THE CYTOGENETIC RELATIONSHIPS OF FOUR SPECIES OF CREPIS BY JAMES A. JENKINS University of California Publications in Agricultural Sciences Editors : E. B. Babcock, W. P. Tufts, E. T. Bartholomew Volume 6, No. 13, pp. 369-400, plate 16, 3 figures in text Transmitted March 28, 1938 Issued December 20, 1939 Price, 35 cents University of California Press Berkeley, California Cambridge University Press London, England PRINTED IN THE UNITED STATES OF AMERICA THE CYTOGENETIC RELATIONSHIPS OP FOUR SPECIES OF CREPIS BY JAMES A. JENKINS INTRODUCTION The cytogenetical investigations designed to throw more light on rela- tionships and phylogeny of the various species in Crepis have progressed along two lines : first, an examination of the chromosomes of the various species ; and second, a study of hybrids. For the most part, the study of hybrids has been confined to those between more distantly related spe- cies, which, in the main, have been sterile. Consequently, the emphasis has been upon the cytology of the F x hybrids rather than upon the geneti- cal basis of the differences between the parental species. There are, however, a number of species groups the members of which are closely related morphologically and have a similar karyotype (Bab- cock and Cameron, 1934). From the morphological evidence, these spe- cies have had a common origin and apparently have not diverged very far from one another. The obvious conclusion is that the similar chro- mosome morphology indicates, in such closely related groups in Crepis, a fundamental similarity of the genes and their arrangement in the various chromosome types. The present paper deals with such a closely related group of species in Barkhausia, the most advanced subgenus of Crepis. Three of these species are insular endemics of Madeira and the Canary Islands; the fourth is a widespread species of northern Africa and Europe, which includes one endemic and one introduced subspecies in Madeira. The three endemic species are Crepis divaricata Lowe, C. Noronhaea Babe., 1 and C. canariensis (Sch. Bip.) Babe. 2 The fourth species is C. vesicaria L., and the two subspecies dealt with in this investigation are C. vesicaria taraxacifolia (Thuill.) Thell. and C. vesicaria andryaloides (Lowe) Babe. 3 1 Crepis Noronhaea nom. nov. = Borkhausia (sic) divaricata var. pumila Lowe, Trans. Camb. Phil. Soc, 4:26, 1831; non C. pumila Rydb., Mem. N. Y. Bot. Gard., 1:426, 1900. Named for Sr. A. C. de Noronha, Director, Museu Eegional, Funchal, Madeira, who sent the seed, collected in Porto Santo, from which experimental cul- tures were grown. 2 Crepis canariensis (Sch. Bip.) comb. nov. = C. Lowei var. canariensis Sch. Bip. ex Webb et Berth., Phyt. Canar., 3:461, 1836-1850; BarTchausia hieracioides Lowe ex Webb et Berth., I.e. det. apud Lowe in litt., sed cf. Lowe, Fl. Mad., 1:559, 1868. 3 Crepis vesicaria subsp. andryaloides (Lowe) comb. nov. = C. andryaloides Lowe, Trans. Camb. Phil. Soc, 4:25, 1831; Borkhausia (sic) hieracioides Lowe, op. cit., p. 27, no. 44 ; B. dubia Lowe, I.e., no. 45 ; B. comata Lowe, I.e., nt>. 46 ; C. comata Banks et Sol. ex Lowe, I.e.; Barkhausia hieracioides et dubia (Lowe) DC, Prod., 7:157, 1838; C. hieracioides et dubia F. Schultz, Flora, 23:718, 1840; C. auriculata Sol. ex Lowe, Man. Fl. Mad., 1:556, 1868. [369] subspec A 1 SO OQ 1 ' l bo Is sho 1 - | 1 2 O P< tub 1939] Jenkins: Cytogenetic Relationships of Four Species of Crepis 371 Crepis divaricata is found only on the eastern promontory of Madeira and there it is nearly extinct, owing to overgrazing by goats. C. Noron- haea is known only from Porto Santo Island, which lies to the east of Madeira. Its chromosomes were reported on by Babcock and Cameron (1934) under the name C. pumila, but this name is invalid. C. canariensis occurs on the two easternmost of the Canary Islands, Lanzarote, where it is abundant, and Fuerteventura. Crepis vesicaria andryaloides is also endemic in Madeira, being found only in the mountains along the north coast and occasionally on the steep slopes exposed to the sea down which it is carried by wind or water. It was finally recognized by Lowe (cf . Man. Fl. Mad. under C. hieracioides and C. andryaloides) as a highly variable species with many intergrad- ing forms, some of which were so extreme that he had previously given them specific or varietal names. For sake of brevity it will be referred to in this paper as andryaloides. Crepis vesicaria taraxacifolia is distributed in northwestern Africa and western Europe. It is polymorphic and several of its forms have been given specific names. A form which occurs in Portugal was described as a species (C. intybacea) by Brotero in 1816 and this form seems to have been introduced by the Portuguese into Madeira at an early date. There it was found and described by Lowe in his Manual (1868) as C. laciniata with two varieties, pinnatifida and integrifolia (the latter occurring here and there with the former, but less commonly) . Taraxacifolia is abun- dant around Funchal, the only port on the island, and in the vineyards around Boa Ventura on the north coast. Since it was found by Babcock along the trail above Boa Ventura, it is inferred that it has spread from Funchal to the north coast by this route. But it was not seen at all in the central highlands, so it probably has been carried by man or animals. The important point is that, having arrived on the north coast, it is hybridizing freely with andryaloides where the two come in contact; and it now seems probable that some of Lowe's perplexing forms (dubia, comata, and even the type of andryaloides) were the products of earlier hybridization. The main object of the present investigation was to study the five species or subspecies from as many different points of view as possible and particularly to state their relationships in cytogenetic terms. Acknowledgments This study was begun in 1931 at the suggestion of Professor E. B. Bab- cock, who supplied the material and facilities for the work. The writer wishes to thank Professor Babcock for his interest and guidance through- out the course of the work, and also to thank Professor R. E. Clausen and Dr. G. L. Stebbins, Jr., for their many helpful suggestions. 