American Rose Annual 1945
The Scientific Basis of Rose Breeding
Dr. Walter E. Lammerts

University of California
Los Angeles, Calif.

Every rose breeder has certain more or less definite ideals or objectives toward which his breeding efforts are directed. Some of these ideals are shared by all rosarians, while others express the individuality of the breeder and may or may not find popular acceptance. Thus in my opinion an ideal red rose, for example, should have the following characteristics:

  1 ) Vigor similar to Radiance
  2 ) Long pointed or urn-shaped buds such as Lulu, Eclipse, or Charlotte Armstrong.
  3 ) Glossy or semi-glossy, hard, leathery foliage.
  4 ) High degree of mildew and rust resistance.
  5 ) Everblooming, rapid "breaking habit".
  6 ) Long cutting stems.
  7 ) Strong neck.
  8 ) Fragrance.
  9 ) Absence of blue fading reaction. World's Fair is ideal in this respect.
10 ) A high-centered open flower with 25 to 35 large petals.

Breeding Plan

Although the rose which perfectly meets these ideals may never be developed, any variety which is a step toward perfection is worthy of introduction. In my own experience most rapid progress toward combining all the above characteristics in one plant may be made by crossing variety A, having such desirable characteristics as mildew resistance and large glossy leaves, with variety B, having double dark red flowers but dull foliage, susceptibility to mildew, and short buds (unfortunately characteristics of most red roses so far introduced).

Simultaneously variety A is crossed with variety C, having long buds, vigorous growth, and deep pink or light red flowers (hence carrying factors for red). Hybrids from cross A x B having wide glossy leaves, a high degree of mildew resistance, and deep pink or light red double flowers are then crossed with pink-flowered hybrids of A x C having the longest buds and most vigor. Large progenies must be grown in order to get the best combination of characteristics in a very few plants.

These sister seedlings with wide glossy leaves, mildew resistance, and fairly long buds, are then crossed together, and in the resulting generation deep red roses having a vigorous growth habit, a high degree of mildew resistance, large glossy leaves, very long buds, and sufficiently double flowers may be expected. In order to understand the need for this type of breeding procedure, a discussion of rose chromosome numbers and behavior as they affect the inheritance of important characteristics is necessary.

Mechanics of Inheritance

Biologists are agreed that the factors determining the characteristics of plants and animals are serially located in certain deeply staining bodies of the cell nucleus, called chromosomes. At an early stage in the growth of the anthers, these chromosome are very elongated thread-like structures having a beeded appearance, and it is believed that the genetic factors are located in these small bead-like chromatin bodies called chromomeres.

Basic wild rose species ordinarily have seven pairs of these chromosomes and so are called diploids; one member of each pair is contributed by the male parent and the other by the female parent. In the formation of the pollen and egg cells this chromosome number is reduced one-half, so that each such gamete (mature sex cell) possesses only seven chromosomes, or one from each of the parental pairs. At the fertilization process, however, the pollen and egg nucleus unite, thus restoring the full number of fourteen chromosomes or seven pairs. In this manner each seed possesses a complete set of hereditary factors necessary for the normal development of the new plant.


Most commercial roses, however, are derived from hybrids involving several different wild species of roses. Somewhere in the history of their development the chromosome number became doubled up so that instead of having only 7 pairs or 14 chromosomes, they have about 28. Since the species of wild roses are rather similar in general characteristics, their chromosome sets are also similar, and have many regions or portions which are sufficiently similar to pair when brought together in the hybrid nucleus.

If species A has a paired chromosome set:


species B may be symbolically presented as having a set:


indicating relatively slight change in the genetic factors located in the seven chromosome pairs represented by the seven letters of the alphabet. Thus when species A is crossed with species B, a variable number of pairs is formed instead of 14 unpaired chromosomes, such as one would expect if no similarities existed in the sets of the two species.

It has been found that occasionally egg cells and pollen grains are formed with unreduced number of chromosomes, i.e., 14 instead of 7. When these unite, plants are formed combining the diploid number of the two parental species, that is having 28 instead of 14 chromosomes, and so are called tetraploid (tetra meaning four, and ploid referring in this case to the basic number seven). There are in these plants of our commercial hybrids now four chromosomes of each kind (represented by the letters A-G), and since chromosome A, for example, can pair with A as well as A, we quite frequently get association of four chromosomes, A A A A instead of pairs. These associations of four are called quadrivalents.

