New Phytol. 37: 72-81. 1938

Calcutta, India


THE taxonomy of the genus Rosa has recently been clarified and the number of forms of specific rank has been considerably reduced. Many of the species are remarkably polymorphic and possess geographic races and ecotypes which are not yet understood. The recognition of well-defined series of parallel variations among the polymorphic species has elucidated the scope of intraspecific variation, potential and actual.

We are indebted to Dr G. A. Boulenger, who, after his retirement as one of the foremost zoological systematists of Europe, devoted his experience and taxonomic acumen to this difficult genus. Dr Boulenger has worked assiduously for fifteen years, revising first the European species (Boulenger, 1924-32) and then the Asiatic roses (1933-6). He distinguishes ten species in Europe and north-west Africa, and ninety-three species in Asia of which fifteen are found in other continents also. In my revision of the North American roses I recognized twenty native species, of these two also occur in Asia. This gives a total of 121 known species for the world (Boulenger, 1936a), an enormous reduction over the number recognized fifty years ago.

When the chromosome number of a species is known Boulenger includes it in his description. He has divided the genus into seven sections. Although his classification was based purely upon morphological characters all polyploid species (except R. gallica) are found in the Section Eglanteriae and the majority of them in the Group Cinnamomeae-Caninae. This section also contains over 50% of the total number of species in the genus. It is most northerly in distribution and contains the most generalized and least evolved types.


The simplest and most primitive species in the Eglanteriae are polyploid and also arctic or alpine in range. The least specialized wild rose is R. acicularis Lindl., which is a circumpolar species containing both hexaploid and octoploid strains. The hexaploid form extends farther south than the octoploid, but the northern limits of the two have not been defined. R. pimpinellifolia L. (the Scotch Briar) is the most generalized species of the mountains of Eurasia, and is the only native rose in Ireland: it is a tetraploid. These two species are the first to come into flower in the spring. Nearly all of the common species of America and Europe belong to the Eglanteriae. Many of them are polymorphic; they are successful species with extensive ranges and are in a state of rapid evolution and for these reasons they have received an excessive share of the attention of rhodologists. Just as the peculiar unbalanced polyploidy of the Caninae, common roses of Europe, led cytologists to consider the genus as a whole to be unique in behaviour, so the occurrence of primitive polyploids with high boreal distribution has influenced rhodologists to hypothecate an evolution of species in Rosa from generalized polyploid forms to specialized diploid ones: an hypothesis to which Boulenger (1936a) has given his assent.

Täckholm (1922) postulated an extinct arctic decaploid rose to account for the unbalanced hexaploid R. Jundzillii Bess. which has seven pairs and twenty-eight univalent chromosomes at diakinesis, and Boulenger also thinks that this is necessary (1936a). I have already pointed out (Erlanson, 1931) that this hypothesis is redundant since an unbalanced hexaploid with only seven paired chromosomes would result from the crossing of two hexaploids with only one set of homologous chromosomes in common. An octoploid crossed with a tetraploid could also give the same type of unbalanced hexaploid.

In 1925 Hurst put forward what Boulenger designates as "sa bizarre classification" based on the hypothesis of five differential septets of chromosomes in Rosa which were represented in nature by "five fundamental diploid species". It is evident that the number five was adopted because the author already accepted the hypothesis of an extinct ancestral decaploid. The geographic distribution of polyploid and diploid species led Hurst to conclude that modern rose species are all descended from this arctic decaploid.

This hypothesis is contrary to the established cytological principle that polyploidy is a secondary and derived condition. Commenting favourably on Hurst's theory of descent from higher polyploids Cockerell (1926) suggested that the hexaploid and octoploid forms had probably been built up from diploids by chromosome reduplication.

The following facts and considerations show that the acceptance of a polyploid ancestry for the modern rose species is not the only alternative.


(1) Primitive diploid relatives

The primitive form R. minutifolia Engelm. is diploid with 2X = 14. It has been placed in the genus Hesperhodos by Cockerell, a classification which is sustained by Boulenger and Hurst. This simple species has a very local distribution in south-western North America and Mexico. It is certainly related to Rosa and also to Potentilla (Boulenger, 1936b) and indicates long-existent polyphyletic lines of descent in the group.