372 University of California Publications in Agricultural Sciences [Vol. 6 Acknowledgment is made of partial support of these investigations by- grants from the Carnegie Institution of Washington and the Rockefeller Foundation ; also to the Works Progress Administration for the services of a typist. MATERIALS AND METHODS The cultures used in the investigation were : (1) One collection of C. divaricata from the eastern promontory of Madeira, Promontory of San Lorenzo, Ilha de Cevada. (2) One collection of C. Noronhaea from seed collected in Porto Santo and grown for one generation in the museum garden at Funchal. (3) One collection of C. ccmariensis from Lanzarote Island in the Canary group. (4) One collection of andryaloides from the mouth of the Ribeira do Inferno on the north coast of Madeira ; the plants or the seeds had ap- parently been washed or blown down from the highlands. (5) Three cultures of taraxacifolia collected in the vicinity of Funchal, on the south side of Madeira. Collections of (1), (4), and (5) were made by Babcock in 1930, and the other two were sent to him at Berkeley in 1931. Crosses were made between all five entities in May and June, 1933, and repeated in the following year ; there was no obvious difference between the results in the two years. The method used was a slight modification of that described by Collins (1922). All root tips were fixed in Miintzing's (1933) modification of Nava- shin's fixative, section at 9/x, and stained either with Haidenhain's iron- alum haematoxylin or Smith's (1934) modification of crystal violet. All meiotic figures were studies in acetocarmine, McClintock's (1929) tech- nique being used. Pollen grains were mounted on a slide in a drop of acid fuchsin dissolved in lactic phenol, a medium which obviated the necessity of sealing the mounts. All the pollen counts were made after the plants had been in bloom for about two weeks. MORPHOLOGY OF THE PARENT SPECIES It is not to be expected that seeds collected from a few plants in the wild would give plants showing the total variability of any one species. How- ever, a comparison with specimens collected in the field showed that the cultures were a representative sample of the total variability of the species. The measurements can be taken as a fair approximation of what is characteristic of these species, and are adequate to establish the rela- tive differences between the species. The species were quite variable in themselves, as would be expected in self -incompatible ("self -sterile") plants. Canariensis and divaricata were more uniform than the other two, and it is interesting to note that 1939] Jenkins : Cytogenetic Relationships of Four Species of Crepis 373 the first two species are the most restricted in their range, and may be considered as relic species. In classifying the plants, there never was the slightest doubt of the species or subspecies to which they belonged. Each fluctuated about a distinct center of variability, and although there was frequently some overlapping in particular characters, yet in the sum total of characters the five entities were quite sharply delimited. The general appearance of all five is illustrated in plate 16, figures 7 to 12. The differences between the species were expressed in all parts of the plants, principally in small differences of size and shape. Some of the more distinct differences are summarized in table 1, for purposes of illustration. In addition to quantitative character differences, there were a number of discontinuous variations peculiar to each of the species or subspecies, as, for example, a purple tip on the ligules of canariensis. All these latter characters were of no obvious adaptive significance ; that is, they could be classified as nonessential. It is clear that the species can only be distinguished by means of a combination of characters. It was not possible, on the basis of external morphology of the plants grown in cultures, to divide the species into groups of a higher category. HYBRIDIZATION Dobzhansky (1937, p. 231) has used the expression incongruity of the parental forms for "mechanisms [including geographic and ecological isolation] which prevent the production of hybrid zygotes, or engender such disturbances in the development that no hybrids reach the repro- ductive stage." Conversely, we may define the congruity (or genetic "compatibility") of two forms as their ability to hybridize and the ¥ t hybrids to form viable gametes (that is, gametes capable of producing vigorous zygotes). There is no simple way of measuring the congruity or expressing it by means of a single numerical value. Two forms may be so incongruous that they fail to produce any hybrid seed, owing to their incompatible reaction systems ; or, on the other hand, they may be fully congruous, as is frequently true of varietal hybrids. Between these two extremes the incongruity may show up at various stages : the F x zygotes may be so weak that they fail to germinate, or if they germinate they may die before maturity; the F x plants may be quite as vigorous as the parental forms but closely approach complete sterility, for example, Nicotiana sylvestris x N. tomentosa, Clausen (1928), Primula verticil- lata x floribunda, Newton and Pellew (1929) ; finally, there are all de- grees of fertility of the ¥ x hybrids, and all degrees of vigor of the F 2 zygotes. Consequently, an estimation of the percentage of viable gametes, in practice, entails : (1) a knowledge of the percentage of seed-setting on 374 University of California Publications in Agricultural Sciences [Vol. 6 O c i O o <1 w f DC % fe H (t Ph fe h •« p u CO T-4 EG <^ h-3 BE 9 PQ u s < B d l e < £ s 5 u >* lj g t £ c «J M Ph 5 c 2 ° § CD >> -w 0) 03 CD QQ C3 P tf 2 t >00(ON d irf f-4 8 i— i o *** d E "5 -° W ^ -M --5 CD *H a § o d %>> ~ > aT !S o3 «3 *» 3 CD 02 « P P 0Q O © *! « CO O tO O > p 3 o3 ft-r-T g I -s bo ft-** P c a; R< & n 0) a >i aT ; — | tf o a flil -M 3 CO 02 CQ * fl 2 ». 13.0 c: 6.0-1 3.2 2.3-4 03 CD •43 a fe to u ^ 3 S 03 02 n I -2 >j 02 *rl CD ?3s ^ ft 02 «2 P "a § •a « 02 02 02 nfl cu 5^ a o a a, cq 02 02 O a a -s, a o 5 P ft ^H J rd i" d T3 d rd "H o3 03 02—1 rj . = O Sm d d d r#T CD O H . !» 4s be 02 _- >^ a CD S. 03 02 g fl § g .d •+= o -p «s -p 0Q >> $ j>> d — GO 03 ar or ; rare ing ba ricate at the basal i— i o3 03 03 G, •— ! 03 7-2 ««^T IO 0. 03 jS jS o3 ^4 03 b i—i 03 ^ 03 _ "* "° pCI fe 03 T-H ^ bC fn S o -3 CN — ! fl 03 c3 gH qj r— ] f_i 03 fl ."2 1-3 S P O CO T CO io o^ *o ooNowdd > £ > Sh sually cordate, quently oblanc late; usually w clasping base o3 -77 -50 .5 .9-3.6 Usually regular] branched upw with foliage e ly up the stem CO f >>^ r-H Q o3 b 3 03 co *-* ways glandul hairs, usually blackish setae P N On* (N H o p < o o£ TH S3 date, s lance blance lly wi •ase larly pwar e eve em CO 3 P 3 bfl -+? "•+3 c3 03 S *"d ally cor metime te, or o te; usua asping b r- ^ "P O O 1 ally reg anched th folia up the s 1 3) E? 02 03 "3 § £ *:*3 P 00 «D (O N CO lO IO CO rH © ^ M N h g.3 ** P d 02 I S S. o3 >» tBi . >> linear or late, rare eolate 03 S -d 2 11 f: glandular frequentl sh setae >5 O fl t^ sually basew liage * j>? 02 mM £3 C3 03 3 g 3 1.5 o »o CO 1 CO -* O "5 r-J 03 03 b 3 03 02 H way hair; blac p CO 00 »0 CO rH © INHNH P P < 1 a, CO oa "§£ 03 S *0 o3 3 4 s ef 4 rH 02 03 s a .§•■§ a ° g 02 3 £ 03 03 § ^ C O — ' o cm. -37 cm. -40 .0 .8-1.3 sually branc base with b liage 1 "So DD 02 g 'S © 1> O IO rH o CO CM CO CN P o 5 •-^ «r> 02 03 , i Leaves of the middle ine leaves e) ) e) d (mdn. (range ^FLORESCENCE Number of heads on the ultimate branch r3 « O 03 11 rEM Height (mdn (rang Spread (mdn, (rang Height/sprea b/0 3 O q) 03 ft 2 ^ ci a 03 03 AULINE Shape caul C3 q o3 03 03 _D rH O m ►H 376 University of California Publications in Agricultural Sciences [Vol. 6 b i ^ o d TJ 3 co d e d o 1 I.