Trans. Nat. Hist. Soc. Northumberland and Durham 5(2): 244-98. (1921)
The Genus Rosa, its Hybridology and other Genetical Problems
J W Heslop Harrison D.SC

      1. Introductory
      2. Orthogenesis in Rosa
      3. Pollen conditions in the Rosae, with some consideration of other features in their Reproduction.
        Pollen Abortion in the Roses.
        Pollination in the Roses.
      4. The Phenhybrids of Rosa.
        History of recognised Rose Hybrids.
        Modern Conceptions regarding them.
        Their uneven Geographical Distribution.
        Their Powers of producing good Seed.
        Observations on the Hybrid Forms.
      5. List of Literature cited.


Species rosarum difficillime limitibus circumscribuntur et forsan natura vix eos posuit — Linnaeus, Species Plantarum.

Species rosarum difficillime distinguuntur, difficilius determinantur; mihi videtur naturam miscuisse plures vel lusu ex uno prures formasse; hinc qui paucas vidit species facilius eas distinguit, quam qui plures examinavit. — Linn. Species Plantarum, 1753.

Although written more than a hundred and fifty years ago, to most botanists, even of to-day, the above extracts from the Species Plantarum appear to be as true as Linnaeus imagined; as a natural consequence, except to the specialist, the genus Rosa is more or less of a tabooed subject.

* These are the actual words used in certain dichotomous tables to separate alleged species!

Nor can this be wondered at when the novice beholds the bewildering array of forms, seemingly alike, masquerading under different names in the carefully labelled herbaria of the rhodologist, or even after he has attempted to name a casually collected rose by means of some flora. Both experiences are [245] disheartening, but to the enquiring mind readily understandable. In the first place, so casually are rose specimens accumulated, and so effectually does the process of drying repress any but the most unreliable of characters, that forms, in reality most diverse, in the end look precisely alike; for example, it is quite an easy matter to produce a form of Rosa coriifolia so similar as a herbarium specimen to a member of the tomentosa group that even to the trained eye they are almost indistinguishable. If then the original descriptions of species be based on such material, as they often enough were, how little can we expect such accounts to square with the roses as they grow in nature, and how futile, therefore, our endeavours to determine a plant by their aid. Even with an ordinary genus a dried specimen may prove troublesome enough although the obstacles to a correct determination are not insurmountable, and so would it be with Rosa were it not that its members—call them species, little species, elementary species, microgenes or what you will—are so excessively variable that another difficulty arises. In some cases no two bushes are exactly alike; by that very fact they practically invite the attention of some individual blessed (?) with the "mihi"—or what Crepin used to call the "bush-mania"— who forthwith sees material for a possible new species. Obviously enough, the most hardened sinner would avoid the error of describing, as distinct, plants differing merely in some microscopic denticle on the back of the main leaf-serrations or in characters of similar value. Far from it, he selects his specimens from a bush as remote from the nearest book description as possible and upon those erects his "new" species. Now in all of its many inconstant characters, the variation in Rosa follows the ordinary frequency distribution of a Quetelet curve, as a very simple test will demonstrate. Collect, for example, a quantity of terminal leaflets from various bushes of the R. lutetiana group; measure their lengths, ignoring absolutely the fact that they are plucked from shrubs with large*, medium, or small leaves, and plot a [246] graph. The result is a variability curve of the most ordinary type. And wherever the character lends itself to such treatment, be it in the degree of biserration of the leaves, the length of the styles, the number of setae on the fruit, the number of glands per square millimetre on the underside of the leaves or in any other feature, the result would be the same.

This being the case, it is clear that if a value representing the condition of any given bush in respect to the sum of its characters could be assigned to it, and such statistics be collected for a host of closely allied forms, then these figures, too, if plotted as a graph, would produce a normal curve of variation. The natural outcome of the failure to recognise this elementary fact in the description of new species on the plan outlined above is that, for the most part, rose species simply exist on paper and at best only apply to bushes lying at the limits of the variation curve proper to their group and at irregular points more or less remote from these; such descriptions, therefore, are worthless.

To meet the circumstances of the case, the scientifically correct procedure ought to be, firstly, the collection and study of innumerable forms, then the selection for description of those falling at or near the crest of the Quetelet curve for their variability, followed, to do full justice, by a careful statement of the range and trend of variation as determined by careful study in the field, completed by details concerning the rose as it grows. Had this method been adopted, independently of one's attitude toward the status of such forms, the number of so called species would have been enormously reduced.

This brings us up sharply against the value to be attached to such forms; are they species or are they not? If not, what are they?

* It is well to emphasise here that, although, treated as above, all such forms would yield a Quetelet curve, they do not pass into one another at the minima of their respective curves; they cannot by any possibility be represented as a connected series of such curves. They are liable to merge at any point and via any or all of their characters.

Anyone studying the roses as living organisms, paying due attention to habit as well as structure, cannot fail to be impressed by the numbers of apparently distinct forms [247] separated by characters stable enough in themselves but nevertheless of relatively minor importance, each form being represented in any given station by crowds of individuals. Furthermore, even in stations vastly different in geological formation, there exist precisely the same types differing in the same way. For example, in two lanes in Mid-Durham on the Coal Measures a certain variety of what I call Rosa omissa occurs in hundreds. This is exceedingly closely allied to R. mollis which grows alongside it. Twenty-five miles away on the Magnesian Limestone of the coast denes identical forms of both R. omissa and R. mollis abound, and the same holds good of the Millstone Grit in Northumberland, thirty miles in the opposite direction. In addition to this locality-constancy both breed true to their special characteristics when due allowance is made for variation, in response to environment, within the limits of their particular range of fluctuating variation. So true to type and its own range of variation is each of these that knowing them only from these stations one could only regard them as genuine Linnaean species. If, however, one assembles with them numbers of similar forms from many stations, near as well as far, one is driven to confess that they merge* imperceptibly into each other as well as into many types of similar level, in this fashion forming a linked-up series of groups with the individual members of the same group substantially alike but each group separable from its neighbours by its own special character or characters.

Having thus equipped ourselves, and having submitted the material to detailed examination, we realise at once that, whilst we possess chains of groups continuous within the limits of the chain, on the other hand each chain is definitely divided from the next in respect to characters with which the comparison which linked up the original groups into [248] parallel chains is no longer possible. Here at length we are up against something so close to a species in the Linnaean sense of the word that the unprejudiced observer can only regard them as such, despite the fact that few rhodologists would look upon the chain as of lower value than a section or subsection, comprising many species.

Admitting the validity of our conclusion, we find that our chain of collected forms assumes the familiar guise of the collective species, whereas the members of that chain sink to their natural level of the Jordanian species or microgene as I prefer to call it. That the microgenes glide into one another depends in all probability, for the most part, on hybridity, as I shall demonstrate later.

* The name Rosa Afzeliana (a reinstatement due to Almquist) is used to cover the two old collective species R. glauca and R. coriifolia, the separation of which was ridiculous in view of the feebleness of the characters employed to differentiate them. The old coriifolia forms massed around var. Lintoni are transferred to their natural allies the Rubiginosae. Similarly R. tomentella and its satellites are removed to the Agrestes. In the same way, and for similar reasons, I have united the old collective groups R. canina and R. dumetorum under R. canina.
In the Villosae I include all the related mollis and omissa forms.

To avoid confusion I propose to call these chains section-species, and on that basis the number I am prepared to admit amongst the forms I have studied in a wild or half-wild condition in the north amounts to eight. These are R. canina, R. Afzeliana*, R. rubiginosa, R. agrestis, R. mollis, R. tomentosa, R. pimpinellifolia, and R. cinnamomea, each of which includes a fairly large number of microgenes, which, as they exist in nature, are far from coinciding with the multitude of alleged species set out with such lavish detail in the various rose floras. Further, for ease in reference and to harmonise my nomenclature as far as possible with that current amongst botanists, I shall designate these section-species by the sectional names Eucaninae, Afzelianae, Rubiginosae, Agrestes, Villosae, Tomentosae, Pimpinellifoliae, and Cinnamomeae respectively.

Again, within these eight sections six, the Eucaninae, the Afzelianae, the Rubiginosae, the Agrestes, the Villosae, and the Tomentosae are so much nearer to each other than to the others that they will always be referred to, when all are intended, under the comprehensive supersectional name Caninae.


So confused are the usual conceptions of geneticists and palaeontologists (and indeed of all who are interested in the matter) concerning the subject of orthogenesis, that it is with considerable hesitation one approaches it at all. Under these circumstances, therefore, before proceeding to discuss it in connection with the genus Rosa, it seems advisable to set forth the prevalent views on the subject as briefly as possible.

The word itself simply implies evolution or development in a definite direction and was adopted by Eimer himself from Haacke's "Gestaltung and Vererbung" to replace his own somewhat cumbersome phrase "bestimmt gerichtete Entwicklung" (= definitely directed evolution). Its more extensive use in genetics and allied subjects proceeds from this action of Eimer's.

Unfortunately, during the past twenty years, instead of remaining fixed, the meaning attached to the term has diverged along two distinct lines and so produced considerable uncertainty as to the exact import of the word. In its simplest interpretation, and that nearest the root meaning, it conveys to one's mind nothing more and nothing less than that certain observations have been made and that these are in harmony with the law of orthogenesis. In other words, the trend of evolution in certain groups of organisms is along definite lines. This appears to me to represent very nearly the value Eimer himself intended to assign to the term, and is the sense in which all palaeontologists and many zoologists (like Towers in discussing colour and pattern variation in Leptinotarsa, Whitman in describing his pigeons, and myself in treating the Lepidoptera) have understood it.


As may be readily perceived, in this original and restricted sense no explanation was or is advanced as to the cause of the phenomenon, although, of course, speculations as to what were the agents responsible for its existence were absolutely inevitable. From such theorising arose the transferred meaning of the word, in which it is employed, not as just explained to include only observed facts, but to cover the theory or theories evolved to account for those facts. The latter usage of the word is followed by Lull in his book on "Organic Evolution" in which he asserts that "Orthogenesis is a theory that variations, and hence evolutionary changes, occur along definite lines impelled by laws of which we know not the origin." Evidently, too, it is in this sense that Morgan looks on the word; otherwise he would not have described orthogenesis as being little more than a "mystic sentiment."

* "The causes of orthogenesis, according to my notion, lie in the action of external influences, such as climate and food, on the given constitution of the organism."