There are three other diploid species all of which have been placed in monotypic genera by some taxonomists. Two of them, Rosa microphylla Roxb. (Platyrhodon) and R. bracteata Wendl. (Ernestella), are retained in the genus Rosa by Boulenger, each in a separate section. The third, R. persica Michx., he separates as Hulthemia, the generic name first proposed for it by Du Mortier (1824). It belongs to the deserts of Central Asia. It has simple, exstipulate leaves and densely prickly fruits. All these forms have probably been separated from the main line of descent in Rosa for long periods of time.

(2) Hybridization in Pleistocene times

Phenological studies in southern Michigan showed (Erlanson, 1930) that in early, mild springs the period of anthesis for individual bushes of early-flowering roses is longer and there is more overlapping between flowering periods of successive species; for it is the onset of the dry, hot July weather that stops flower production.

At the time of the greatest extension of the Pleistocene ice-sheet the mean annual temperature was only a few degrees lower than it is to-day in the regions affected. Any of the roses found in Michigan now could have thrived there then on unglaciated areas. At one time the ice-sheet extended all across Michigan as well as northern Illinois, Indiana and Ohio, and roses would have been confined to the southern borders of the latter states. During the Ice Age the days lengthened in spring in these regions just as they do now, but the spring temperatures would be lower until later in the year; a long spell of cool spring weather must have been the rule, and rose species had a long flowering period with ample opportunities for interspecific hybridization. The same conditions would hold in Europe. Thus the ice-sheet tended to bring northern and southern species together both in space and in time. Chilling encourages the production of unreduced gametes, and it is reasonable to postulate the production in Pleistocene times of amphidiploids, tetraploids and of triploids in which duplication gave hexaploids. Crossings between tetraploid species followed by duplication may well have given rise to the octoploid R. acicularis as in octoploid Aegilotriticum (Kihara & Katayama, 1931) or the hexaploid type crossed with a diploid species may have produced the octoploid.

When the Pleistocene ice-sheet was fully extended over Europe and America species belonging to the section Eglanteriae were no doubt to be found along its southern boundary. There is no reason to exclude the coexistence of primitive diploid and polyploid species, as well as more specialized diploid forms in warmer latitudes-the ancestors of the modern section Synstylae and other roses with southern distribution.

(3) Phytogeography of certain roses

The distribution of the hexaploid species in America is continuous and stretches from the Aleutian Islands and Alaska south to the northern border of California, to the mountains of Colorado and across from Alberta to northern New York (see Erlanson, 1929, Fig. 2). As I stated in 1929, it would appear as though the hexaploid type has not arisen more than once, if at all on the American continent .... The hexaploid types of Cinnamomeae may have entered America from north-eastern Asia and spread south and east over the continent" (Erlanson, 1929, p. 482).

The greatest geographical range of a rose in America is that of R. Woodsii Lindl., a simple diploid form which stretches from the coast of Alaska to Chihuahua, Mexico. R. blanda Ait., which is related to R. Woodsii and gives fertile spontaneous hybrids with it, belongs to north-eastern North America. It extends from Pennsylvania to Anticosti, and to Illinois, the Dakotas, Manitoba and Hudson Bay (see Erlanson, 1929, Fig. 4). This is a diploid rose which can thrive under almost arctic conditions but it has not spread far to the south.

(4) Generalized diploid types and polyploid relatives

Both R. blanda and R. Woodsii are related to R. cinnamomea L. which is the most generalized and primitive diploid Eurasian species with a very extensive range. Boulenger, Crépin and also Rydberg were struck by the similarity between R. cinnamomea, R. blanda and R. nutkana Presl. The latter is a hexaploid species of north-western North America. Crépin placed R. cinnamomea and R. nutkana in the same group in the Cinnamomeae (Boulenger, 1936, p. 140). No one has ever suggested that diploid R. cinnamomea is descended from hexaploid R. nutkana, and phytogeography points to the opposite conclusion. Thus the simple diploid type, R. cinnamomea no doubt descended from a diploid ancestral type which may have given rise also to R. Woodsii, R. blanda and R. nutkana (hexaploid). The latter three species (combined perhaps with R. rugosa Thunb.) I consider to be ancestral to the tetraploids R. Durandii Crépin, and R. californica S. and C. on the Pacific Coast, to the tetraploid R. arkansana Porter of the Great Plains, and to R. carolina L. and R. virginiana Mill. of the north-eastern United States, as well as to the diploids R. palustris Marsh., R. nitida Willd. and R. foliolosa Nutt.