& o l-H ^ a H © o ff « rSS (3 © o •* co 'tf' CM i-H °? CO rH CM 9®. CO CM CM 00 © IO o a S * S 03 (B O CO *C CM Oi »C I— ( T— 1 »0 CM CM CM CM d CD r-5 CM r-i r ~' *~ ' 1— 1 T— * &T a O » a .13 -v 2 CD pd g >> ^ CD "3 •1 o © 00 CO 111 CM s CO CO e 9* csi ^ ^ T-H ^d r2 * 6X),0 ^ r3 CM 9 9 CO o © © i-h CM 1 CO CO 9^ 9 9 00 "*' 1— 1 1— I CM CM <■* © CM i-i Tt< i-H rH rH CD CD .2 CD S -5 d 9 d h ° 5 -a £ e o O r-l O « I— I T— I CO r-l CO > o H-> CD ^O .i-i OJ CQ ^ — i £ o3 03 o 00 co © d a c3 9 r— 1 1—1 00 CO 1—1 § O CD 03 -rj a <^ CO CM i-i Oi 00 *o i—i i— i TJH O CM > o 03 ■*=• Q CD Si rd 3 1 9 CD rJ rn O o o3 ^d. -9 to i § 02 1939] Jenkins: Cytogenetic Relationships of Four Species of Crepis 377 a 5 d 5 cu o H t-i ? X2 cu »o »o d o o o o O a o3 d o cu CO O > 3 g CU s --* > o IO o d d >> IO o t>. Oh IO o eo eo d d 00 co 4< oo od 1 1 1 »c o io id J 00 o t- ( rH O IO rjl CO -P ■ d 'S E CU o CU h H b£ ^2 £ o oo CU t^. IO CM +3 >d d >> 00 « up ,4 bO 00 © b- CO i— 1 o o 73 d IO CO CO o IO IO 53 Tti <<*' d d 1 OS t> CM CM l— 1 i-H 03 > d id r-t i— 1 "*' CO > a d -? a s d > a * o o o ill o o IO o i CO S *: a a 9 a n a 9 (M O CO IO a o o a «p a . CU CO CO d d o o 1— 1 T-H d d 02 ■"" l 1 IO Tt< r-i d CO CM Ft CD § ^ 1 » a s -*=> w w O'aT O'cu -d d bfi d bfl OQ +3 T3 d . ^ d / ~ s cu CU /-N be ^* cu bfi >~v / ~" N O CU a 2 a 2 C W) o3 o bC S d bo d be "W ' V ' s * s_^ "SI 1 a, G Sh "9 S a 2 rd bfi -3 %■$ o3 o3 bfl d cu u o "3 1 -*» s- bfi 1 -4-3 « ^ -O 02 J=l cu 02 2 fl — cu 2 c ^3 CU p o3 cu i © go bC ^ 3 a O ^ . ^ O 3 1 1— 1 cu o3 CU o3 "< << h- 1 ^ pq Ph 378 University of California Publications in Agricultural Sciences [Vol. 6 the F ± plants, and (2) the raising of an F 2 population in order to de- termine whether or not the gametes are capable of producing vigorous zygotes. For a complete conception of the congruity there are also neces- sary: (3) an estimate of the percentage of hybrid seed obtained from crossing the species, (4) a record of the germination of that seed, and (5) a knowledge of the percentage of ¥ t plants that grew to maturity. The data on congruity for the ten possible hybrids are given in table 2. It might be well to discuss at some length each of the measures. TABLE 2 Percentage of Hybrid Seed-setting, Percentage of Germination of the Re- sulting Seeds, Percentage of Morphologically Good Pollen on the Fi Plants, Percentage of Open Fertility of the Fi Plants, and Germination of the F 2 Seeds, of the Ten Possible Hybrids Cross Hybrid seed-setting Germination of hybrid seed "Good" pollen on Fi Fi Fertility Germination of F2 seed 1 * 5 4 5 T-A* N-A 54 (l)f 44(2) 42(3) 42 (4) 38 (5) 36 (6) 36 (7) 24(8) 16(9) 10 (10) 71(5) 86(1) 76 (3) 62(7) 43 (10) 68 (6) 48(9) 75(4) 53 (8) 86 (2) 81(1) 46 (5) 44(6) 10 (10) 19(9) 58 (2) 56 (3) 33 (8) 35(7) 53(4) FairJ (3) Fair (4) Poor (8) Good (1) Very poor (10) Good (2) Fair (5) Fair (6) Poor (9) Fair (7) 35 88 N-T 20 D-A 47 D-N 54 D-T 67 D-C 43 A-C N-C T-C 50 Average 34 67 44 Fair 44 * T, vesicaria subsp. taraxacifolia; A, vesicaria subsp. andryaloides; N, Noronhaea; D, divaricata; C, canariensis. T-A includes both combinations, namely, taraxacifolia 9 X andryaloides c? and andryaloides 9 X taraxacifolia d". The other nine combinations also include the reciprocals. t The numbers in parentheses refer to the relative order of the observation in magnitude array, begin- ning with the highest. t Excellent, 76-100 per cent; good, 51-75; fair, 26-50; poor, 3-25; very poor, 1-2. The percentage of hybrid seed-setting (table 2, column 1) is merely the ratio, expressed in percentage, of seeds obtained to the number of florets emasculated. Approximately one hundred florets were emascu- lated in each cross, with conditions kept as nearly constant as possible. No difference was noted in reciprocal crosses. It may be significant that when the percentages are arranged in magnitude array, crosses between taraxacifolia and andryaloides stand at the top ; these two subspecies are hybridizing in nature and are producing many intermediate progeny. There is practically no difference between the various divaricata and Noronhaea crosses, and finally, the crosses involving canariensis are all 1939] Jenkins : Cytogenetic Relationships of Four Species of Crepis 379 at the bottom of the list. The numbers and the samplings of the native populations of the species are not sufficient to establish any precise de- gree of relationship on this basis. The percentage of germination (table 2, cols. 2 and 5) of the F x and F 2 seed was very little different from that of the parents ; the average germi- nation for the four species, including the two subspecies of vesicaria, was 45 per cent — varying from 25 per cent for canariensis up to 86 per cent for andryaloides. There was no obvious difference in the reciprocal crosses, and the average for the various crosses did not differ markedly from the average for all the crosses. Practically all the ¥ 1 seeds that ger- minated grew into plants which lived to maturity ; the few that died in the course of the experiment did so from causes far removed from in- harmonious combinations of genes. Fertility (table 2, col. 4). — A study of the fertility of the hybrids is complicated by the situation in the pure species, where, with the excep- tion of subsp. taraxacifolia, the parents are completely or almost com- pletely self -incompatible. However, an abundance of seed was obtained when the heads of sister plants were rubbed; also, the fertility of the open-pollinated plants was high, particularly when exposed to the visi- tations of insects. An attempt was made with canariensis to see whether there were definite intrasterile-interfertile groups within the species, as East (1932) and others obtained in Nicotiana and other genera. The results did not conform to a simple scheme. If there was a single series of fer- tility allelomorphs, their clear-cut expression was modified by other factors, either modifying genes or fluctuations of the environment. It was difficult to measure the degree of fertility accurately ; so the percentage of seed-setting was estimated as belonging to four groups : exceUent (76-100), good (51-75), fair (26-50), and poor (1-25) ; the last class was subdivided into a classification of very poor (1-2) . The open seed-setting on the F x hybrids was markedly poorer than that of the parents, where the open seed-setting was usually 100 per cent, or at least in the excellent class. Also there was a decrease in the average amount of morphologically good pollen (table 2, col. 3) ; although, even in the parents, which usually had from 80 to 100 per cent of apparently good pollen, sometimes there was as low as 20 per cent, in spite of the fact that the flowers were selected when the plants were at the height of their blooming season (that is, after they had been in flower for about two weeks). As a consequence of this variability, the amount of good pollen could not be used as an index of fertility. If there was some simple relationship existing between the amount of apparently good pollen and the seed-setting, it would have been very difficult to establish without an extensive statistical study. 