Such, then, are the differences in interpretation existing in the works of present-day writers, but they are in nowise attributable to Eimer. He very carefully distinguished between his actual illustrations drawn from a wide field in the Lepidoptera to explain the principle, and the original causes thereof. Briefly, without appealing to the inner driving force or principle of Morgan, or the "urge", vis-a-terge, "kick" or "vital force" quoted by the same author as synonyms of orthogenisis, Eimer announced that "Die Ursachen der bestimmt gerichteten Entwicklung liegen nach meiner Auffassung in der Wirkung äusserer Einflüsse—Klima, Nährung—auf die gegebene Konstitution des Organismus."* By this dictum Eimer, without equivocation, commits himself to a belief in the inheritance of acquired characters. Although I made no definite pronouncement it will easily be deduced from my work on the Geometrid genus Oporabia that I was driven to similar views to account for various facts concerning lepidopterous larvae.


In the present paper, with but the barest indications of possible causes, I intend to demonstrate as fully as possible that the development of section-species, microgenes and varieties in the genus Rosa, and more especially in the supersection Caninae, proceeds on orthogenetic lines even if of a rather peculiar type; in doing so I believe that I shall be giving the first illustration of the law derived from the phanerogams. However, despite the emphasis on the Caninae, it must not be supposed that the action of the principle is confined to them; on the contrary the Pimpinellifoliae and Cinnamomeae exhibit the same phenomenon. Thus, as far as possible these latter groups must be dealt with, but owing to the circumstance that they are at best stranded and possibly more primitive groups, badly represented in the North of England and greatly diminished in species throughout the range of Rosa, very little material remains for study; whatever has survived the vicissitudes of the past falls exactly in line with the Caninae.

* It must not be forgotten that I regard the glabrous R. glauca + the hairy R. coriifolia as forming the section species R. Afzeliana. However, in naming microgenes, to remove misunderstandings in the minds of English readers, the nomenclature of Wolley-Dod's list is employed throughout this paper.

My attention was first drawn to the subject some years ago by the occurrence of a green-leafed form of Rosa coriifolia* on the railway banksides near Billingham, South Durham, which ran down to the microgene frutetorum. In spite of its proving on examination to possess all the characters proper to the aggregate R. coriifolia, as I then termed the hairy-leafed members of the Afzelianae, I could not help being struck by its extraordinary resemblance to that segregate of the Tomentosae which I take to be R. tomentosa, Smith. In consequence, influenced by its ambiguous appearance which had caused me at first sight to look upon it without question merely as R. tomentosa, I refrained from including it in my "Wild Roses of Durham" but laid it aside for future study. The same season I made a direct comparison between it and [252] fresh specimens of undoubted R. tomentosa gathered at Lamesley, North Durham, when I was amazed to discover that practically in every character, leaf shape, serration and colour, flower colour, size and shape of fruit, colour of bark, and finally, habit of growth the two coincided; whatever differences were perceptible lay in the technical characters employed to differentiate the section-species Afzelianae (= corrifolia + glauca) from the Tomentosae. Subsequent investigations revealed that this microgene of the Afzelianae abounded throughout the coast denes of Durham, in the hedges along the bridle path between Greatham and Cowpen, and in North Yorks., and further that its latitude of variation was much less than is common to most rose microgenes. The next development occurred in late August of the same year when a very similar clump of bushes was examined close to Greatham village; these bore fruit just changing colour and were therefore determined without consideration as Rosa frutetorum. Imagine my dismay when, on making detailed examination at home, I had to confess that the plant appertained to R. glauca. The tale did not end here. In the following year, working in Hawthorn Dene (another of the beautiful Durham coast ravines), I found a fourth rose, once more referable to R. glauca, but in this case distinguished from the form just described and the similar R. frutetorum and R. tomentosa by its very conspicuous blue green, pruinose leaves. Later, in the same expedition, I collected a R. coriifolia variety of precisely the same type with leaves covered with a waxy bloom, as well as a Rosa mollis very close indeed to it. Next day I brought all these alongside R. frutetorum and R. tomentosa from Lamesley. In the features referred to as linking R. frutetorum and R. tomentosa together, save in the one point of leaf colour (and of course in their special section characters), the agreement was complete.

* A species-type may be defined as the model of a group of microgenes of similar facies, with representatives, in each section, possessing the characters proper to that section added to those peculiar to the species-type.

To get parallel Eucaninae forms proved far from difficult, for of the bushes growing along a disused waggon way at Harraton, North Durham, more than 50 per cent. of the total, [253] and at least 90 per cent. of the Eucaninae, belonged to a microgene R. flexibilis simulating the green R. glauca discussed previously, whilst a much smaller number were of the related blue-green glabrous and blue-green hairy types. Thus, to all intents and purposes, I had traced the passage of microgenes of like type to the frutetorum through all the section-species except the Rubiginosae and Agrestes. With the former section I have never made the link, simply because it is only a denizen in our counties, whilst with the latter the connection came later when I recognised that the microgenes regarded as glandular Eucaninae were actually Agrestes. Rosa vinacea then fell into its natural place as an Agrestes representative of the frutetorum—tomentosa—flexibilis alliance to which I apply the name "species-type."*

By similar stages other linked species-types were discerned, although not every one was recognised in all the sections. For instance, whilst the lutetiana—dumetorum fraternity had yielded all four possible combinations in the Eucaninae and Afzelianae, only certain Tomentosae provisionally named by me R. pseudo-cuspidata went with them—and so with others.

This is, of course, the glandular R. coriifolia var. Lintoni of British authors.

In all I determined seventeen definite species-types occurring in three or more groups, with a less number represented in two only, all of the section-species being involved except the Rubiginosae. The members of that section, for reasons already stated, could not with certainty be attached to any forms occurring locally. Recently, however, I have received a number of Norwegian roses from Traaen which included a glauca form determined by him as Rosa Traaeni Almquist. To this the Rubiginosa form R. Lintoni, which is common hereabouts, and the Eucanine R. sylvularum present [254] a close resemblance. Even then I was sceptical as to the Rubiginosae showing to the full the capabilities of the other sections, so much was I influenced by the uniformity of the full green R. rubiginosa growing sporadically in a subspontaneous way in our hedges. I simply couldn't imagine its yielding glaucous leaved varieties. This last stumbling block was likewise removed by Traaen. Thanks to him, I now have a splendid series both of the glabrous and of the hairy pruinose Rosa glauci-rubiginosa and R. supra-rubiginosa, whence we see that R. rubiginosa can produce all four minor forms described below.

* Prior to any serious study of the material I had often been at a loss as to how to treat the blue-green lutetiana and dumetorum forms running down to the same name as their green relatives.

In this fashion the facts which were destined to give rise to, and then impress, the idea that species formation in the Rosae was on an orthogenetic foundation were accumulated.* We are now in a position to expand and systematise the notion.

* And in more southern stations for mountainous habitats, as Gelmi so clearly explains in the case of Rosa glauca in the Trentino

Whatever the common ancestor of the Caninae was, it appears, if one judges from its geographical distribution, to have separated early into three divisions, each of which in turn gave rise to two allied sections, the pairs being Afzelianae-Eucaninae, Rubiginosae-Agrestes, Villosae-Tomentosae. And the remarkable thing about these pairs is that, in every instance, the second members of the pairs bear the same relation to the first, the Eucaninae differing from the Afzalianae as the Rubiginosae differ from the Agrestes and the Villosae from the Tomentosae. In each case the first section of any particular pair possesses a broad, flattish, hemispherical woolly head of stigmas almost sessile on and obliterating the disk (text fig. 1, b), whilst the second displays less hairy and, often enough, glabrous styles, more or less protruding from the disk, thus leaving it clear and distinct (text fig. 1, a); similarly in the first section-species the sepals are more or less erect and subpersistent (figs, 1, 5, pl. xvi; figs. 7, 8, pl. xvii; fig. 10, pl. xviii; fig. 33, pl. xx; and pl. viii B.), whilst in the [255 ] second they are reflexed (rarely spreading) and deciduous (figs. 9, 12, 13, pl. xviii; figs. 14, 15, 16, 17, pl. xix; fig. 20, pl. XX; figs. 26, 27, pl. xxi), falling before the fruit ripens. Finally, the first section of any given pair ripens its fruit markedly in advance of the second. It will thus be apparent that one section-species of each pair of the Caninae is fitted for self pollination and northern* latitudes, and the second for insect pollination and more southern climes, thereby indicating that the parallel evolution of the allied pairs from the primitive Caninae ancestor was possibly a response to environmental factors. In this parallel movement we discern the first hint of the workings of the unfolding principle of Elmer's.

Text-fig. I. a. R. flexibilis (Eucaninae).
b. R. Reuteri (Afzelianae).
To show difference in stigma heads.

But progress did not end here; in these groups (as in others not studied by me) in Asia, America and Europe, a gradual advance has occurred with the evolution of a multitude of species-types, each species-type marking a definite halting point in the onward march. Once again, as these stages reveal themselves, we are bound to admit that throughout the sections there is complete agreement, for each species-type is common to all; a further unfolding of the latent powers of the sections has manifested itself. Again the evolutionary principle followed is strictly orthogenetic in its action.

Next, in each section-species, every species-type breaks into two microgenes, in one of which the foliage is of a full green [256] colour, and in the other of a glaucous hue due to its secretion of a waxy bloom. Finally, each of these two forms in its turn yields two varieties, one with the leaves glabrous and the other with them more or less hairy. Only one explanation of all these marvellous coincidences appears at all feasible, and that is, that in the derivation of the sections from their hypothetical common ancestor, in the segregation of the section-species into species-types, in the break of the species-type into microgenes, and in the splitting of the microgenes into varieties, we have a wonderful exposition of evolution on definite lines or, in other words, of orthogenesis.

As is obvious, no attempt has been made to trace the phylogenetical connection of the various species-types, although that would entail no great labour. One could commence with the neutral type of leaflet exemplified by R. Traaeni and from it trace by degrees in direct lines every form of leaflet from narrow to broad, from uniserrate to multiserrate, from forms with connivent teeth to those with divergent, and so on with every possible combination, just as Eimer followed up the wing patterns in the Lepidoptera. To do so, in the absence of exact evidence as to the true arrangement or even the correct starting point, would be unprofitable, for it would simply produce a scheme a fit target for gibes of the same nature as assail the palaeontologist when he arranges his fossil reptilia and amphibia in orthogenetic series. He is always accused of making his arrangement to suit himself, irrespective of the time of appearance on the earth's surface of the creatures with which he deals. My omission cannot affect in the slightest the genuine orthogenetic course of evolution visible in Rosa, any more than a failure to indicate the order of the figures in the "cat's cradle" would prove that they had not been evolved in a definite order.