Since one oftens finds a variable number of pairs in the hybrids between the diploid species, it is not surprising that a variable number of quadrivalents is observed in the tetraploid rose varieties, though usually only three or four are seen, the rest of the chromosomes being in associations of three (trivalents), paired (bivalents), and unpaired chromosomes (univalents). The number may theoretically vary from 7IV to 14II depending on the amount of pairing in the original species crossed and the length of the chromosomes involved.

Practical Application

By now no doubt you wonder what all this has to do with breeding the ideal rose. It so happens that Captain Thomas is about the best source so far found of large glossy leaves and resitance to mildew. Not only is it disease resistant on the Pacific Coast, but in a letter Dr. J. Horace McFarland states, "Captain Thomas, as a climbing hybrid tea has shown satisfactory hardiness in this climate (Pennsylvania) and has not been at all bothersome with respect to disease proclivities. I don't think I have ever seen a mildewed leaf on it, nor does it seem anxious to take up blackspot infection." Incidentally, it is a very lovely single deep yellow rose and so is well worth planting in its own right if you like single blooms.

Captain Thomas is then our variety A. Crimson Glory is a fine red rose but in common with all other red roses is very susceptible to mildew and has rather small dull foliage. This will be our variety B in the hybridizing scheme outlined above. Charlotte Armstrong, derived from a cross of Soeur Thérèse and Crimson Glory, has a long bud, vigorous growth, and carries factors for red, i.e., will be variety C. All three of these varieties as well as many others of value in breeding better roses are tetraploids, that is, they show a variable number of quadrivalents. Hence even characters dependent on a single genetic factor and accordingly very simple in their inheritance in diploid crosses, show a much more complex segregation and variability in degree of expression as seen in the hybrid plants.

Dominant and Recessive Factors

Fortunately for the progress of rose breeding, several very important desirable characters have been found to depend in their expression on the action of one, or at most a few factors which behave as dominants. These are long symmetrical urn-shaped buds (such as are found in Soeur Thérèse and Eclipse), glossy leaf, double flower, and mildew resistance. Genetically speaking, dominant characters are the ones which appear substantially unchanged in the hybrids obtained by crossing to varieties which do not have the characters in question or any latent or recessive factors for them. The latent characters or the ones which disappear in the hybrids are called recessives.

Climbing habit is also dominant to dwarf bush habit and dependent on the action of a single factor, so that by backcrossing an otherwise desirable climber or pillar type to the commercially desirable bush type, one can always recover the recessive dwarf bush habit in about one half of the progeny; and among these dwarf bush types a rather small percentage will have the desirable features of the climber.

Glossy vs. Dull Foliage

Let us now consider the effect of quadrivalent formation on the inheritance and expression of glossy vs. dull foliage. Intercrossing of dull-leaved varieties always results in plants having dull foliage; hence we know that dull foliage is recessive, since dull x glossy results in glossy and dull-foliaged plants. However, there are two distinct types of behavior as exressed in the ratio of glossy to dull, depending on which varieties are selected as glossy-leaved parents. This fact may be seen by study of the following examples:

Type I — Crosses

Glossy-leaved Variety
Dull-leaved Variety
Dull Glossy
Mrs. Sam McGredy Herbert Hoover (35038)
Herbert Hoover (37148)
Mrs. Sam McGredy Soeur Thérèse 92 114
Sanguinaire C.P. Kilham 37 34
Sanguinaire Joanna Hill 21 19
Sanguinaire Night 50* 20
Total 231 229
Expectation 15:13 ratio 246 213
*Some plants in this population were definitely accidental selfs.

Type II — Crosses

Glossy-leaved Variety
Dull-leaved Variety
Dull Glossy
Captain Thomas Radiance 7 12
Captain Thomas Crimson Glory 3 21
Captain Thomas Lulu 6 28
Captain Thomas Soeur Thérèse 28 67
Goldenes Maines Soeur Thérèse 14 33
Total 58 161
Expectation 3:11 ratio 47 172

These results may be explained by the fact that the chromosomes carrying the dominant factor G which causes glossy foliage are able to form quadrivalents. Students of chromosome behavor have found that the chromosomes divide in the early thread stage before the reduction division, so that instead of only four chromosome threads being present there are actually eight, and each factor is thus represented eight times instead of the original four. Furthermore, factors sufficiently distant from the spindle fibre attachment of the chromosome (which is the mechanism by which the chromosomes are separated from one another) may cross over freely from one chromosome to the other. However, the final result is that only two chromosomes of the eight are segregated to each pollen or egg cell, thus giving the total of 14, two of each of the seven different kinds previously designated A to G.