380 University of California Publications in Agricultural Sciences [Vol. 6 Under the most favorable conditions of seed-setting, open-pollinated heads sometimes appeared on some F 1 plants in which all the possible embryo sacs had developed. Consequently, the normal procedure in gonogenesis being assumed, there is every indication that under certain circumstances all the female gametes in these particular hybrid plants are capable of functioning. Conclusions from hybridization. — (1) All five entities hybridize very readily. (2) The crossability of the various species and subspecies, taken in pairs, was about the same, though there was a suggestion that cana- riensis was less congruous ("compatible") than the others. (3) The hybrid seeds germinated as well as those of the parents, with little or no difference between the individual crosses. (4) The hybrids,- on the whole, were less fertile than the parents, and there was less morphologically good pollen, but the correlation was not obvious. (5) Hybrids involving taraxaci folia were slightly more fertile than the others. (6) Under cer- tain circumstances, all the gametes in some hybrids were capable of functioning. CYTOLOGY All five species and subspecies had eight chromosomes at the mitotic metaphase, confirming the observations of Babcock and Cameron (1934) . Although several plates from each were carefully measured, there -& it/ w a b Fig. 2. Somatic metaphase from root-tip cells of : a, Crepis vesicaria subsp. taraxacifolia; b, Crepis canariensis x C. vesicaria subsp. andry- aloides F,. x 2500. All parts of this figure were drawn with the aid of a camera lucida at a magnification of 3750 times, from permanent preparations, and reduced one-third in reproduction. seemed to be no constant difference either in morphology or in total length of the various chromosome types. The differences observed be- tween them were small and could easily be explained on the basis of variations in fixation, age of the cells, twists, etc. The somatic chromo- somes of taraxacifolia are illustrated in text figure 2, a, which would serve equally well for any of the parents (see also Babcock and Cam- eron, 1934). 1939] Jenkins : Cytogenetic Relationships of Four Species of Crepis 381 In the first-generation hybrids the somatic metaphase chromosomes appeared to be the same as in the parents. In size, staining capacity, and morphology the maternal and paternal elements could not be distin- guished. Both of the D chromosomes in the hybrids had a satellite, that is, there was no indication of amphiplasty as reported by Navashin (1928) and Hollingshead (1930) in more distant species hybrids in Crepis. Text figure 2, b shows a somatic plate of the F x hybrid canaden- sis x andryaloides, which is essentially similar to the parents. Meiotic chromosomes. — Both in the hybrids and in the parents, at first meiotic metaphase four bivalent chromosomes were regularly seen, all of which disjoined in a normal fashion (see text fig. 3, b and c). Also, the second meiotic division was normal, and comparable, in all respects, in the parents and in the hybrids. Fig. 3. Meiosis in Crepis Nororihaea x C. canariensis Y t : a, diplotene showing four paired elements; b, metaphase, showing four typical bivalent chromo- somes ; c, anaphase, showing four chromosomes passing to each pole, x 1700. All parts of this figure were drawn with the aid of a camera lucida at a mag- nification of 3400 times, from acetocarmine preparations which were squashed by light pressure, and reduced one-half in reproduction. The parents frequently have a single nonterminal chiasma at early diakinesis (see text fig. 3, a) , the minimum required to maintain pairing. Earlier stages were not examined in detail in regard to chiasma fre- quency, as it is almost impossible to distinguish between twists and chiasmata in acetocarmine preparations. At late diakinesis there was usually one, sometimes two chiasmata, and rarely three. This is rather surprising in view of the great length of the Crepis chromosomes. An- other curious fact is that there was little terminalization until late diakinesis or early metaphase. Recently, Darlington (1931) has regarded the relative frequency of chiasmata formed in the first meiotic division of the parents and the hybrids between them as a measure of the genetic homology of the chro- mosomes. The work of McClintock (1933) on nonhomologous association and Beadle (1933) on asynaptic maize, and of Kihara (1929) and others 382 University of California Publications in Agricultural Sciences [Vol. 6 on the influence of temperature, would throw considerable doubt on this measure of relationship. Nevertheless, if this criterion has any value whatever, these species are very closely related. The cytological evidence strongly indicates that the five entities have a similar arrangement of genes in the various chromosome types. In other words, there have been no large duplications, translocations, or other rearrangements that in any way interfere with normal meiosis. HYBRID SEGREGATION The five entities were crossed in all possible ways, making ten different hybrid combinations, each including the reciprocal hybrid. F 2 cultures were grown from all except two combinations, namely, taraxacifolia- canariensis and Noronhaea-canariensis, in which no self ed or sibbed seed was obtained. The behavior in all the hybrids was remarkably similar and of a type frequent in crosses between closely related forms. For the sake of brevity, the general behavior of the hybrids will be described and illustrated by data from only one hybrid combination, namely, taraxaci- folia-divaricata. The Fi-generation plants were variable, but no more so than the par- ents. The character differences were manifested in either of two ways : (1) they were more or less intermediate between the parental averages, or (2) the influence of one parent was more pronounced. The greater num- ber of the characters were of the intermediate type, which included prac- tically all the distinctive species differences ; for example, height, flower size, achene size and color, and beak length. Each hybrid was distinctive, and it was easy to determine the parental species by an inspection of the hybrid; thus no species was entirely dominant over any other species. The characters that showed dominance were, for the most part, those which had no apparent adaptive significance ; for example, anthocyanin patterns, pubescence, color patterns (see pi. 16, figs. 1-6 for the appear- ance of the rosettes of another series of hybrids) . The Fi-generation plants were just as vigorous as the parents, but no more so. This lack of hybrid vigor is probably to be explained by the consistent cross-pollination of the wild species, which makes them highly heterozygous. In the F 2 , most of the characters followed the blending type of inheri- tance, even most of those in category 2 above. This shows that the specific and subspecific complexes were made up of a large number of genes, and that most of the characters, if not all of them, were influenced directly by many genes. A few characters, those determined by a single gene differential, showed dominance in the F 2 . The characteristics of the blending inheritance in these species hy- brids, as illustrated in tables 3-9, may be roughly summarized as follows : 1939] Jenkins : Cytogenetic Relationships of Four Species of Crepis 383 (1) In both F x and F 2 there was a continuous range of expression of any one character difference, with the majority of the individuals inter- mediate. The mean of the F 2 population was similar to that of the F 1# (2) The range of the F 2 variation was about that of the parental ex- tremes, with no well-marked occurrences of transgressive inheritance. This lack may have been due to the small numbers, as the F 2 population in any one cross did not exceed one hundred individuals ; on the other hand, it may have been due