Independently of my researches, Almquist, another worker in the same field, travelling by the reverse route to mine, arrived at precisely the same conclusions, but as far as classification is concerned, far outdistanced me. Commencing [257] by observing the break of the glaucous and the green microgenes of certain Afzelianae into their smooth and hairy forms, he followed its occurrence throughout that group, and in the end proved its universality in the rose sections. With the huge mass of material at his command, and attacking the matter purely from the standpoint of the systematist, he proved that there were 31 species-types and fully characterised them, utilising, in addition to the features employed by me, the forms and relationships of the leaflets on the flowering sprigs as the basis of his scheme. On this foundation he constructed a tabular synopsis of the sections, species-types, microgenes and varieties; but in this many blank spaces exist.

All of this he carried out without the slightest inkling of its genetic import: nevertheless his table can be used as a kind of Mendeléeff's Periodic Classification to predict the existence of roses as yet undiscovered. Prior to this, however, on the evidence of my own work, I had outlined and lectured on a similar scheme evolved, after prolonged study, from my seventeen or more species-types. Moreover, I had employed my own arrangement with conspicuous success to predict the occurrence of a great number of forms (chiefly Agrestes) allied to Rosa blondeana and R. sclerophylla which I afterwards discovered in South Durham and North Yorks where that group abounds.

The similarity of the tables developed by both Almquist and myself to that drawn up by Mendeléeff to illustrate the Periodic Law in classifying the chemical elements cannot be too strongly emphasised. Almost to the most minute detail they agree; we have the vertical groups of the chemist agreeing with my section-species, the horizontal series or periods with our species-types. Furthermore, if we allow for the division of the six members of the Caninae into three pairs of more nearly related section-species, we have an immediate parallel to the divergence of the groups in Mendeléeffs table into two subgroups. For instance, the relationship between the Eucaninae and Afzelianae is much the same as that between [258] the Lithium-Sodium-Potassium subgroup and the Copper-Silver-Gold subgroup. Again, the Semivillosan R. faroensis and R. orientalis, with R. pimpinellifolia, suggest in a striking way the transitional elements in the chemist's group eight. Lastly, it would not be a surprising thing should it turn out in the end that the parallel nature of the two tables is founded on chemical differences of the same type.

Thus Almquist working as a systematist and I as a geneticist, arrived at essentially the same conclusions and announced them at practically the same period.

To fit the Cinnamomeae into the general arrangement was not difficult, for the only two forms of that section known to me form by far the most beautiful and clearest illustration of the differences between the clear green and glaucous microgenes I have examined. Rosa pimpinellifolia (figs. 18, 19, pl. xix) on the other hand, completely defeated my efforts. After cytological examination, which determined that the Caninae were endowed with a chromosome complement of 28 (haploid numbers) and the Pimpinellifoliae with 14, I regarded these discrepancies as the cause of the unconformability. However, the acumen of Almquist in devising his table had not been lacking here. He points out that the leaflets on its flowering shoots have in no case advanced beyond the stage of the lower leaves of R. Traaeni. Any comparison of its leaf characters with those of other roses can only be with those of such lower leaves. If we recognise this, then it agrees with the non-committal and primitive Traaeni leaflet referred to previously as a possible point of derivation of all leaf forms.

*Rostrup in his Flora of the Faroes was completely baffled by this form; instead of naming it he simply says "perhaps canina"!

Further, he adds the interesting remark that its stem with its closely set prickles stops short at the second year's stage of other roses. In both of these points, leaflets and armature, R. pimpinellifolia reveals itself as a juvenile species such as is seen elsewhere in the genus Clematis. Another very important fact also emerges; although R. pimpinellifolia is the only rose known from Iceland, and although the Icelandic form is [259] indistinguishable from our Durham and Northern Scottish forms as I have satisfied myself by direct comparison, in the Faroe Islands* it is replaced by a very curious variety approaching the Villosae. This plant Almquist terms a half-Villosan and regards as a mutant of R. pimpinellifolia in that direction. Very likely in that view he is correct. In my mind, however, we are encountering a further step from the "juvenile" species R. pimpinellifolia to the adult of R. mollis. Nevertheless, let me hasten to explain, this does not exclude the possibility of mutation in any way. In the ontogeny of R. mollis it cannot be denied that the species evolves from its seedling stage to its final condition in a determinate direction. R. pimpinellifolia and R. faroensis (as well as R. orientalis), as has been pointed out, represent halts in this ontogeny, so that their phylogeny recapitulates the ontogeny of the Caninae. Hence, along vastly different lines, these microgenes conform to the usual orthogenetic development of the species within the genus. Only one point remains and this portion of my work is finished; it must not be overlooked that in the species-types common to the sections, we may be dealing with similar segregates rendered possible by crosses made ages ago between forms carrying the same sets of coupled unit characters.


Pollen Abortion in the Roses.

As I have just observed in the previous portion of this paper, the persistence of species-types common to all the sections may arise through segregation from previous crosses—a contingency demanding, amongst other considerations, special investigations as to possible hybridity. The first of these to [260] be set in motion were those attacking the matter from a cytological standpoint followed by an examination of the pollen conditions in the genus. Only the latter research will be discussed here in view of the enormous amount of material which has accumulated as a result of the earlier enquiry.

Very early indeed in the study of plant hybrids was their sterility observed, and determined in general to depend on their badly developed pollen. With the recognition of this fact, naturally enough, imperfect pollen was soon regarded as an adequate test of hybridity, as Dutrochet insisted in 1832. Moreover, Gaertner elaborated the point a little later, and stated it in terms differing but slightly from those current today. To test the reasonableness of these is far from difficult. If one takes pollen from known natural hybrids like those between Linaria repens and L. vulgaris (= L. sepium), between Rubus idaeus and R. caesius, between Rosa mollis and R. pimpinellifolia, between Primula officinalis and P. acaulis, and examines it microscopically, a huge proportion, if not the whole of the grains, is found to be malformed or shrivelled, and thus shows a conspicuous lack of cytoplasmic contents. Not five minutes ago I plucked a wild hybrid between two Primulas and was astounded to find that quite 90 per cent. of its pollen was in this state — an extraordinary fact when one realises the usual conception of the relationship between the parents and one's easy assumption that, if any hybrid is completely fertile, this one is.

With these facts as a basis the contents of the pollen sacs of certain rose microgenes were critically examined, when they, too, showed in most cases the same type of pollen abortion, although the proportion of sound grains varied with the form. One of the worst offenders in this respect appears to be Rosa mollis var. caerulea (text fig. 2, c): a little better behaviour is displayed by Rosa flexibilis (text fig. 2, b) and so on until we reach Rosa rugosa (text fig. 2, a) when practically every grain is perfect.


Text-fig. 2. a. Pollen of Rosa rugosa x 300.
b. Pollen of R. flexibilis x 250.
c. Pollen of R. mollis var caerulea x 250.

These, and other observations, by the close analogy they bear to the circumstances ascertained to hold with recognised hybrids, urge one strongly to the view that similarity in pollen condition is correlated with similarity in genetical constitution. In other words, the conclusion becomes irresistible that the bulk of so-called rose species are hybrids likewise. The sole difference lies in the fact that in the known hybrids their condition is patent and betrayed by every point in their structure, whereas in most of the roses, owing to the closeness of the parent forms, the hybridity is latent. Thus it can only be manifested during that delicately balanced test of parental equality gametogenesis, when the lack of homology in the parental chromosomes ends in the result just described, i.e., pollen abortion.

Although only a few rose microgenes were submitted to cytological examination, in a much greater number the pollen was critically examined and the proportion of good grains determined; what were the results the appended table discloses.

[262 ]


Section-Species. 0% - 10%
good pollen.
10% - 30%
30% - 50%
50% - 75%
75% - 90%
90% - 100%
AFZELIANAE subcristata (all bad)
fugax (all bad)
coriifolia (type of Fries)
EUCANINAE Inconspicua
lutetiana x Lintoni
flexibilis fallens senticosa  
RUBIGINOSAE Hybrid Lacy Ashton
Hybrid Anne of Geierstein
AGRESTES [Borreri*]          
VILLOSAE caerulea (all bad)   submollis omissa mollis cleistogamous
TOMENTOSAE tomentosa (type of Smith) sylvestris
(all near 30%)
PIMPINELLIFOLIAE  Hybrid mollis x pimpinellifolia         pimpinellifolia
var. rosea
CINNAMOMEAE cinnamomea plena       cinnamomea [rugosa*]
SYSTYLA     arvensis      
TOTALS 14 12 4 2 3 4
* The microgenes placed in square brackets are placed by many other authors in different sections; the nomenclature, however,
agrees otherwise with that of British authors.


From this table the astounding position emerges that, of 36 of our wild and subspontaneous roses and two garden hybrids produced from them, only four possess perfect pollen, three approach these, whilst no less than 75 per cent. have more than half of their microspores imperfect, and still more significantly, some 40 per cent. of the whole can only display at best less than 10 per cent. of well-formed pollen capable of bringing about fertilisation. Lastly, seven forms, two certainly hybrids, produced practically no good pollen. And matters are much worse than the table suggests, for whilst only a comparatively small number of the microgenes showing aborted pollen were examined quantitatively, and therefore tabulated, all of those with above 50 per cent. good grains are included.

Whence, if pollen conditions may be invoked as a criterion of hybridity, it appears certain that almost all of our wild roses are latent hybrids, Rosa pimpinellifolia, R. mollis, R. omissa and R. senticosa being more or less honourable exceptions amongst the genuine natives, and R. cinnamomea and R. rugosa amongst the introduced and half-wild forms.

At the lower end of the scale the most noteworthy feature lies in the preponderance of the Afzelianae in the worst class, in the massing of the Rubiginosae in their neighbourhood, and the presence of the Villosa form caerulea in the same category. In all of these cases the matter is complicated by the occurrence of facultative apomixis. Despite this, as with all other microgenes, it appears almost certain that the pollen sterility is the outcome of latent hybridity, and the apomixis, therefore, a secondary effect of the same. Any discussion on this point will be reserved until the latter phenomenon comes up for treatment.

But if so many of our roses are thus contaminated, and Rosa pimpinellifolia,. R. mollis, R. omissa, R. senticosa, of the real natives, and R. rugosa and R. cinnamomea of the others, have to a very great extent escaped, there must be some reason for their favoured position. Let us carefully consider each of these [264] and its habits, to extract, if possible, any clue to their comparative immunity. Amongst them four sections are represented, R. pimpinellifolia belonging to the Pimpinellifoliae, R. mollis and R. omissa to the Villosae, R. senticosa to the Eucaninae, and R. rugosa and R. cinnamomea to the Cinnamomeae.