In the varieties giving the Type I ratio only one of the quadrivalent chromosomes carries the dominant factor G; the others have the recessive factor g. Hence before the division, the quadrivalent may be represented by G1g2g3g4, the numbers referring to the four members of the quadrivalent. These then divide into eight and may be represented as follows: G1G1, g2g2, g3g3, g4g4. Any two of these may be segregated at random to the gametes, giving the following possible combinations:

G1 g2 g2 g3 g3 g4 g4
G1 G1G1 G1g2 G1g2 G1g3 G1g3 G1g4 G1g4
G1   G1g2 G1g2 G1g3 G1g3 G1g4 G1g4
g2   g2g2 g2g3 g2g3 g2g4 g2g4
g2   g2g3 g2g3 g2g4 g2g4
g3   g3g3 g3g4 g3g4
g3   g3g4 g3g4
g4   g4g4

Summarizing all the gametes having dominant G or GG and those lacking it, we have 1 GG plus 12 Gg or 13 G-bearing gametes to 15 gg gametes. Accordingly, on crossing a glossy plant of the above genetic constitution to dull, gggg, which forms only gg gametes, we expect to get a ratio of 15 dull to 12 glossy to 1 very glossy, which is substantially the result obtained in Type I crosses summarized above, though fluctuations in individual progeny results occur. Nevertheless, all fit the approximately 1:1 ratio quite closely except Sanguinaire x Night, where the excess of dull-leaved plants may possibly be due to accidental self pollination since Night was used as the female parent.

When the two dominant G factors are present, the quadrivalent G1G2g3g4 divides to form G1G1, g2g2, g3g3, g4g4. Any two of these may segregate at random, giving us 6 gg gametes to 16 Gg gametes to 6 GG gametes, or as regards dull vs glossy, a ratio of 3 dull to 11 glossy, when grossed to dull-foliaged varieties. This is approximately the ratio obtained in the Type II series of crosses. Among the glossy-leaved plants there are two grades of glossiness, 3 having the constitution of GGgg and so being much glossier than the other 8 which have the Gggg constitution. From the practical breeding point of view, the important facts are that glossy leaf is dominant and is inherited as a unit character, i.e., depends on a single factor, and so is not lost or diluted no matter how many backcrosses to dull varieties one may wish to make. Due to quadrivalent association, it, as well as other dominant factors, may vary strikingly in its degree of expression, depending on whether it is present in the simplex Gggg, duplex GGgg, triplex GGGg, or quadruplex GGGG form. Modifying factors may limit slightly the degree of expression of the glossy factor, but their importance in this case is very minor.

Mildew Resistance

In the progeny of crosses between Crimson Glory, which when selfed gives only susceptible seedlings, and Captain Thomas, one gets varying degrees of mildew resistance. Letting 0 equal no infection and using grade 4 to express a high degree of susceptibility or infection, a small progeny when tested by the floating leaf method under ideal conditions of growth of mildew mycelium gave the results shown in Table I.

Table I
Grades of mildew resistance and number of plants in each grade obtained by crossing
Crimson Glory x Captain Thomas.
Grade of resistance 0 0- 1 2 3 4 TOTAL
Number of plants in each grade 3 12 5 3 1 1 25

Classification of a much larger population under field conditions when mildew was severe on susceptible varieties gave substantially the same results. Only a few were completely immune but many could be classified as good, i.e., showing a high degree of resistance. Evidently mildew resistance is dominant and rather simple in its inheritance. It may then be handled as a unit character and combined with any other desirable characteristics.

Bud Length

The evidence so far collected indicates that the long urn-shaped bud is dependent on a dominant factor or factors since short-budded varieties such as Crimson Glory give only short-budded seedlings when crossed with other short-budded varieties such as Captain Thomas. Symbolizing long bud by L and short bud by l, Soeur Thérèse has the factorial composition LLll. The various grades of bud length can then be represented as follows:

LLLL   very long bud
LLL l   longer than Soeur Thérèse
LL l l   Soeur Thérèse
L l l l   Crimson Glory, Night other red roses
l l l l   very short globular buds

Soeur Thérèse when self-pollinated gave 1 plant with extremely long buds about twice the length of Soeur Thérèse; 5 plants with very long buds; 21 plants with buds as long as Soeur Thérèse; 22 plants with medium long buds; 7 with short, and 1 with very short buds.