to the lack of dominance in the various gene series. (3) There was no recovery of types corresponding to the parents in all or most of the characters. In many hybrids, particularly where there are chromosomal difficulties, the intermediate types are eliminated, that is, they are unfavorable combinations. The fact that no parental types were recovered indicates (a) that the intermediate combinations, or at least a large number of them, were able to survive, and (b) that there were a great many genie differences between the two parents. (4) No new characters appeared in either F x or F 2 , which would indicate that the gene systems in all the species and subspecies were essentially the same. The new combinations in the hybrids merely altered the ex- pression of the existing characters. It is now known that multiple genes can bring about such a continu- ously varying F 2 , that is, several genes, each with a small increment, affecting the same character. It is very probable that species that are highly heterozygous will have a blending type of inheritance, or at least a variable expression, for most if not all characters. It is only in rela- tively homozygous lines that it would be possible to obtain a sufficient number of clear-cut segregations to reveal the genetic basis of such small character differences as those found in these species. It would be a long and tedious piece of work to put these interspecific differences on a Mendelian basis, and the task would be greatly complicated by the pres- ence of self -incompatibility. Data were taken on a number of more or less clear-cut character differ- ences between the species and subspecies. With the small number of in- dividuals, and in the limited time, it was not possible to work out the factorial bases of these characters beyond saying that they are gene determined but do not conform to a simple Mendelian scheme. The re- sults in one are more or less typical of them all, and for the sake of brevity only one will be illustrated. Both taraxacifolia and Noronhaea have a conspicuous red stripe on the dorsal surface of the outer row of ligules, which was somewhat vari- able in its expression. The other three, namely, canariensis, divaricata, and andryaloides, had no stripe, and it did not appear in any of the six possible hybrids between them. In all the crosses in which one parent had a stripe and the other had 384 University of California Publications in Agricultural Sciences [Vol. 6 < CO a g a ft Q H K <5 ft pi 02 m p a a a H *! gn s « g P a B B I ft 9 H S >-h o a P w 3 5 £ o O CM i>- ^ CO CO CM 1— t CD cm T-H TH tH CM CM CO (M CM CN H CO 00 CO 00 rH IC T-H t~ >OHNN r— 1 to CO *0 CM "ti 1— 1 «o CO N CO CO i— i T»* CM t>- rjH CM r— I eo CO N CO 00 o ,M lOHiOO T— 1 o CO «3 t* * i—i eo 02 X) X >> £ 13 c3 o3 03 CD o3 *o -M uZ o3 '13 a 8 2 CO .5 c3 3 1 •5i 3 - . P h fe U- o Q < a Eh g ** a CO £ a < PQ <-} w ft o H ft N S "tf O H § s is 8 H S w ft 8 a ft ft co o a Bfi ■-i g a ft ft Q fe g o < o ~ P < ft M *< d K 1 1 1 ft H fe «e< O « CO g S5 ft u |Z| B g a ^ O O CM l> o rt* CO CO IN ft *-* T-H o CM 00 "5 CM »0 CD H t-H CM U3 CM CO *C -«*< O O »o CO l-H 00 t^ CM Ci CO i-H OS CM CO CO OS ^ -^ l-H 64 CM CO -* CM o OS CO *0> OS 00 eo ■* ti iO CN T-H CO CO *0 CM i-i OS eo io eo >o eo CO CO t-H »C O eo <* "* es ^ * CO T-H eM E» T3 X! : 03 >> A o3 *o T3 fi o3 03 -g .2 o3 s H X o o3 c3 > £3 0. •H 03 H < OQ QHPh £- 1939] JenJcins : Cytogenetic Relationships of Four Species of Crepis 385 <; a 8 ■ 03 H t 3 pq « OD 9 s g « g w a a M B 1 H Q JL I 8 a w e8 CO • eo CM . e>» ■* h if) h rl< e» co M (N N CO o CO T* CM • oo ^ 1-1 so T3 X! >> J2 TJ 53 C J2 s I 8"S 03 o3 > *-< P E- fn pt & OQ g a 2 B § ft 3 w 8 I > B d« « 8 3 ^ O ^ _ o h S O M (M CM CO O H T-H 00 «o CO o ^ US CO © i-i o o lO ta io § N NO00 •*< 00 CM -^ OS o N W D »C T-H •O CO CM »C i-l 1—1 © i-i CM OS CO «o CO OS CO CO © "* oo • T-H T— 1 T3 c £> >> A T3 a cj s 0) o3 a GG o3 IS jl c3 c3 > *3 ■Si o3 - . 1 P r- I* fe 386 University of California Publications in Agricultural Sciences [Vol. 6 H H H Q fe * pq Sg < 9 Si H a w h g s & Ph h 02 O o o H £ 2 ™ 3 p W § O P5 ^N tf 1 o CM £ o CO § H 1— 1 «* 1— 1 CO e* eo o 1— 1 1—1 CO 00 CM ■ 0 1-1 •«*< H rt MOO ~ > ca d 3 03 o .2 'o s !l § S > fH •h d - , c E- fr fc g * o <1 GO ft o w G S3 H ft O 03 o fe 3 . £ MMHO) o © lO o> o CM rj< OS *} i-i CM i-i 00 00 o OS CO "* oo i—i «o H lO CD O) t~ - i-l 1-H ^ to »o CM CO U5 © T* us T3 ^ >> M T3 d a 8 : os © c3 o a DQ 3* . — i 03 03 03 > b 1939] Jenkins : Cytogenetic Relationships of Four Species of Crepis 387 < < a 3 W r} fa s i I s £2 M > GO © '3 01 a CO si ■+= sfi c T > 3 .2 "c 1 s K P^ fe 388 University of California Publications in Agricultural Sciences [Vol. 6 none, the F x segregated 59 plants with a stripe to 14 without, and the F 2 segregated 99 plants with a stripe to 43 without. In the F 2 the deviation from a 3 : 1 ratio is not quite significant, since a deviation as large as this would be expected in slightly more than one out of ten, from chance alone. Progeny were obtained from 5 of the 14 plants in F x that did not have a stripe. Four of the 5 segregated F 2 plants with a stripe, indicating that the F x plants carried the red-stripe gene though it was not expressed. The fifth F x plant did not have a stripe, nor did any of its 10 progeny. There is some evidence that the taraxacifolia plant used in the original cross was heterozygous, as 2 of its progeny out of a total of 9 were with- out a stripe. If this latter progeny is excluded from the total for the F 2 , the ratio is 99 : 33, a perfect agreement with a 3 : 1 ratio. It is most probable that the F x parents of those F 2 progenies that seg- regated red-striped plants although their F x parents had none, did not have the proper genetic background for the gene to express itself, but that the F 2 recombinations did supply the favorable background ; that is, the presence of this character not only requires the presence of the gene in the dominant condition, but also requires a definite genie back- ground, much as was observed of Harland's (1935) crinkled dwarf in Gossypium, which showed different expressions of the character, de- pending upon the particular genie milieu in which it had to develop. Further evidence for this theory is the wide range of expression of the character in F 1? where some plants had such a slight expression that they were difficult to distinguish from normal, and an even wider range in F 2 , where some of the latter plants were so intense in their expression that the color showed through on the upper side of the ligule. In crosses between taraxacifolia and Noronhaea, both of which had a stripe, all the F x showed the stripe ; and out of 14 F 2 plants 3 had no stripe and all these occurred in the progeny of the same F t plant. This lack of the red stripe may be due (1) to a slight expression which was overlooked in the classification, or (2) to the wrong background for the visible expression of the gene. In spite of the fact that the numbers are small and the evidence some- what conflicting, it appears that the stripe may be referred to a single dominant gene which behaves in a normal Mendelian manner ; though the expression of the character is dependent, to an appreciable degree, upon modifying genes. It is probable, for instance, that if the red-stripe gene were introduced into the divaricata background by repeated back- crosses, it would segregate in a normal Mendelian fashion but might not show the same dominance relationships or expression that it shows in taraxacifolia. There is probably enough difference in the genie back- ground to suggest that the expression of this gene would be modified in the new background. 