Rosa pimpinellifolia is the sole representative of its group within our islands; whatever variability it displays is slight, and no forms attaining the rank of a microgene exist. Again, its flowering period is manifestly earlier than its congeners'; furthermore, it prefers stations abhorred by its relatives, wind-swept moors, mountain scars, coast ravines, cliffs and sandhills being its chosen habitats. Hence, geographically, and in habit, R. pimpinellifolia is powerfully secluded from extraspecific influences, so that almost inevitably its purity is ensured to a very considerable extent. However, its seclusion is not perfect; in many northern localities it encounters R. mollis and R. omissa, not only geographically, but also in season; more rarely it meets with the Eucaninae and Afzelianae. Hybridity with all of these is not excluded, but so remote are its affinities from them that almost uniformly its hybrids are patent and non-fertile; they cannot perpetuate their kind, and do not form a significant factor in our rose flora. Their rarity is even more pronounced in Central Europe where R. pimpinellifolia rarely collides with possible partners. R mollis and R. omissa are similarly isolated. Northern in range, and early in flower, their purity is, in general, maintained but is far from being absolute. They can intercross to some extent, but only slightly, with the Eucaninae and other groups, so that the Villosae, in the main, produce perfect pollen. Still, forms like R. caerulea, undeniably a member of the Villosae, do crop up and hint by outward characters, as well as in their pollen, at their development from previous crossing.

Rosa senticosa, however, proves more difficult of solution. A typical Eucanine, it grows in the same districts and flowers [265] with its relatives; it may, therefore, cross freely with them. Nevertheless, one feature shuts it off from them. Its curiously lower habit of growth renders it an unfit subject for the hedges and thickets in which its nearest relatives revel; it delights in forming a low scrub of compact little bushes in deserted localities. Pollination, then, if the microgene is at all prone to allogamy, will most likely take place through the agency of low-flying insects which, only in rare instances, will have visited other Eucaninae, Afzelianae and the like; therefore contamination, although not prevented, is at a minimum.

Rosa cinnamomea and R. rugosa need not detain us long. Both are forms isolated geographically and otherwise, R. cinnamomea being Arctic and Alpine and possessing no closely allied relatives, and R. rugosa being a native of Japan. Seclusion even more powerful than with R. pimpinellifolia is at work to prevent their crossing with other forms, and thus their condition harmonises with that of that species.

If our explanation of the superior genetic purity of these forms, as deduced from the state of their pollen, is warranted, a necessary corollary immediately arises. Monotypical genera and species completely secluded by any circumstance whatever from outside influence should invariably produce perfectly formed pollen grains.

To test this I have prepared and examined pollen of wild Narcissus pseudo-narcissus (from an isolated colony growing on the heather in Farndale, North Yorks, far away from any cultivated form), Pyrus Aucuparia (likewise from the moors) Primula farinosa, Trientalis europaea, Ulex europaeus and Parnassia palustris, all far removed from the disturbing influences of allied species. In every case the microspores were absolutely perfect.

The circumstances attending Pyrus Aucuparia are particularly interesting. In Europe it is an isolated species, but in Canada it has several related representatives like Pyrus americana and P. sambucifolia, and in these the pollen is far from normal. [266]

In addition, if this holds good, the converse ought to be true: wherever any genus is represented in a given locality by many related species their pollen, by its variable size and cytoplasmic content, should suggest hybridity. Such is indeed the case has been demonstrated by Jeffery for the Onagraceae, Hoar for the Rubi, Standish for the Crataegi and Cole for certain cultivated Rosae in America. Of these I have made an independent examination of Epilobium (a notorious hybridiser in the Onagraceae) and of the Durham Rubi, and my results agree exactly with the accounts given by these workers.

Thirdly, a totally distinct consideration of the genus Rosa tends to confirm the same view. If roses of such remote affinities as Rosa pimpinellifolia and R. lutetiana, Rosa Rugosa (from Japan) and R. foliolosa (from Texas) can be successfully crossed; if such complexes as the Gottfried Keller briar composed of {H. P. (Pierre Notting x Tea Madame Berard) x Persian Yellow} x (Madame Berard x Persian Yellow) can be built up; if this latter can then be crossed with satisfactory results with the H. P. Charles Lefebre, what possible obstacle can there be to intersectional crossing between the Afzelianae, Eucaninae, Tomentosae, Villosae, etc., or intermicrogenic crossings between the various members of any one of these? The failure to hybridise is unthinkable. Its occurrence is simply masked by the close relationship of the parents and the fertility of the offspring. Only rarely would it be detected otherwise than by the state of the pollen. However, I have performed this feat once myself in the case of a cross described later between R. lutetiana and R. Lintoni where the hybrid, which ran down in the books to the dumetorum form aciculata, grew sandwiched between what careful study in the field proved to be its parents. Another hybrid, R. mollis x R. canina forma, involving the Eucaninae and Villosae, was brought to my notice by Traaen; this he collected under circumstances much the same at Brevik in Norway. [267 ]

Finally, to prove experimentally that crossing between various members of the supersection Caninae was possible, I transplanted several microgenes to the garden where they flourished and blossomed freely. After castrating several of their flowers, I cross-pollinated them and enclosed them in strong paper bags. Fertile seeds were obtained in nearly all cases, and their numbers and possible parentages are set out in the table on the next page.

The seeds thus recorded have been planted and appear to be germinating normally; the resulting plants, if hybrid in nature, will not be available for study for a long time.

A glance at the table will prove conclusively that hybridity, in spite of its ostensible absence, is at least feasible within the limits of the Caninae, and serves to confirm the idea deduced from the pollen condition and the irregular course of its development, backed up by other weighty facts, that within the Rosae we are dealing with multitudes of latent hybrids.

* Although to my mind this does not explain the species-type, it may nevertheless be the correct view. Independently of these possible linked genes others do exist in Rosa. That a detailed analysis of their genetical behaviour has not been made depends on the slow growth of the plants; at any rate the attempt is being made. The following characters appear to be linked: glandular leaves are always biserrate; forms with deep red flowers have hairy foliage; erect and suberect, persistent and subpersistent sepals always accompany woolly flat heads of stigmas and early ripening fruit, whilst dilated and broad bracts are correlated with short peduncles.

Thus the production of new forms is rendered possible whenever cross-pollination brings together gametes of different potentialities. This would account for much of the variability of the genus and would, at first sight, offer some explanation of the segregation and passage of the common species-types throughout the groups. We have only to imagine in the production of various inter-Canine crosses the interaction of a series of linked* genes, involving the characters proper to any given species-type, and the whole matter is elucidated. A moment's thought, however, recalls the fact that any special species-type passes through species the purity of which is unassailable, and occurs in regions so far asunder as to place the original crossing, explanatory of the present segregation of the type, far back in Miocene and Pliocene times. Thus the hybridity, the existence of which we have almost certainly proved, whilst capable of explaining some of the peculiarities of the persistent species-type, cannot account for all. The species-type idea is absolutely independent of hybridity, although that phenomenon may tend to its perpetuation.

[268 ]

TABLE II. — Showing numbers of good seed per fruit in certain pure species and crosses.

(Average of 12)
(All fruit fell)
(Average of 15)
(All fruit fell)
(Average 9)
TOMENTOSA TYPE 31 Fruit full
No count made
LUTETIANA 7 16 16 Fruit full
no count made
GLAUCA VAR. REUTERI 13 (and many
chaffy scales)
PIMPINELLIFOLIA 35 36 Fruit full
no count
Fruit fell
33 15
RUGOSA Fruit full
no count
Pollen parent unknown—
Self-pollinated or apogamous?
27 and 7 chaffy scales
(Average of 12)
APOGAMOUS 8 Fruit full
no count

Hidden hybrids as the roses would seem, some at least are to be reckoned within the ranks of those classified under normal conditions as true-breeding. Excluding proved apomictical forms, Rosa omissa, R. mollis, R. vinacea, R. flexibilis, R. Lintoni, R. frutetorum, R. pimpinellifolia and R. rugosa have all been shown to reproduce their kind with reasonable fidelity within the permissible limits of fluctuating variation known for the group — an observation which, regarded in the light of the behavour of Burbank's many "compounds," cannot exclude their basal hybridity.

Pollination in the Rosae.

Investigating as I was the possible effect of cross-pollination in the roses, I was compelled to study the matter in the field, when some very interesting and illuminating evidence was secured. Very early indeed I discovered that, to say the least, pollination in Rosa was conducted under peculiar circumstances. Every morning at 7 a.m. practically every young flower, no matter what its species, provided that its stigmas were mature enough to receive pollen, was already pollinated, and this maturity, since the roses are homogamous, was almost always shown at that hour. Thus it appeared almost certain that, if pollination was effected by insects, it could only be though the action of Noctuidae flying at dusk and dawn, or through Diptera and Hymenoptera busying themselves at daybreak. To determine which was responsible I paid special attention an hour or so after sunset to the blossoms of the day, and to those just ready to burst. At that time, as if by magic, every flower young and old was folded up for the night. Unless then brought about by casual day-fliers like the Noctuids of the genus Miana the agency of moths must be ruled out. There remained then the operations of Diptera and Hymenoptera to be considered. I therefore got up earlier, at 4 a.m. (GMT), before any insects were at work, when I found that even then every newly expanded R. pimpinellifolia had its stigmas powdered with pollen from its own overarching stamens. At the same time those of R. mollis, R. omissa and R. Lintoni were quite untouched. A little later even their anthers dehisced, after which, unless insect guests performed the necessary operation, on the maturing of the innermost stamen whorls they curved over and deposited their precious dust. In many cases indeed even this curving motion as a mode of self-pollination proved superfluous, for almost without exception, as the flowers mature, several stamens are locked between the stigma heads so that their pollen, as it is shed, of necessity falls on the adjacent stigmas. By one method or another, therefore, in default of outside agencies, self-pollination is automatic.

* Unless the shining secretion on the exposed disk of the Eucaninae, Agrestes and Tomentosae is nectar the Rosae do not secrete that substance.

