiflora-tenuifolia-anomala Veitchii triangle illustrates this point. By far the greatest amount of asynapsis occurs in P. tenuifolia × anomala Veitchii; hence on genetic grounds these two species should be considered the most remote from each other, with P. albiflora occupying a position between them. On morphological grounds, however, the three species are more or less equidistant, but P. anomala is somewhat intermediate and nearer to P. albiflora, while P. tenuifolia is particularly remote from the latter species with which it forms a hybrid with almost perfect pairing. The difference between these species are discussed elsewhere (Saunders and Stebbins 1938; Stebbins 1938). There is a somewhat closer correlation in the case of P. triternata Mlokosewitschii, which is quite remote taxonomically from both P. anomala and P. tenuifolia, and forms with both of them hybrids having a very irregular meiosis.

Nevertheless, the pairing affinity of the chromosomes is of little or no value for determining phylogenetic relationships between the species of Paeonia. Equally useless in this respect are both the morphology of the somatic chromosomes and the degree of structural differentiation between them as determined by the "structural hybridity" of their hybrids. Although there is some correlation between structural hybridity and failure of pairing, the former is probably of minor importance as a cause of the latter since in many species and hybrids one or the other is present by itself.

For example, the heterozygosity for many small inversions of P. Delavayi, anomala, and triternata does not reduce the metaphase chiasma frequency of these species at all; in P. albiflora × anomala Veitchii a smaller amount of structural hybridity is associated- with a great reduction in chiasma frequency; while P. albiflora × tenuifolia, in which there is a considerable reduction in chiasma frequency, a marked reduction in pollen fertility, and practically no seed setting, shows the presence of only one inversion.

Asynapsis, therefore, is probably due to genic differences, but these differences are more or less independent of the ones responsible for the observable morphological differences between the species, and have been important chiefly in producing genetic isolation, with the consequent independent mutation of different ancestral stocks separated by such genie factors for sterility and for failure of chromosome pairing. Further evidence tor this interpretation of the significance of asynapsis in the hybrids is that those hybrids with the least irregular meiosis are between species very well isolated from each other in other respects, and vice versa. For instance, P. tenuifolia and P. albiflora have in nature always been separated from each other geographically by thousands of miles; their periods of blooming are a month apart; and they are very difficult species to cross artificially (Saunders and Stebbins 1938); on the other hand P. tenuifolia and P. triternata occur in the same region, bloom at the same time, and are relatively easy to cross; but are isolated by the extremely abnormal meiosis and the complete sterility of the hybrid between them. In Aquilegia (Anderson 1931) specific isolation is produced by geographic barriers alone; in Tradescantia (Anderson and Sax, 1936) it is produced in part by sterility factors independent of chromosome pairing, while in Paeonia it may be produced either by geographic barriers, by different periods of flowering, by incompatibility in mating, by cytogenetic incompatibility which results in asynapsis and pollen sterility, or to a less extent by gross structural differences in the chromosomes. In Paeonia, therefore, as in Aquilegia, and probably most other genera, the evolution of the diploid species has been brought about primarily by the accumulation of a large number of small genetic changes, which for lack of a better understanding of them may be termed gene mutations. These are responsible for changes in external morphology and in the physiological makeup of the species. In addition, genetic changes affecting chromosome behavior at meiosis as well as the development of the gametophyte have occurred, but these have been independent, at least in the frequency of their occurrence, of the mutations responsible for morphological changes. In many cases, 'however, they have produced the isolation necessary for the evolution of two or more different species in the same region. The third type of change—gross alteration of the chromosome structure by means of inversions, reduplications, and translocations—has occurred very frequently, inversions having been much the most frequent. Alterations of this type have been responsible chiefly for the "group-discontinuity" which has formed the various forms or varieties of P. Delavayi, P. anomala, and P. triternata but, unaccompanied by other agents, have not been effective in the actual differentiation of species.

The writer wishes to express his thanks to Dr. A. P. Saunders for supplying the material for this study, and for much technical assistance.


1. The somatic chromosome number of the species of Paeonia is either an = 10 or zn = 20, and the chromosomes are similar in size and morphology in all of the species.

2. Certain abnormalities of meiosis, chiefly the occurrence of fragments and chromatin bridges as a result of crossing over in inverted segments, are found in all of the species, but are most abundant in P. triternata, anomala, and Delavayi, and less so in P. albiflora, Emodi, and suffruticosa.

3. The metaphase chiasma frequency ranges from 1.58-1.93 in the various species, and the terminalization coefficient from 0.41-0.52. There is no correlation between the relative chiasma frequencies at metaphase and the percentage of asynapsis.

4. In one form of P. anomala multivalents resulting from segmental interchange involving a large segment of a chromosome were found.

5. Of the diploid hybrids, P. albiflora × tenuifolia and P. albiflora × anomala Veitchii show few abnormalities of. meiosis except a lower chiasma frequency and a higher percentage of univalents, but in the others a relatively high frequency of bridge-fragment configurations and of multivalent configurations containing triple or quadruple chiasmata occurs. This indicates that these hybrids are heterozygous for a large number of small inversions and for duplications of segments. These are analyzed in one hybrid, P. Delavayi var. lutea× suffruticosa, and several different types of configurations resulting from the presence of inversions and duplications in the same chromosome group are described.