1939] Jenkins : Cytogenetic Relationships of Four Species of Crepis 389 DISCUSSION These five entities can be readily distinguished from one another by observation ; the morphological differences between them are expressed as many small differences in size, color, and shape, affecting almost all plant organs. There are a few outstanding qualitative differences be- tween them, though these are, taxonomically, of a minor nature ; for example, the purple tip on the ligules of canariensis, the ligules wither- ing white in andryaloides. Two of the species, canariensis on Lanzarote Island of the Canary group and Noronhaea on Porto Santo. Island, are geographically well isolated from the other two on Madeira. Of the Madeiran species, divari- cata and vesicaria andryaloides occupy different ecological stations, the former being found only on Promontory San Lorenzo, which is an island at high tide, and the latter in the northern highlands of Madeira. In spite of the fact that these species have occupied these regions for a long time, there is no evidence that they have ever hybridized in nature. Vesicaria taraxacifolia, on the other hand, is undoubtedly of more recent advent on the island, probably having been introduced by the early settlers around Funchal, on the south coast. It is well established there and has spread to the north side of the island, particularly around the vineyards; furthermore, it is an "aggressive" weedy type and is spread- ing. In the north-central part of Madeira, where taraxacifolia and an- dryaloides have come into contact, numerous intermediate forms were collected and observed by Babcock in 1930. These are undoubtedly nat- ural hybrids. But there was no evidence that taraxacifolia had spread to the eastern end of the island and hybridized with divaricata. The fact that all five entities have the same karyotype and that the chromosomes apparently mate up chromomere for chromomere in the meiotic prophase of the hybrid, with no subsequent irregularity, would indicate that they have essentially the same genie arrangement. It is difficult to prove that all five have the same number of genes, though the evidence points to this conclusion. There may be minute rearrangements and even lack of some particular genes in some of the species ; however, if this is so, it is not reflected either in the pairing of the hybrids or in a constant elimination of large proportions of gametes. The hybridization experiments demonstrate that the species and sub- species are able to exchange genes readily. The hybrids are produced without difficulty and show a fair measure of fertility ; only a compara- tively small proportion of the hybrid recombinations are incapable of surviving. The hybrid cultures gave every evidence of being as vigorous as the parental species, and were quite as vigorous as the progeny of some natural hybrid derivatives of taraxacifolia and andryaloides. 390 University of California Publications in Agricultural Sciences [Vol. 6 Al l the evidence is consistent with the view that the five entities have a great many gene differences, though the number must remain prob- lematical. The most probable assumption is that all the species and subspecies possess the same number of genes, but that there are many different combinations of alleles. In any one species there must be a considerable proportion of heterozygous genes, and since the range of variability in the F 2 for most characters is roughly twice that of the parents, there must be a higher proportion of heterozygous genes in the hybrids between the species and subspecies. The prevalent type of F 2 segregation for any single character can be satisfactorily explained on the basis of four or five genes with incom- plete dominance. But it is not likely that even a probable estimate of the total number of genie differences could be obtained by multiplying the number of character differences by four or five, as we know that many genes, if not all, may influence several characters. It is quite possible that a very few genes influencing growth rates at slightly different periods of development could produce a large array of character com- binations. It would require a very long and extensive breeding program to establish with certainty the number of genes influencing any one char- acter difference. If the total number of basic genes available to these species is desig- nated as a, b,c,d,... n, and it is assumed that each gene may have sev- eral alleles, which may be designated as a 1 , a 2 , a 3 , . . . & k ; & 1 , & 2 , b 3 , . . . & k , etc., the gene population of each species would contain all n genes, but many, if not a majority, would be represented by two or more alleles clustered around what "Wright (1932) calls an "adaptive peak." Two different specific combinations coming together in a hybrid do not upset the gene balance, but many of their segregation products (recom- binations) are not equally viable : some are so unbalanced as to produce lethal or very weak combinations ; in other words, they fall in the "adap- tive valleys." The genes in one species may be transferred to another, and although not all the hybrid combinations are equally successful and many are eliminated, it is possible that new and still more harmonious combina- tions might be built up. Some offspring might even be adapted to new habitats and start an independent line which in time might become eco- logically distinct. All the evidence indicates that the isolating mechanisms that have been built up between these species are due to gene incompatibilities, which undoubtedly have arisen by mutation over a long period of time, and there is no indication whatever of any chromosomal rearrangements. This is rather surprising in view of the ease with which quite radical rearrangements were obtained by Navashin and Gerassimova (1936) 1939] Jenkins : Cytogenetic Relationships of Four Species of Crepis 391 through the ageing of Crepis seed, which would seem to be a natural process. In Crepis there are several groups of morphologically closely related species with a similar karyotype (Babcock and Cameron, 1934) . It is nat- ural to assume that some of the differences in chromosome morphology between the groups are due to chromosomal rearrangements. Miintzing (1934) found evidence of one inversion between C. divaricata and C. dioscoridis, the former from subgenus Barkhausia and the latter from subgenus Eucrepis. There have been two additional instances (unpub- lished) : C. oporinoides x patula, two distantly related species of the sub- genus Eucrepis; and C. canariensis x oporinoides, the former of sub- genus Barkhausia and the latter of Eucrepis. In the latter two hybrids there were extensive translocations, but there were also very obvious dif- ferences in the karyotypes, which would lead one to suspect that there had been translocations. It is