Independent of this definite mechanism for utilising their own pollen, the same effect is frequently brought about in diverse ways by insect visitors. Firstly, every flower bears its own colony of the widely distributed Taeniothrips primulae — an insect, contrary to the indications of its name, very impartial in its predilections. This Thrips rarely leaves its own particular flower, but spends its time in wandering indiscriminately amidst stigmas and anthers, and thus unavoidably carries pollen from the latter to the former. Secondly, insects from elsewhere, whilst in some measure rendering cross-pollination possible, since they are always pollen eaters and not nectar* gatherers, crawl at random from anther to anther, touching the stigmas as they pass, and deposit strange as well as adjacent pollen with the same result as before. Thus, to a very unexpected degree, the roses prove autogamous; that cross-pollination can occur, and in a manner likely to secure hybridity, the following journeys of the bees Bombus pratorum and Andrena trimmerana will give adequate proof.

(1) Journey of a worker Bombus pratorum: — Rosa mollis var. caerulea, R. pimpinellifolia, R. omissa.

(2) Journey of a second worker B. pratorum: — R. mollis type, R. mollis var. caerulea.

(3) Journey of a female Andrena trimmerana: — R. pimpinellifolia, R. mollis var. caerulea, R. mollis type, Saxifraga lingulata, Rubus idaeus, Aquilegia canadensis.

How generously the roses are patronised by insects, and therefore some degree cross-pollination possible, can be determined from the table on pages 272 and 273.

With such a crowd of guests as this, and guests of such cosmopolitan tastes, could one be surprised, even where autogamy and apomixis to a greater or less extent prevail, if hybridity became exceedingly frequent?

[272, 273]




Rosa pimpinellifolia Taeniothrips primulae Eristalis tenax
Syritta pipiens
Platychirus albimanus
Bombus pratorum
Andrena albicans
Andrena trimmerana
Pieris rapae
Rosa glauca var. subcristata Taeniothrips primulae Helophilus pendulas
Syritta pipiens
Rosa var. Lintoni Taeniothrips primulae Syritta pipiens
Polietes lardaria
Bibio laniger
Platychirus albimanus
Odynerus parietinus
Odynerus sp.
Bombus pratorum
Bombus terrestris
Rosa lutetiana Taeniothrips primulae Rhyngia
rostrata Syrphus ribesii
Volucella bombylans
Syrphus pyrastri
Syritta pipiens
Platychirus albimanus
Degeeria sp.
Lasius niger
Formica fusca
Megachile circumcineta
Coelioxys elongata
Apis mellifica
Bombus hortorum
B. pratorum
Halictus cylindricus
Miana fasciunula
Rosa dumalis Taeniothrips primulae Helophilus pendulus
Platychirus peltatus
P. aiblinanus
Syrphus inanis
Eristalis tenax
Bombus terrestris Simaethis fabriciana Psallus variabilis Panorpis germanica
Rosa dumetorum Taeniothrips primulae Syrphus pyrastri
S. inanis
Helophilus pendulus
Eristalis tenax
Rhyngia rostrata
Andrena trimmerana
Apis mellifica
Bombus hortorum
Megachile circumcincta
Halictus cylindricus
Rosa rubiginosa var. comosa Sericomyia borealis Bombus var. lucorum
B. pratorum
Acdrena trimmerana
Apis mellifica
Rosa mollis Taeniothrips primulae Syritta pipiens
Polietes lardaria
Bibio laniger
Platychirus albimanus
Syrphus ribesii
Helophilus pendulus
Vespa norvegica
Apis mellifica
Andrena albicans
A. wilkella
A. trimmerana
Bombus pratorum
B. var. lucorum
B. sylvarum
Ralietses cylindricus
Pieris rapae
Macroglossa stellatarum
Rosa mollis var. caerulea Helophilus pendulus
Degeeria sp.
Syritta pipiens
Bombus pratorum
Rosa omissa Taeniothrips primulae Helophilus pendulus
Lucilia caesar
Platychirus peltatus
Pyrellia lasiophthalmia
Syrphus spp.
Bombus terrestris
B. hortorum
B. sylvarum


In such a genus as Rosa one would scarcely anticipate the occurrence of cleistogamous flowers, so little need does there appear for their existence. Their development in Viola spp., Oxalis acetosella, Lamium amplexicaule and, to a less extent, Drosera rotundifolia and D. anglica is quite understandable. With the rapid growth of the surrounding vegetation in spring these species run grave risks of being overwhelmed and thus prevented from receiving the necessary insects. To avoid this contingency this very specialized method of self-fertilisation has been evolved. No such fate seems to threaten Rosa omissa, so that the discovery of a considerable number of cleistogamous flowers in that microgene occasioned great surprise. In general structure these were quite comparable with similar flowers of Viola hirta.

Their ovate lanceolate sepals were perfectly normal, but the involute petals, of the usual shape when straightened out, were reduced in dimensions to 8 mm. by 4 mm. In colour, they were greenish white. Otherwise, both the male and the female organs were quite ordinary. Most significantly, the amount of fully developed pollen attained the surprising figure of 100 per cent. — a contrast with the normal flowers of the same bush which yielded 70 per cent. good grains.


The defective nature of the pollen in many rose microgenes seemed, at first sight, to constitute a very effective barrier to their reproducing their kind, and therefore stimulated enquiries as to how the affected forms did so. To this end bushes known to have produced very defective pollen were marked in July and visited in September. Without exception, these individuals then bore, if possible, a greater supply of hips than their companions. Further, in one instance, a Rosa rubiginosa, the plant grew five miles from another member of the genus. Only one explanation seemed to account satisfactorily for the observed facts, and that was that in Rosa some form of apomixis was prevalent.

To put this to the proof, in the following season certain blossoms on each of the shrubs transferred to the garden were deprived of their stamens (care being taken to remove those clasped between the styles) and enclosed in stout paper bags. The roses thus treated comprised representatives of Rosa mollis, R. mollis var. caerulea, R. omissa, R. rubiginosa var. comosa, R. coriifolia var. Lintoni, R. pimpinellifolia, R. rubiginosa Penzance hybrid Lucy Ashton. In the same year, in June and July, a considerable number of forms were similarly castrated in an unfrequented lane running along an old waggon way abandoned a hundred years ago. Remote from the colliery village as this lane is, too many people use it as a short cut to admit of the possibility of leaving the twigs bearing the experimental flowers exposed in paper bags. I therefore very carefully cut the stigma heads off the mutilated flowers. The bushes so dealt with were R. mollis, R. omissa, R. tomentosa var. sylvestris, R. lutetiana, R. dumetorum, R. coriifolia var. Lintoni and R. glauca var. subcristata. In the garden all of the fruit fell save for those on the two Rubiginosae and certain of R. var. caerulea. On the waggon way matters did not pursue the same course; by August 8th all of the Villosa forms had fallen, and by August 15th the R. tomentosa var. sylvestris and R. var. Lintoni had followed. Only hips of R. dumetorum, R. lutetiana and R. glauca var. subcristata thus remained. A month later, of these, R. lutetiana and R. dumetorum alone persisted, accompanied by a solitary R. glauca. In October the surviving glauca hip had vanished; then, since they were now ripe, the two lots of Eucaninae fruits were removed for dissection. When examined, not a single fruit contained anything save a few useless chaffy scales. In the Rosa caerulea from the garden the fruit contained eight seeds on the average against 20 in the case of those pollinated normally; the two Rubiginosa supplied perfect seeds not differing widely in numbers from those flowers left exposed to chance pollination, self, insect, or otherwise.

The position of the Rosa glauca var. subcristata needs special attention. Unlike what occurred with the Villosae, Tomentosae and Eucaninae, the beheaded fruits began to swell as usual and gave every promise of abundant seeds, as was disclosed when examples were opened for examination. The future wholesale failure was far from being suspected, and it was with considerable dismay I beheld the gradual dropping of the hips. The reason, however, was only too apparent after dissection; on account of the sessile head of stigmas, to ensure their complete removal, I had cut too deeply into the fruit so that decay and insect enemies could work their will unhindered, thus causing its premature fall with its store of half-ripened seeds.

From these experiments we perceive that at least Rosa mollis var. caerulea, R. rubiginosa var. comosa, R. rubiginosa, hybrid Lucy Ashton and R. glauca var. subcristata are apomictical. All the other microgenes tested would seem, at first sight, to depend for their successful seeding on pollination either by means of their own sound microspores or by foreign pollen conveyed by insects. However, experiments carried out during 1920 prove that all our local microgenes examined, save Rosa arvensis and R. pimpinellifolia, are to some extent facultatively apomictical.

Although not submitted to experiment, Rosa Borreri seems likewise to be apomictical. A certain shrub, whose pollen proved wholly defective when put under the microscope in June, was nevertheless decked with a brave array of crimsom globes in October. One of these carried at its apex the unopened flower bud. This seems proof positive that in this particular blossom some type of apomixis had taken place, and suggested that the other fruits originated in the same manner. Of course in this, as in all the other instances, the apomixis may be of the facultative order.

The proof of apomixis existing in Rosa does not rest solely on my work, for Dingler has proved it to occur in an unnamed Rubiginosa form, whilst Lundström demonstrated its existence in varieties of R. glauca and R. coriifolia.

A careful comparison of the degree of apomixis with the position of the affected forms in Table I., giving statistics for pollen abortion, tends to emphasise the very close connection between the two phenomena; almost without the faintest possibility of error they are the coupled effects of a common agency.

Now arguing from the analogy of known or patent plant hybrids, either produced artificially or detected in the field, we have urged that pollen abnormalities in Rosa and genera similarly affected depend on hybridity in a latent condition, or at least in a state not readily determined as such. This view was strengthened by the discovery that a dumetorum form, classed as R. var. aciculata, was in reality a cross between R. var. Lintoni and R. lutetiana, and that its pollen was exceptionally bad. Carrying the argument to its legitimate conclusion, we are compelled to take up the view that in the Rosa apogamy, or whatever form of apomixis they present, is a phenomenon originating in hybridity — an opinion advanced from other data by Ernst.

*Except a dubious example possibly referable to the Rosa stylosa.

Quite recently Gates has attempted to show that apogamy was intimately bound up with tetraploidy and octoploidy, whilst earlier Strassburger and Rosenberg, formulating the same idea in a cruder form, tried to trace a connection between apogamy and a high chromosome number both in species of the present genus and in those of related genera. In particular, Rosenberg insisted that the Rosa species with a haploid number of eight chromosomes — R. livida, R. cinnamomea, and certain R. canina forms — were "sexual," whilst those with 16 — R. canina var. persaliciolia and R. glauca — were apogamous. In the first place, Rosenberg's figures are absolutely fallacious. Every* rose I have examined appertaining to the Tomentosae, Afzeliamae, Eucaninae and Rubiginosae have been endowed with a diploid number of thirty five, and is, therefore, pentaploid, whilst the Villosae and Pimpinellifoliae possess twenty-eight, and are thus tetraploid, as Rosenberg would have discovered had he pursued their cytology further.