Several interspecific hybrids of varying degree of fertility have been obtained between members of the Cinnamomeae (Erlanson, 1934, Fig. 20). Unfortunately no studies of synapsis in the F1, nor of comparative chromosome morphology in the genus, have yet been made. In an extensive study of the genus Crepis, Babcock (1934) and his co-workers find that the processes involved in the evolution of species are: (1) chromosome transformation, (2) amphidiploidy following interspecific hybridization, (3) autopolyploidy, (4) gene mutation. In Rosa all these processes have certainly also played important parts. Pairing behaviour indicates that translocation, reduplication of chromosome segments and the elimination of fragments frequently occur.

R. gymnocarpa Nutt. ranges from British Columbia to the hills of southern California. This rose often appears strikingly similar to some forms of R. Woodsii. It is a diploid like its close relative in Asia R. Beggeriana Schrenk. The latter is widespread in Asia between 30° and 50° latitude North at altitudes of 1500-5500 ft. (Boulenger, 1934). In Boulenger's opinion the fact that the styles, sepals and disk of the hypanthium are deciduous from the ripe fruit makes R. gymnocarpa and its Asiatic relatives "un petit group très naturel". He believes that they are related to the species in his section Pimpinelli-Suavifoliae. The deciduous character could well be due to a single mutated gene, and although significant taxonomically it does not necessarily indicate much change phylogenetically.

R. acicularis (octoploid and hexaploid) is difficult to distinguish from R. blanda in regions where both of them grow. The hexaploid type gives semi-fertile tetraploid hybrids freely with R. blanda. R. blanda var. hispida Farwell is distinguished from R. acicularis by a larger inflorescence, longer flowering laterals, more stamens, a later flowering date and a longer fruit ripening period. The evolutionary tendency in Rosa is from simple to compound inflorescence, towards more numerous stamens and later flowering time.

(5) Behaviour of polyploids and diploids compared

When the relative advantages and disadvantages of the diploid and polyploid conditions are taken into consideration, particularly in a partially self-sterile genus, the hypothesis of parallel evolution of R. acicularis and R. blanda from a common ancestor is feasible.

The characteristics which distinguish R. acicularis and R. blanda may partly be due to the difference in their chromosome numbers. The physiological effect of polyploidy on the plant as a whole is usually to slow down the growth (Lindstrom, 1936); it may also confer greater adaptability as in polyploid Dianthus (Rohweder, 1934).

Among American rose species the higher polyploids respond at once to a slight rise in temperature and put forth foliage and flower buds very early in the spring. The European tetraploid R. pimpinellifolia also comes into flower precociously, but its diploid relatives R. xanthina Lindl. (syn. R. Hugonis Hemsl.) and R. Primula Boulenger flower synchronously with it or even earlier in America. Both Boulenger and I noticed that the more primitive species are the earliest flowering roses. They have short flowering laterals and a one-to few-flowered inflorescence. The polyploid condition which slows up growth would render it difficult for polyploid roses to flower and to mature fruit in boreal latitudes if they produced long flowering laterals. This primitive characteristic is thus selected by the environment. Hagerup (1932) found that polyploid species appear to be adapted to more extreme conditions of cold and drought than related diploids. In Chrysanthemum the decaploid species is apparently arctic.

A plant of octoploid R. acicularis from Alaska was transferred to Ann Arbor, Michigan, where it leafed out soon after the first April thaw and was frequently badly damaged by frost in May. In some seasons all the floral primordial tissue was destroyed and no flowers were produced. The alternating thaws and frosts which are a feature of continental climates between 40 and 50° of latitude North, may militate against the spread southwards of octoploid races in Rosa. The Alaskan octoploid thrived at Pasadena (Erlanson, 1934) in southern California, and came into flower early in February, 3 1/2 weeks before the hexaploid type.

It would seem as though polyploids are best adapted to arctic-alpine conditions. If the polyploids in Rosa were produced from unspecialized diploid species in Pleistocene times, as suggested above, they have remained in the habitat to which they are best suited by migrating northwards as the ice-sheet retreated.

Giant diploid grains occur as 0.3-2 % of the pollen in some plants of R. blanda and R. Woodsii (Erlanson, 1934), but have not been observed in the more specialized diploid species. Autotriploids were found in diploid cultures of both these species (Erlanson, 1933). One culture of R. blanda contained a non-hybrid tetraploid as well as an autotriploid (Erlanson, 1934).