also natural to assume that the species within any one group have essentially the same arrangement of genes, particularly in view of the fact that the differences in chromosome number and morphology between groups are quite striking. Besides the present group, only one other closely related group of species with a similar karyotype has been in- vestigated. Cave (1936) studied four such species: Crepis foetida, C. commutata, C. eritrieensis, and C. Thomsonii, with essentially the same result, namely, that there was no evidence of rearrangements. Conse- quently, the assumption of a similar arrangement of genes is borne out in these two investigations. Whether or not it is true in the whole genus will have to be determined by further research. Nevertheless, these two instances materially strengthen the evidence for the assumption that similar chromosome morphology, of closely related species within this genus, indicates structural similarity; accordingly, karyotype studies are valuable in determining genetic relationships. These species and their close relatives have had a very complex evolu- tionary history, involving repeated isolations and hybridizations ; so that it is impossible, with the available evidence, to trace their phylogeny in any detail. Since there are few qualitative variations differentiating the five species and subspecies, and since almost all these variations are pres- ent in at least two of them, it is probable that the majority of the specific differences were present in the ancestral stock. Nevertheless, some char- acter differences have undoubtedly arisen since the separations, for ex- ample, the purple tips of the ligules in canariensis, and it is quite likely that the quantitative differences have been emphasized in the passage of time. The uniformity of the environment on the islands would tend to- ward uniformity and less evolution of the species, once they became established in a favorable habitat; furthermore, the relatively small 392 University of California Publications in Agricultural Sciences [Vol. 6 numbers characteristic of most island species would also automatically tend toward still more uniformity (Wright, 1932). Since these entities could not be arranged into groups of a higher cate- gory, they must have migrated to their present situations at different times, or must have come from forms which had already differentiated ; either would involve separate migrations. This, with the fact that the nearest relatives of canariensis, Noronhaea, divaricata, and andryaloides are C. Fontiana, from northwest Africa, and C. Bourgeauii, from south- west Spain, would lend support to Cockerel's (1928) hypothesis that the indigenous flora of Madeira came from the northeast and, since these are oceanic islands, that the facilities for transportation have been available at all times. It might not be out of place to speculate on the probable future of these species, granting that the forces working today continue to operate in the same way. Distinct geographic barriers prevent an interchange of genes between canariensis and Noronhaea, and their allied species. It is reasonable to assume that they will continue to differ progressively from each other and from the Madeiran group, and, if they are able to survive, will continue to pile up genetic differences which will decrease their congruity. The situation in Madeira is somewhat different. Andryaloides and taraxacifolia are at the present time forming hybrids, backcrosses, and complicated segregates. This "hybrid swarm" seems to have many vigor- ous and robust plants. It is probable that andryaloides, being a more primitive relic, to judge from its perennial habit, restricted range, and strict ecological requirement, will in the end be "swamped" by the more aggressive taraxacifolia. It is worth noting, however, that this is a very slow and gradual process. Since it is highly probable that some of Lowe's peculiar forms were hybrid derivatives, the two had come into contact over a century ago, perhaps much earlier. Yet most of taraxacifolia (in Madeira) and presumably most of andryaloides (in the highlands) are still unaffected by the mingling of the two at certain points. But taraxacifolia is known to be a montane plant in other countries ; hence, in all likelihood, it will gradually invade the highlands, and the mingling of the two will continue. In any event, andryaloides will have contributed a number of new genes to the invader, and this will afford possibilities of segregating out new combinations of characters that are even better than the present taraxacifolia combinations, making this latter subspecies even more ag- gressive. Divaricata, with its more restricted range and apparently more primi- tive characters (perennial habit, large flowers and leaves) and more homozygous expression, which is probably due to the more limited num- 1939] Jenkins : Cytogenetic Relationships of Four Species of Crepis 393 bers in the species, will very probably die out because of overgrazing by goats, or it may contribute something to the andryaloides-taraxaci- folia complex ; so that ultimately there will be one polymorphic species with many ecological types. In other words, hybridization may produce a degradation as well as a multiplication of forms, and this makes the probable phylogenetic history of any species a complex one. Those forms which are separated by both geographic and sterility bar- riers are, unquestionably, species in the Linnean sense. If the geographic barriers are present without the sterility barriers, it is a matter of opin- ion what will be the most useful and satisfactory way of treating the groups without doing too great violence to the convenient morphogeo- graphical system, and at the same time incorporating as much of the genetical data as possible. Goodwin (1937) with a very similar situation in Solidago feels that on morphological grounds, and for the sake of con- venience, the species should be kept distinct even though they do form fertile hybrids. In a recent paper Clausen, Keck, and Hisey (1936) have proposed a scheme based on Turesson's (1929) genecological system. The ecospecies (Linnean species) are the smallest units which are kept apart by the aid of an inner genetical balance mechanism ; that is, their hybrids are partly sterile. The ecotypes (subspecies) produce fertile hybrids and are kept apart through their geographical or ecological isolation. In other words, the main point is whether the forms have fertile or only partly fertile hybrids. The difficulty in this scheme is that, in practice, the fertility varies from to 100 per cent, and somewhere along this range a more or less arbitrary point must be selected in order to divide the species from the subspecies. In the present investigation three of these five entities maintain their morphological distinctness mainly through geographical and ecological isolation, and the other two, andryaloides and taraxacifolia, are becom- ing merged. It is clear that the final decision on the taxonomic treatment of such genetically close entities or systems must involve some arbitrary definitions. Since practical considerations must inevitably go along with every scientific approach to these problems, the fact of geographic or ecological