*i.e., if we regard the haploid chromosome number of seven encountered in R. arvensis and R. rugosa as the base.

These facts support, to some extent, Rosenberg's conception of the linking of apogamy and high chromosome number but nullify the attempt to correlate apogamy and tetraploidy quá tetraploidy — at any rate in this genus; for within its limits we have on the one hand the tetraploid* Pimpinellifoliae microgenes pollinate normally, and on the other hand the tetraploid Villosae apomictical. Except by analogy, since none of the British Rosae are octoploid species, their evidence neither confirms nor confutes the possible relation which Gates, on the basis of Strassburger's work, draws between octoploidy and apogamy in Eualchemilla.

Returning now to the dependence of apomixis on hybridity, let us attempt to discover how the latter can induce the former. Since the supersection Caninae and the allied Synstylae include both sexual and apomictical microgenes, and further since the former is the more primitive and normal state of the Rosaceae, it is a reasonable assumption to make that the apomictical roses are derived from the sexual types. Moreover, it seems a fair deduction from the foregoing work to regard many rose microgenes as latent hybrids. If they are hybrids they can be between forms whose physiological divergence is slight or between forms where it is great. Suppose a cross of the second type to take place. Arguing from the analogy of the Linaria hybrids, or, appealing to the animal world, of the two Oporabia dilutata-autumnata hybrids in which in the reduction division few or no chromosomes find partners, we may anticipate similar happenings in the F1 generation of our wide rose cross.

Hence, in the pollen formed, in place of the reduced number of chromosomes we have the full somatic number. This pollen, again, if the power of the Linaria pollen or of the autumnata-dilutata spermatozoa is any criterion, will be defective and therefore impotent — deficiencies leading to its collapse. In the same way, in megaspore formation we have a similar failure of reduction, so that when the egg nucleus appears finally, even prior to fertilisation, it may carry the full unreduced chromosome complement. What is more natural than to assume that, spurred on by the stimulus of heterozygosis, this cell is capable of development just as if it had been normally fertilised? Even were the pollen functionally active the position here is unaltered; the only difference would be that apomixis would be facultative.

Thus an apomictical race would come into being capable of little variation and therefore faithfully reproducing its kind, like the apomictical Rosa rubiginosa mentioned previously, the seeds of which have been planted and from them young shrubs grown.

If, however, the cross was between roses of less remote relationship, then in the maturation division of their F1 hybrids dislocation more or less serious of the ordinary mechanism may occur, some chromosomes finding mates, others vanishing and so on exactly as described above. Hence, whilst many microspores will collapse, others will possess nuclei carrying normal or nearly normal chromosome numbers. These latter will be perfectly active, and like the similar sperm of the autumnata-filigrammaria hybrids differing but little in power from those of a pure species. If events in the female apparatus follow a parallel course the mature embryo sac will, in many cases, contain the egg-cell and egg-apparatus complete and prepared for the advent of the male nucleus. Fertilisation is thus possible, and apomixis unnecessary. If, however, the chromosomes are the heredity bearers, and their distribution in the micro- and mega-spore development anomalous, then the potentialities of the fertilised egg-cells will vary greatly, and thus account for the huge range of variation seen more especially in the fertilised roses.

Should this be the correct explanation of the genesis and continuance of apomixis in the roses, then it offers an adequate explanation of constant hybrids in the plant kingdom, and also of the multiplication of "species" in such critical genera as Hieracium, Rubus, Mentha and Salix. [280]

With dominance in the apomictical forms, and dominance followed by segregation in the sexual microgenes, it is clear that the notion of the species-type remains unaffected.


Linked up in some way with apomixis, and depending like it on the presence of unreduced chromosome numbers in the cells of the egg-apparatus, in many roses these cells, as well as the egg-cell proper, develop into embryos. Thus Rosa produces for our benefit a further complication, that of polyembryony. Concerning this I cannot say more at this juncture, for work in this direction is not complete.


*Since this above was written Jebe of Christiana has indicated other similar hybrids.

So homogeneous in their traits are the various members of the supersection Caninae, and so delicately graded are the relations between the sections, that only in exceptional cases could intersectional hybrids be detected as such. If that is true of such wide crosses as those between sections, then the extreme difficulty of distinguishing hybrids of the various microgenes within the same section by ordinary macroscopic features will be plain. Further, if hybridity between parents so related does exist (and from what has been said previously we know that it does), only pollen conditions under ordinary circumstances will betray it. Nevertheless, even that criterion will fail to give the faintest indication concerning the microgenes which gave rise to any given plant. Clearly then, this chapter, which will deal with wild plants patently hybrid, cannot in general include inter-Canine crosses; still, as we shall discover, two* such are brought forward, although their recognition was due to exceptionally favourable conditions. Moreover, since the work described here was carried out in northern counties, where the only other section occurring in quantity sufficient to give a reasonable prospect of hybridity is the Pimpinellifoliae, we have to consider, in the main, crosses between it on the one hand, and the Afzelianae, Eucaninae, Rubiginosae, Agrestes, Villosae and Tomentosae on the other. To determine hybrids of such origin would, at first sight, seem to present little, if any, difficulty, so sharp is the break between the latter six sections and the Pimpinellifoliae. In spite of this, so used were rhodologists to the gentle way in which their multitudinous species appeared to merge into one another, that even in such cases they followed, without hesitation, their usual proclivities for the erection of new species. Any hybrid on its discovery thus passed with its real nature unrecognised into the wordy chaos which overwhelmed all rose forms—species, microgene, variety and hybrid alike. This being so, before referring to our local forms, we shall endeavour to trace the steps by which the real genetical character of these forms was ultimately established.

History of recognised Rose Hybrids.

*Words in square brackets added by me.

Their history may be stated to have commenced with the finding of Rosa hibernica in County Down. This rose, although collected by Templeton in 1795, remained undescribed until 1803 when he published a short diagnosis in the Transactions of the Dublin Society; independently. Smith followed this by another and only slightly more comprehensive one in English Botany, tab. 2196 in 1810. Borrer's detailed description in British Flora III, p. 231 (1833) supplements both but, significantly enough of his attitude towards it, he ends his remark with the following phrase: — "A satisfactorily distinct species intermediate in habit and in size of leaves between Rosa spinosissima {pimpinellifolia) and R. collina b. Woods, but in its fruit and in the presence of setas nearer to the former." As Woods applied the name "collina" to a Eucanine form this was an excellent representation of its relationship in the light of the knowledge of the "thirties" of the nineteenth century. Baker in his Monograph (1867) could add little to these descriptions but, nevertheless, brought forward new localities in England, as well as in Ireland, and [282] had, moreover, already published an account of his new variety glabra in his earlier "Review." Yet one remark in the "Monograph" seems, when viewed in connection with later developments, to be very illuminating. He says: "in general habit, when in flower, the ordinary glabrous English form has just the same sort of resemblance to typical canina that Doniana [Sabini]* has to mollissima [mollis]."

Six years later, in 1875, Dr. Christ, the Swiss botanist, once again raised the subject in the Journal of Botany, and deliberately propounded the question "What is Rosa hibernica Smith?" To this he supplied the answer, "Ce rosier est un hybride entre le R. canina, L., et le R. pimpinellifolia, L.," adding as his reasons its mixed characters (armature of spinosissima, fruit of canina, etc.) and, most important of all, its manifest sterility — the then recognised criterion of the interspecific hybrid. Furthermore, he pointed out that Fries had determined the glabrous form of Rosa hibernica (= var. glabra, Baker) to be Rosa spinosissima x R. canina.

At the time these views were not adopted universally, and those contrary to them were never absolutely confuted; nevertheless, no one to-day disputes their validity.

Events moved on similar lines with the parallel hybrid to this in which R. spinosissima [pimpinellifolia] and R. tomentosa (sens. latiss.) took part, for in 1809 Smith brought forward in Flora Britannica, as a new species, a Scotch rose which he termed Rosa involuta. This was speedily followed by the description of those alleged species by Woods in 1816, Rosa Sabini, R. Doniana and R. gracilis, to wit. In this case, the critical eye of Lindley perceived glimmerings of the truth very early, for, in his "Monograph" (1821), we find him asserting his doubts as to whether, after all, this is not a product of R. tomentosa var. mollis [R. mollis] — a remark repeated by Hooker in Flora Scotica.

Still more allied forms turned up, Rosa Wilsoni erected by Borrer in 1835, coronata collected by Crepin in Belgium in 1858, R. var. Robertsoni described by Baker in his "Review" in 1864, and R. gracilescens, R. Smithii, R. laevigata and R. occidentalis by the same worker in the Monograph in 1869.

In none of these cases, save R. Wilsoni, was hybridity suspected. That curious form, however, aroused suspicions in the mind of Borrer, for in speaking of it he asks "Can it be a hybrid product?" and follows this with the remark, "Mr. Wilson finds several bushes of it, which discourages such an idea."

Thus a fact that, after due examination by a competent geneticist, would have strengthened his earlier suspicions served to lull them with Borrer. Had the phenomenon of heterosis been recognised then, the remarkable speed with which these hybrid roses spread vegetatively, and in doing so give rise to clumps independent of the parent bush, would inevitably have directed attention to the fact that all the Wilsoni bushes occurred along the Menai Strait, and radiated from one centre.

Once again Christ supplied the key to the derivation of Rosa involuta and its satellites. By careful analyses of its characters, coupled with other pertinent facts, he proved beyond cavil that it had sprung from a crossing of Rosa spinosissima and R. tomentosa (sens. latiss.)

If we remember that in making these decisions the widest view has been taken of Rosa canina regarding it as equivalent to the Afzelianae + Eucaninae, and of R. tomentosa by looking on it as including the Villosae and Tomentosae, on the same basis one more combination is possible, and that is Rosa pimpinellifolia x R. rubiginosa. Precisely as in the other two instances, roses of this parentage had actually been encountered, Rosa biturigensis described by Boreau in 1857, and R. involuta var. Moorei by Baker in 1869. Moreover, the relationship of the former plant had already been determined in a fairly accurate sort of way, since Boreau terminates his description with these words, "Looks like R. pimpinellifolia but nearest rubiginosa in its characters, differing therefrom in its prickles, earlier white flowers, and smooth peduncles, etc. This form occupied a great area, and if it were a hybrid as Schultz asserts, one would scarcely anticipate its being more abundant than its parents."