The best assumption seems to be that doubleness is dominant but also quantitative in its expression. The various grades of doubleness may then be set up as follows:

dddd   single 5 petals. When selfed only singles with 5 petals are obtained.
Dddd   7-10 petals. When selfed one may occasionally get roses with 30-50 petals; i.e., those having constitution DDDd.
DDdd   15-25 petals — as Mrs. Sam McGredy
DDDd   30-50 petals
DDDD   very double, 80 petals or more

Color Inheritance

Dark maroon-red flower color is dependent on recessive factors. [?!] The nearest approach to red obtained when deep red varieties such as Crimson Glory or Night are crossed to yellow varieties, is a rose-red to Tyrian rose. The flowers also fade rapidly to magenta-red. Only by backcrossing to maroon, nonfading reds such as World's Fair, can one recover a large percentage of plants with dark maroon-red color.

Orange-yellow, yellow, white, and scarlets are also recessive in their inheritance, i.e., give pinks when crossed to magenta-red or maroon-red roses. In other words, magenta-pink or Tyrian rose to rose-red is the dominant kind of color in roses. Deep yellow such as that found in Goldenes Mainz is recessive to light yellow. Hence to recover deep yellow colors, one must backcross to the deep yellow varieties. Also white is recessive to cream, buff, or light yellow. Many of our so-called "white" roses, such as Sir Henry Segrave, Odine, and Alice Stern, are really creams and dominant to true white in their color expression. Hence to recover the latent white in hybrids with these varieties it is necessary to backcross to white.

The very popular "yellow and silver reverse" type of petal color such as characterizes roses like Condessa de Sastago and Contrast is recessive to self-color. It may be of interest to note that a few yellow and silver reverse plants were recovered in the hybrids of Captain Thomas and Crimson Glory, indicating that both of these varieties carry these recessive factors. In all cases of color inheritance, the expression of the color is of course complicated by the fact that the factors determining the desired color may lie in chromosomes which are able to form quadrivalents. Hence the number of plants which need to be grown to recover a desired recessive color such as deep buttercup-yellow are much larger than is necessary in diploid crosses.

Finally it should be mentioned that characteristics such as vigor, fragrance, thorniness, strength of neck, length of cutting stem, width of leaf, and shape of bud and open flower seem, in the present limited state of our knowledge, to be the result of the interaction of many factors and so are usually intermediate in their expression and quantitative in their type of inheritance. One should therefore always cross plants having such dominant factors as glossy leaf, long bud, double flowers, and mildew resistance to varieties having vigor, comparative freedom from thorns, strong necks, well-shaped buds and flowers, and wide leaves as the final step in the breeding program. Backcrosses will also probably be necessary before the ideal desired combination is obtained.

CybeRose note: Since Lammerts' time, much has been learned about the inheritance of fragrance. First off, the most potent of the Rose perfumes (rose ketones) are produced by the oxidative degradation of yellow carotenes. A bright, non-fading yellow rose is not likely to be richly perfumed. You have a choice: see the carotenes or smell them. It is interesting to note that among the yellow Tea-Noisettes, the most deeply colored varieties also have large flowers and little or no sweet scent. The small-flowered varieties fade rapidly and smell very sweet.

The "bicolor trait" is confusing. A rose may restrict its red pigment to the upper epidermis, but will not appear to be a bicolor unless the mesophyll (tissue between the upper and lower epidermis) is reflective/opaque. Other roses have more red pigment on the reverse of the petals ('Mrs E. G. Hill' and Moore's 'Dresden Doll' are examples). And still others, like 'Baccará' and 'Independence' are pigmented both front and back. If such a doubly pigmented rose also has the reflective/opaque mesophyll, the color should appear brighter, without being bicolor. Therefore, in Lammerts' cxamples, the bicolor effect of 'Condessa de Sastago' and 'Contrast' might be recessive to both (a) pigmented lower epidermis and (b) non-reflective mesophyll (which may be inherited quantitatively, though I have no data on that point. I can say that some bicolors ('Snowfire' and 'Oso Easy Cherry Pie') have cleaner reverses than 'Love'.

When I first read this paper, sometime in the 1970s, I assumed that Lammerts knew what he was writing about. But when I gained access to a large rose garden, I soon learned that he was just assigning "genotypes" to accomodate his preconceptions. In fact, diploid roses show the same range of doubleness as the tetraploids. Start with any single flowered variety (dd) and the very double 'Clotilde Soupert' (DD). Now, try to convince someone that all other Teas, Chinas, Noisettes and Dwarf Polyanthas with more than five petals must be Dd.