isolation may properly serve as an adequate basis for the recognition of divaricata, Noronhaea, canariensis, and vesicaria. Their internal isolating mechanisms are as yet only imperfectly developed; that is, they are only on their way to becoming distinct species. However, the fact that they are to some degree incongruous and with continued iso- lation will probably become more so as time goes on, together with their morphological distinctness, would seem to be enough to establish them as distinct species. Nevertheless, there is still the possibility that, should the territory of 394 University of California Publications in Agricultural Sciences [Vol. 6 any one be invaded by another, the two thus coming together would un- doubtedly hybridize. It would remain for the botanist of that time to determine, through field studies and an investigation of the viability and fertility of the hybrid derivatives under natural conditions, the fate of the two species involved. SUMMARY The cytogenetic relationships of four closely related species of Crepis, namely, C. canariensis, C. divaricata, C. Noronhaea, and C. vesicaria subspp. andryaloides and taraxacifolia, were investigated. The evidence presented was derived from (1) a detailed morphological study of the parents and the hybrids between them, (2) a comparison of the somatic and meiotic chromosome situation of the parents and of the hybrids, and (3) the inheritance of a number of characters in the first- and second- generation hybrids. Among the five entities there were a great many morphological differ- ences which affected all parts of the plant. In the hybrids by far the greater number of these differences appeared to be the result of the presence of a large number of multiple genes. All five had a similar karyotype and the chromosome behavior in the hybrids was similar in every respect to that in the parents. The internal isolating mechanism between them was found to be incomplete, although varying degrees of congruity between them were indicated by the comparative fertility of the hybrids. For practical taxonomic purposes, the fact of geographic or ecologic isolation warrants the recognition of C. divaricata, C. Noron- haea, C. canariensis, and C. vesicaria, as species ; whereas andryaloides and taraxacifolia must be considered as subspecies of vesicaria, because they are hybridizing in nature and are losing their morphological dis- tinctness. 1939] Jenkins : Cytogenetic Relationships of Four Species of Crepis 395 LITERATURE CITED Babcock, E. B., and Cameron, D. R. 1934. Chromosomes and phylogeny in Crepis. II. The relationships of one hundred eight species. Univ. Calif. Publ. Agr. Sci., 6 : 287-324. Beadle, G. W. 1933. Further studies of asynaptic maize. Cytologia, 4:269-287. Cave, M. S. 1936. Cytological and genetical investigations involving Crepis foetida, C. com- mutata, C. eritreensis, and C. thomsonii. Unpublished thesis, filed in the Uni- versity of California Library. Clausen, J., Keck, D. D., and Hiesey, W. M. 1936. Experimental taxonomy. Carnegie Inst. Wash., Ann. Rept. Div. Plant Biol., 1935-36:208-214. 1928. Interspecific hybridization and the origin of species in Nicotiana. Zeitschr. f. ind. Abst. u. Vererb., Suppl., 1 : 547-553. COCKERELL, T. D. A. 1928. Aspects of the Madeira flora. Bot. Gaz., 85:66-73. Collins, J. L. 1922. Culture of Crepis for genetic investigations. Jour. Heredity, 13 : 329-336. Darlington, C. D. 1931. The analysis of chromosome pairing in Triticum hybrids. Cytologia, 3 : 21-25. DOBZHANSKY, TH. 1937. Genetics and the Origin of Species (Columbia University Press, New York), xvi + 364 pp. East, E. M. 1932. Studies on self -sterility. IX. The behavior of crosses between self -sterile and self -fertile plants. Genetics, 17:175-202. Goodwin, R. H. 1937. The cyto-genetics of two species of Solidago and its bearing on their poly- morphy in nature. Am. Jour. Bot., 24:425-432. Harland, S. C. 1935. The genetics of cotton. Pt. XIII. A third series of experiments with the crinkled dwarf mutant of G. oarbadense L. The cross barbadense crinkled x hirsutum crinkled. Jour. Genetics, 31 : 21-26. HOLLINGSHEAD, L. 1930. Cytological investigations of hybrids and hybrid derivatives of Crepis capil- laris and Crepis tectorum. Univ. Calif. Publ. Agr. Sci., 6:55-94. Kihara, H. 1929. Conjugation of homologous chromosomes in the genus hybrids Triticum x Aegilops and species hybrids of Aegilops. Cytologia, 1 : 1-15. 396 University of California Publications in Agricultural Sciences [Vol. 6 Lowe, K. T. 1868. A Manual Flora of Madeira and the Adjacent Islands of Porto Santo and the Desertas (John van Voorst, London), vol. 1, xii + 618 pp. McClintock, B. 1929. A method for making aceto-carmine smears permanent. Stain Tech., 4:53-56. 1933. The association of non-homologous parts of chromosomes in the mid prophase in Zea mays. Zeitschr. f. Zellforsch. u. mik. Anat., 19:191-237. Muntzing, A. 1933. Apomictic and sexual seed formation in Poa. Hereditas, 17 : 131-154. 1934. Chromosome fragmentation in a Crepis hybrid. Hereditas, 19:284-302. Navashin (Nawaschin), M. 1928. "Amphiplastie" — eine neue karyologische Erscheinung. Zeitschr. f . ind. Abst. u. Vererb., Suppl., 2:1148-1152. Navashin, M., and Gerassimova, H. 1936. Natur und Ursachen der Mutationen. III. Ueber die Chromosomenmutationen, die in den Zellen von ruhenden Pflanzenkeimen bei deren Altera auf treten. Cytologia, 7:437-465. Newton, W. C. F., and Pellew, C. 1929. Primula Jcewensis and its derivatives. Jour. Genetics, 20 -.405-467. Smith, F. H. 1934. The use of picric acid with the gram stain in plant cytology. Stain Tech., 9:95-96. TURESSON, G. 1929. Zur Natur und Begrenzung der Arteinheiten. Hereditas, 12 : 323-334. Wright, S. 1932. The roles of mutation, inbreeding, crossbreeding and selection in evolution. Proc. Sixth Int. Cong, of Genetics, 1 : 356-366. EXPLANATION OF PLATE PLATE 16 Rosette Leaves 1. Crepis canariensis. Note the almost entire margins and the winged petioles. 2. Crepis vesicaria subsp. taraxacifolia x C. canariensis F 4 . Note the slenderer petiole and the rounded apex characteristic of taraxacifolia. 3. Crepis divaricata x C. canariensis Y t . Note the dissection on the upper half of the leaves very frequently found in divaricata. 4. Crepis Noronhaea x C. canariensis Fj. Note the lyrate leaves and the slender petiole frequently found in Noronhaea. 5. Crepis vesicaria subsp. andryaloides x C. canariensis F a . Note the somewhat modified pinnate dissection characteristic of andry- aloides. 6. Crepis Noronhaea x C. divaricata. Note the characteristic divaricata-like dissection as in 3. 1-6 are approximately %2 their natural size. Mature Plants 7. Crepis canariensis. 8. Crepis vesicaria subsp. andryaloides. 9. Crepis divaricata. 10. Crepis vesicaria subsp. taraxacifolia, spreading form. 11. Crepis vesicaria subsp. taraxacifolia, erect form. 12. Crepis Noronhaea. The plants 7-12 are growing in 6-inch pots. [398] UNIV. CALIF. PUBL. AGR. SCI. VOL. 6 [JENKINS] PLATE 1 6 I. canaricnsis 4. F, Noronhaea X canariensis * 2. F, taraxacifolia X canariensis 5. F, andryaloides X canariensis 3. F, divaricata X canariensis 6. F, Noronhaea X divaricata 7. canariensis 8. andryaloides 9. divaricata V " N-V 10. taraxacifolia §#> 1 1. taraxacifolia m. 12. Noronhaea [ 399 ]