Once again we learn that the stimulus imparted by heterozygosis had sufficed to emphasise an origin correctly determined from other considerations to be hybrid, but that lack of knowledge of such an impulse had caused its indications to support the opposite view.

Baker in the same way recognised the true affinities of his plant for he says, "I cannot tell whether this is more like Sabini or rubiginosa..... It comes very near to the French Rosa biturigensis," Recollecting that Sabini is a pimpinellifolia-tomentosa hybrid, we are bound to confess that this closely approximates the truth, although the resemblance to Sabini caused its description as an "involuta" form.

In spite of all these shrewd surmises, it was left to Crepin to point out 25 years ago the exact value to be attached to these plants, and his opinion was confirmed by Marshall's discovery of a biturigensis form in Kent, which could only arise from a crossing between the Rubiginosae and Pimpinellifoliae. A little later, in 1897, Barclay gathered the same hybrid in Perthshire, and further repeated his good luck at Port Seaton, in Haddington, in 1910. In addition, Prof. Traill detected the same plant at Turriff in Aberdeenshire, as did Miss Hayward at Melrose.

Here again the powerful stimulus of heterozygosis had played its part, and, to such an excellent observer as Barclay, the fact caused much thought. I cannot do better than quote his exact words in discussing his find. "I had the good fortune to fall in with a rich colony of the hybrid, consisting of twenty or thirty great clumps spread over nearly a mile of the coast. ... It has often been remarked that hybrid plants frequently excel their parents in size and vigour. This was strikingly exemplified at Port Seaton. The hybrid there forms magnificent bushes, [285] growing to a height of ten to twelve feet, twice the usual height of the sweet briar, which again is usually a taller plant than the Scotch rose. Like the other hybrids which I have mentioned it forms clumps rather than bushes. In this point we see clearly the influence of R. pimpinellifolia. This sends out numerous suckers from its root, which themselves produce roots, and soon form a thick cluster of daughter plants around the parent stock. All its hybrids partake of the same character, and I have seen a clump of R. involuta extending for a distance of over 20 yards, the whole evidently having arisen from one original single plant."

Modern Conceptions regarding them.

*I only admit two, for I regard the first pair and the second pair as each forming an aggregate species.

Since Christ made his original determinations the study of roses has been revolutionised, and, in spite of the silly erection of multitudes of "sawdust " species, microgenes which actually exist, and therefore can be recognised in the field, have been described. Further, their systematic positions have been accurately made out. As a result we have the separation of the Afzelianae from the original Eucaninae, the Agrestes from the Rubiginosae, and the Tomentosae from the Villosae. No longer does it suffice to say that Rosa hibernica = R. pimpinellifolia x R. canina. R. canina, when that diagnosis was first made by Fries, was a very comprehensive aggregate. To British authors not obsessed with the splitting mania, it now breaks into four* collective species, R coriifolia, R. glauca, R. canina and R. dumetorum. Each of these possesses numerous microgenes, each of which may be involved in a cross with R. pimpinellifolia. Still, let us note, the opportunities for crossing are not very great; differences in habitat and time of flowering interpose very effective barriers, so that even to this day hybrids of this type remain very rare. They do, however, occur both in the north and in the south of our island. If they appear in the south, since R. glauca (agg.) [286] and R. coriifolia (agg.) are boreal groups, the second parent is bound to be either a canina or a dumetorum form. Should the leaflets be hairy beneath, then dumetorum is the second parent; if glabrous, then canina; which microgene of either is not determinable except under specially favourable conditions in the field, or when very striking features characterising any special microgene reappear in the hybrid.

*On May 29th, 1920, I discovered an undoubted glauca x pimpinellifolia on a ledge just under a raven's nest on Falcon Clints, Co. Durham. The only Rosae at that elevation (1,600 feet) were R. pimpinellifolia and R. glauca (forma?).

Matters in Scotland, Ireland and in the North of England generally are on a different footing. In these localities all four aggregates flourish, so that hairy leaflets simply assert that the plant is either dumetorum x pimpinellifolia, or coriifolia x pimpinellifolia; glabrous leaflets, on the contrary, imply glauca or canina x pimpinellifolia. Fortunately enough, seclusion of special forms in mountainous districts, or in coast ravines, occasionally assists in a diagnosis; so, too, does the almost universal impotency of glauca pollen. Therefore, whilst the occurrence* of a glauca x pimpinellifolia cannot be referred to as impossible, any glabrous hibernica, no matter where gathered, is almost certainly the product of a crossing between R. pimpinellifolia and some canina microgene.

Again, but rarely indeed in my experience does any coriifolia form come into contact with pimpinellifolia, so that the matter is narrowed down to circumstances not differing widely from those in the south. Nevertheless in one or two ravines on the Magnesian Limestone in Durham, the rose flora comprises only Rosa mollis, R. coriifolia var. frutetorum, and R. pimpinellifolia var. spinosissima. Any so-called hibernica there must be generated from the setose-peduncled form of R. pimpinellifolia and R. var. frutetorum, as in the case of the specimen figured in Plate XII.

With Rosa involuta and its allies matters are complicated in much the same way. Christ, as we mentioned, looked upon [287] it as a tomentosa x pimpinellifolia hybrid — a designation now wholly insufficient. The R. tomentosa of those days, although shorn of the R. mollis, was yet wide enough to cover many microgenes now transferred to the Villosae. Even allowing for this far-flung R. tomentosa it is as certain as anything can be that the names involuta, Sabini, Doniana and occidentalis, if not others, were applied to plants of mollis x pimpinellifolia parentage as well as to tomentosa (agg.) x pimpinellifolia. Of course, southward of a line from Yorkshire to Wales, the latter origin was certain: northward of this nothing was sure; mollis, omissa, tomentosa in the guise of any of their segregates might very well enter, although in descending order of frequency; firstly, because the period of flowering in R. mollis more nearly synchronises with that of R. pimpinellifolia; secondly, because those two species, as in the coast denes of Durham, occupy common habitats, and lastly, because tomentosa vera and its microgenes thin out with extreme rapidity northward.

In Durham and Northumberland, R. rubiginosa is certainly not a genuine native, and never, as far as I know, comes into contact with R. pimpinellifolia; these counties, therefore, cannot produce R. biturigensis or any of its varieties. For the specimen figured (Plate VIII. A) I have to thank Mr. Barclay, and judging from its general appearance, I have very little doubt but that in this case, and in the vast majority of others, the hybrid had for its parents R. pimpinellifolia and the R. rubiginosa form comosa. In the species-section Agrestes no hybrids have been reported.

The only other wild rose phen-hybrids known to me are inter-Canine, and they include the lutetiana x coriifolia referred to above and the mollis x canina from Norway. At this juncture I have no intention of going into minute details of these or of the numerous other hybrids I have seen growing naturally, nor indeed do I intend, in general, to give full descriptions at all; whatever remarks I have to offer will be reserved for the concluding portion of the paper, when, for the first time, many of these crosses are figured. [288]

Their uneven Geographical Distribution.

One very striking feature of the rose hybrids we have considered thrusts itself upon us, and that is the manifest inequality of their geographical distribution. For every single hybrid shrub discovered in the South of England, dozens have been detected in the North and in Scotland and Ireland. As I have explained, this proceeds from the nearness of mollis and pimpinellifolia in their flowering season, and secondly from the lateness of spring in these latter areas tending to mass, as it were, the periods for blossoming of all the forms very close together. Great as this disparity is with us, it is still further accentuated on the Continent. There, rose hybrids of the patent type are extremely unfrequent, and this, I think, depends on an exaggeration of the facts governing their occurrence with us, involving a much more sharply marked period of flowering in the earlier pimpinellifolia, aided by the restricted area inhabited by the northern Rosa mollis.

With a renewal of northern climatic conditions, as in Norway, hybrids in the genus attain the same frequency as in the northern poitions of our island. Nothing could surpass the magnificent set of mollis x pimpinellifolia hybrids I have obtained from various stations in the west of Norway.

Their Powers of producing Good Seed.

Probably one of the causes leading to the tardy admission of the hybrid nature of the forms discussed above was the erroneous ideas which were curent as to their fertility. This fertility was so contrary to the preconceived notions of hybrid sterility which held sway that it served to mask the real state of affairs. To a very great degree the vogue of the idea, and the failure of books to disclose the almost uniform barrenness of recognised rose hybrids, arise from the fact that, as just shown, such hybrids are infinitely more prevalent in the north; in consequence, they are usually collected by visiting botanists shortly after the fall of the sepals. Then, owing to the effects of heterosis and the exceptional display of flowers provided by the bushes, everything seems normal. Visit the same bushes [289] a very short while later, and note the changed scene. Gone is the plentiful promise of fruit, and all that remains for our inspection is a shrub abundantly endowed with leaves and a site strewn with countless shrivelled, immature fruits—at least this has been Barclay's experience as well as mine. Figs. 28-31, Plate XXII., will illustrate the stage at which the shrubs most accessible to me cast their hips.

Personally, I have only encountered two fertile wild hybrids, one the lutetiana x corrifolia var. Lintoni already mentioned, and a mollis x pimpinellifolia from Corbridge, Northumberland. Fruits of the former are to be seen in Fig. 25, plate xxi., and of the latter in Fig. 30, plate xxii., whilst those of their putative parents appear in Figs. 8, pi. xvii.; 16, 17, 19, pl. xix.

Mr. Barclay's experience of fertile hybrids is even more limited than my own. Whilst, favoured by his more northerly home, he has inspected many more hybrid bushes than I, he has only fallen in with one fertile specimen, the parentage of which he was able to ascertain with absolute precision to be R. omissia x R. pimpinellifollia. The shrub in question, classed as an involuta form, grows near Auchterarder railway station, Perthshire, and a fruiting twig plucked from it is shown on Plate X., and fresh 1919 hips on Figs. 21 and 22, whilst Fig 23 (plate xx.) depicts last season's fruit from the actual parent omissa. With his Haddington pimpinellifollia x rubiginosa matters are apparently the same, and the plant yields an enormous supply of perfectly ripened fruits. However, if one opens them another tale is told; only in a few cases do they contain more than a mass of chaffy scales. Barclay, nevertheless, gives the facts concerning a pimpinellifolia x rubiginosa reared artificially by Wilson, of St. Andrews. This undoubtedly fruited satisfactorily, and, moreover, its seeds germinated so freely that an F2 generation has been grown. One plant of this lot leant unmistakably towards R. pimpinellifollia. Many other garden hybrids of R. rubiginosa, like Lucy Ashton, Anne of Geierstein, etc., which [290] I have examined, likewise yielded fertile seeds. The connection between these observations and those of Barclay, in view of the phenomenon of apomixis displayed by R. rubiginosa, seems significant.