6. The three species which are heterozygous for a relatively large number of inversions are morphologically relatively polymorphic, while the others, except for the existence of horticultural varieties, are less variable morphologically. This is in agreement with the postulate of Darlington on the effect of small inversions on morphological variability.

7. The apparently higher chiasma frequency of Paeonia species and hybrids grown in central New York as compared with the same forms grown in England is ascribed to climatic differences between the two regions.

8. The relatively large number of duplications found in Paeonia species is made particularly evident in hybrids with a low chiasma frequency, due to reduction in competition. Evidence against the interpretation of the multivalents found as resulting from interchange of many small segments is presented.

9. The degree of pairing and the amount of sterility in the various Paeonia hybrids is little or not at all correlated with the degree of morphological similarity between the parent species. Asynapsis and sterility have been important factors in producing isolation between the species, but have taken place more or less independently of the mutations responsible for the morphological differences between them.


Anderson, E., 1931 Internal factors affecting discontinuity between species. Amer. Nat. 65: 144-148.

Anderson, E., and Sax, K., 1936 A cytological monograph of the American species of Tradescantia. Bot. Gaz. 97:433-476.

Belling, J., and Blakeslee, A. F., 1924 The configurations and sizes of the chromosomes in the trivalentsof25-chromosomeDaturas.Proc.Nat.Acad.Sci.lo: 116-120.

Bridges, C. B., 1933 Salivary chromosome maps. J. Hered. 26: 60-64.

1936 The Bar "gene" a duplication. Science 3:210-211.

Catcheside, D. G., 1932 The chromosomes of a new haploid Oenothera. Cytologia4: 68-113.

Cooper, D. C., and Brink, R. A., 1931 Cytological evidence for interchange between non-homologous chromosomes in maize. Proc. Nat. Acad. Sci. 17: 334-338.

Dark, S. 0. S., 1936 Meiosis in diploid and tetraploid Paeonia species. J. Genet. 32:353-372.

Darlington, C. D., 1929 Chromosome behavior and structural hybridity in the Tradescantiae. J. Genet. 21: 207-286.

1936a The analysis of chromosome movements I. Podophyllum versipelle. Cytologia 7: 242- 247.

1936b Crossing over and its mechanical relationships in Chorthippus and Stauroderus. J. Genet. 33:465-500.

1937 Recent advances in cytology. Second edition. 671 pp. London, J. A. Churchill, Ltd.

Dermen, H., 1933 The origin and behavior of the nucleolus in plants. J. Arnold Arb. 14: 282- 323.

Dobzhansky, Th., and Tan, C. C., 1936 Studies on hybrid sterility. III. A comparison of the gene arrangement in two species, Drosopkila pseudoobscura and Drosophila miranda. Z. i. A. V. 72:88-114.

Emsweller, S. L., and Jones, H. A., 1935 Meiosis in Allium fistulosum, Allium cepa, and their hybrid. Hilgardia 9: 277-288.

Heitz, E., 1926 Die Nachweis der Chromosomen. Zeitschr. Bot. 18:625-681.

Hicks, G. C., and Stebbins, G. L. jr., 1934 Meiosis in some species and a hybrid of Paeonia. :Amer. J. Bot. 21: 228-241.

Langlet, 0. F. I., 1928 Beiträge zur Zytologie der Ranunculazeen. Svensk. Bot. Tidskr. 21: 1-17.

Levitzky, G. A., 1931 The karyotype in systematics. Bull. Appl. Bot. PI. Breed. (U.S.S.R.) 27: 220-240.

McClintock, B., 1933 The association of non-homologous parts of chromosomes in the mid- prophase of Zea Mays. Z. Zellf. mikr. Anat. 19:191-237.

Müntzing, A., 1934 Chromosome fragmentation in a Crepis hybrid. Hereditas 19: 284-302.

Philp, J., and Huskins, C. L., 1931 The cytology of Maltkiola incana (L.) R. Br. especially in relation to the inheritance of double flowers. J. Genet. 24:359-404.

Richardson, M. M., 1936 Structural hybridity in Lilium Martagon album× L. Hansonii J. Genet. 32:411-450.

Saunders, A. P., and Stebbins, G. L., jr., 1938 Cytogenetic studies in Paeonia I. The compatibility of the species and the appearance of the hybrids. Genetics 23: 65-82.

Sax, K. 1932 Meiosis and chiasma formation in Paeonia suffruticosa. J. Arnold Arb. 13: 375- 384.

1937a Effect of variations in temperature on nuclear and cell division in Tradescantia. Amer. J. Bot. 24: 218-225.

1937b Chromosome inversions in Paeonia suffruticosa. Cytologia, Fujii Jubil. Vol.: 108-114.

Stebbins, G. L., jr., 1938 Notes on the systematic relationships of the old world species and of some horticultural forms of the genus Paeonia. Univ. Calif. Publ. Bot. (in press).

Stern, C., 1936 Interspecific sterility. Amer. Nat. 70:123-142.

Sturtevant, A. H. and Dobzhansky, Th., 1936 Inversions in the third chromosome of wild races of Drosophila pseudoobscura and their use in the study of the history of the species. Proc. Nat. Acad. Sci. 22:7.

Whitaker, T. W., 1934 An improved technic for the examination of chromosomes in root-tip smears. Stain Techn. 9:107-108.