There are no reliable data on the genetics of Rosa. Allopolyploidy provides protected loci for gene mutations in new directions (Lindstrom, 1936). Yet it is almost impossible for a single autosomal recessive mutation to attain a homozygous condition and be selected out in a cross-breeding hexaploid or octoploid (Haldane, 1930). Dominant mutations cannot even be fully expressed when in the presence of four, five or more unmutated allelomorphs. Thus it is possible that the polyploid condition itself has been a bar to evolutionary progress for millennia in Rosa. However, types such as R. acicularis and R. nutkana may act as sources and reservoirs of new genes which are brought to light in the stable diploid descendants of unbalanced hybrids between them and diploid R. blanda and R. Woodsii. For example, R. blanda from northern Michigan was found to be extremely heterozygous and to throw seedlings which were tender 300 miles south of their natural habitat (Erlanson, 1929). The greatest degree of tolerance to different climates is shown by the simple diploid roses R. blanda and R. Woodsii. There are within R. Woodsii several distinct races which are adapted to specific environments as found in Drosophila by Timoféeff-Ressovsky (1936).

(6) Physiological factors that limit range

A factor which must be potent in limiting the southern boundary of boreal roses is the necessity for a period of after-ripening of the seeds at temperatures near 0° C. (Crocker, 1926). Hardiness, at least in some species, seems to be correlated with delayed development of the pro-embryo in the full-grown seed. Some races of R. Woodsii, as well as R. rugosa, do not exhibit this peculiarity, but this may have kept R. blanda and R. acicularis from spreading any farther south; under cultivation they thrive well anywhere in the United States.

R. nutkana from the State of Washington was unable to put out normal foliage and flower buds in the dry heat of southern California. The new growth was stunted and shrivelled and flowering was much delayed (Erlanson, 1934, Fig. 2). Other races of this species have done well in the moister atmosphere of Santa Barbara in Father Schoener's garden.

The more southern diploid American species such as R. palustris, R. foliolosa and R. setigera Michx. need a longer growing period before they flower, and for fruit ripening, than could be obtained far to the north. The two latter species are not viable much farther north than Detroit, Michigan.


There is no primitive tetraploid species in America comparable to R. pimpinellifolia of Eurasia. The American tetraploid rose species fall into three distinct groups, each of which is more specialized in habitat preferences and inflorescence type than the related diploid form (Erlanson, 1929): (1) R. carolina and R. virginiana in the northeastern region; (2) R. arkansana with a circumscribed area of distribution in the prairie region from Alberta and Saskatchewan south to Texas; (3) R. californica which is practically confined to California. All except R. californica are smaller than the related diploid. R. arkansana is semi-herbaceous and has a dwarf ecotype R. alcea Greene (Erlanson, 1934) which can withstand 60° below zero Fahrenheit in Canada (Wright, 1937). They are all plants of upland habitats as contrasted with the stream-bank and swamp habitats of the more primitive diploid and hexaploid species. The tetraploids come into flower after the related diploid species and usually continue to produce flowers in terminal corymbs on the season's turions throughout the season. R. acicularis, R. blanda, R. nutkana and R. Woodsii have a strictly limited flowering period of about 2 weeks.

In Eurasia there are specialized and highly evolved tetraploid roses represented by R. gallica L. (the Damask Roses) and some garden forms of R. chinensis Jacq. which are probably amphidiploids.

It is evident that tetraploid roses have arisen more than once on the American continent. These tetraploids probably arose from partially sterile diploid hybrids by duplication and appear to be of more recent origin than the higher polyploids and simple diploids (Erlanson, 1930). Their existence weakens the hypothesis of progressive evolution by descent of lower from higher chromosome numbers in Rosa.


Geographic distribution and cyto-taxonomy of Rosa together with data from the genetics and physiology of polyploids indicate that there have always been diploid lines of descent in the genus related to the polyploid lines.

Polyploidy is advantageous in arctic-alpine conditions, but it may slow up evolutionary changes. Physiological explanations for different types of range limitations are offered.

Hybridization of hexaploids with related diploids followed by the elimination of unpaired chromosomes in the descendants has no doubt been an important modus operandi in promoting the appearance of polymorphic diploid races with novel characteristics.

It is unnecessary to postulate a unique phylogenetic descent for modern rose species from a hypothetical arctic decaploid in order to account for cytological and phytogeographical conditions in the genus.