Lastly, in the lengthy series of examples of mollis x pimpinellifolia and of mollis x canina (forma?), received by me from the Norwegian rose student Traaen, practically every hip dissected had its quota of neatly shaped, well filled seeds; very scarce indeed were the chaffy scales so numerous in the majority of hybrid fruits. Considered along my fertile mollis x pimpinellifolia cross, these Norwegian specimens from Mostero, etc., by their fertility, throw some light on the relationship between the component forms, and suggest that R. mollis and R. pimpinellifolia are much more closely allied than one naturally expects. If we couple this suspicion with the fact that R. pimpinellifolia can throw a semi-Villosan form in the Faroe Islands, the near genetical dependence of the one on the other seems well established.

Ripened fruits and fertile seeds, therefore, in the widest rose crosses, are the exception rather than the rule, and we must dismiss as inaccurate the optimistic views of Smith,


(1) R. pimpinellifolia x R. lutetiana.

I found a plant, unfortunately badly damaged by fire, referable to the above cross in a dene near Horden, Durham. It was growing at a wood edge where the tiny sandhill form of R. pimpinellifolia (var. spinosissima) came into contact with a mass of a rose with very dark glaucous leaves, running down to the variety glaucescens of R. lutetiana. To the latter the hybrid approached very nearly in the colour and serration of its glabrous leaves, although its armature was midway between that of glaucescens and spinosissima. This is the R. hibernica var. glabra of Baker. Almost certainly the rose recorded by me from Cowpen Bewley as pimpinellifolia x glauca in the same plant.

(2) R. pimpinellifolia x R. dumetorum. Plate XI.

This also occurred on the Magnesian Limestone in Durham and at different points. None of the bushes bore fruit on August 14th except that figured, and even in this case 95 per cent. had fallen. A careful study of this bush, and comparison with its neighbours carried out in the field, incline me to consider it as a cross between pimpinellifolia and R. dumetorum (sens. str.) None of the bushes attained the height of a normal R. dumetorum, but one very low one straggled over a tremendous area in the manner characteristic of rose hybrids; in connection with the former observation, the flattening influence of the prevailing north-east winds from the sea must not be lost sight of. Even oaks and ashes yield to it.

(3) R. pimpinellifolia x R. coriifolia var. frutetorum. Plate XII.

One of the coast ravines adjoining the Black Hall Rocks protects enormous thickets of R. mollis, R. coriifolia var. frutetorum, Ligustrum vulgare, and on the sandy patches R. pimpinellifolia var. spinosissima. The coriifolia leaflets were densely clad with a tomentum both above and below, and whilst undeniably more decided in its biserration than usual, the plant could only be assigned to the variety named. Near by grew the present hybrid in the shape of a young bush; attaching due weight to the surrounding plants and to its leaf, prickle and other features, its parents are as given. This rose is R. hibernica var. laevigata. I have a plant substantially the same from maritime localities in Norway.

(4) R. pimpinellifolia x R. rubiginosa var. comosa. Plate VIII.

The plant in the sketch is from the East Lothian colony of Barclay and was collected in September. At this date its fruit was fully developed and perfectly ripe, as is apparent from the figure, although its good seeds are few. For Barclay's remarks on this hybrid see page 284.

(5) R. pimpinellifolia x R. mollis. Plate IX.

The example shown was obtained at Corbridge, Northumberland, on July 16th, 1919. Even at that early date its fruits were reddening and contained an adequate stock of perfect seeds. The early ripening is in itself enough to determine the mollis parent, because that species exhibits its array of cernuous crimson globes weeks before its allies of the Tomentosa and Villosa sections. Even were this not so, the neighbouring bushes were practically all macrophyllous R. mollis which the hybrid resembles strongly in fruit, foliage and prickles. The central plant was linked up by underground stems with other clumps covering nearly a score of yards of a steep bankside.

On paper, the bushes fall under Rosa Sabini, but since that name covers hosts of crosses between R. pimpinellifolia and the Villosa and Tomentosa allies, it is quite an unsatisfactory designation.

(6) R. pimpinellifolia x R. mollis (from Norway). Plate XV. B.

Compared with the bulk of the British examples resulting from the same crossing, the most peculiar point about these Norwegian specimens is the preponderatingly microphyllous character of which the photograph gives an excellent notion. Although I possess about 40 examples selected from as many bushes, the majority depart but slightly from the plant in the figure. Its most noteworthy characters are its fertile seeds, rounded leaflets, long setose peduncles, prickly fruit, the strongly bipinnate pair of sepals and the strong reddish coloration of stems, petals, bracts and sepals, the latter feature recalling certain varieties of R. mollis. Furthermore, attention must be directed to the enormous and fairly regular development of the prickles of the main stem, as well as to the half-opened flowers, which caused Smith to apply the name involuta to parallel forms. This trick of exhibiting half-expanded blossoms is not universal; most that I have studied behaved quite normally. [293]

Two or three specimens, of which the inset on the left of Plate XV, B represents a leaf, are so abnormally microphyllous as to suggest as their possible origin a back cross between such a plant as I figure and Rosa pimpinellifolia.

(7) R. pimpinellifolia x R. omissa. Plate X.

Again I have had to rely on Mr. Barclay for the specimen placed before us in this plate. It was collected from his fertile Auchterarder bush; from its fruit, foliage, and glandular development, assisted by a reference to the bushes growing with it, Barclay satisfied himself that its parentage was that just stated. Its exaggeration of the prickle armature on fruit and peduncles is very noteworthy, and is typical of many pimpinellifolia hybrids. Since, in many instances, this surpasses in strength that of its parents, it caused much tribulation in the minds of early workers. No doubt the phenomenon is to be attributed to the action of heterosis.

(8) R. pimpinellifolia x R. omissa. Plate XIII.

Once more Northumberland provided the specimen illustrated. This was gathered near Slaley on the same day as the pimpinellifolia x mollis described previously. On that date, July 16th, 1919, the fruit shown was about the best left on the bush. Under the plant, on the clayey bankside, thousands of immature fruits lay rotting, and others were ready to follow them.

Relying on book descriptions, this plant would appear to be R. Sabini, and has been so called by such a competent observer as Baker. Comparison with Plate X. will reveal what a heterogeneous mass comes under that name. My determination of its origin depends on critical examination of the plant and its neighbours. Baker and Tate refer in their "New Flora," 1868, to the very bush, and localize it so exactly that no mistake is possible; this means that this plant has persisted in its original station for 60 years at least.

(9) R. pimpinellifolia x R. tomentosa var. sylvestris. Plate XIV.

The present plant grows in a shady lane leading from Hawthorn Dene to the village of Hawthorn, Co. Durham, and was first observed on August 13th, 1919.

Unlike the previous specimens we have brought under examination, it was a drawn-up, feeble-looking plant, and differed most obviously from all the other hybrids of the group in my possession by its bright green, heavily glandular foliage. At the date in question, its unripe hips were just commencing to fall. This suffices, in itself, to indicate that its second parent was not the R. mollis or R. omissa of the Corbridge and Slaley plants. Independently of this the Villosae and Tomentosae were but weakly represented in microgenes in the vicinity, although, just as in other coast denes, R. mollis covered acre after acre, and excelled any other form in individuals. The only Tomentosae variety about was R. tomentosa var. sylvestris, and this coincided in colour and type of foliage, and in gland and prickle development with our plant. In a flora one would see it determined as R. gracilis, so that if all plants passing under that name are genotypically alike they are derived from a crossing between R. pimpinellifolia and R. var. sylvestris.

(10) R. mollis x R. canina (forma?) Plate XV. A.

Without very close inspection, in the form of its glabrous leaflets, the flexuose nature of its internodes, the more or less reflexed sepals, this plant has a strong canina look. Examination of the setose fruit and peduncles, its woolly flattish head of stigmas, its copiously compound serrate leaflets, its dilated stipules, just as strongly insists on the influence of R. mollis, whilst the prickles are a very obvious compromise between prickles of the straightish, thin mollis type and the stout uncinate ones of the typical Eucaninae. Altogether, the plant is a good intermediate between the plants named. [295]

(11) R. coriifolia var. Lintoni x R. lutetiana. Fig. 25, Plate XXI.

This forms a stout luxuriant bush. In its flowers the colour of the corolla seems to be a delicate pink intermediate between the white of lutetiana; and the clear rose of Lintoni; similarly, too, the shape of the petals strikes the mean, for they are neither so broad as those of lutetiana nor so narrow as in Lintoni. In the calyx, the sepals are more strongly glandular and pinnate than those of lutetiana and to a great extent approach Lintoni. The latter form has a broad woolly head of stigmas sessile on the disk; this has a much narrower head, slightly larger, subspherical, far from sessile, and weakly hispid; in lutetiana, as we know, the protruding stigma head is nearly glabrous. The lengths of the peduncles are intermediate. The bracts, stipules and prickles, in colour, shape and other minor characters, are not so far removed from Lintoni.

The leaves are dark green and keeled and remind one of lutetiana; still, the teeth, whilst not so marked, bear the three or four glandular denticles of Lintoni. Influenced by the same form they exhibit beneath a feeble hairiness on the primary veins, but only rarely do glands appear; when they do, they resemble the reddish ones of Lintoni.

On July 21st. 1919, the sepals of the hybrid were horizontal, those of the adjacent Lintoni half erect, and those of lutetiana reflexed. At that time Lintoni fruits were the largest by far, and, as a matter of fact, both the hybrid and lutetiana were still flowering.

To sum up, as a bush and broadly speaking, it is liker lutetiana and was at first regarded as a phase of that microgene; in its minuter details it follows Lintoni.

I call it Rosa hybrid promissa.

For all the figures, except those in the text, and Plate VIII., fig. B, and Plate XV., I am indebted to Miss Monica M. Hull.

Armstrong College,
March 25th, 1920.





Fig. 9. Fruit of Rosa dumetorum (Eucaninae).

Fig. 10. Fruit of R. omissa (Villosae) just ripe.

Fig. 11. Fruit of R. omissa in October.

Fig. 12. Fruit of R. biserrata (Eucaninae) just ripe.

Fig. 13. Fruit of R. micrantha (Agrestes) Just ripe