Bull. American Iris Society, 156:79-83 (1960)
Differences Among Rebloomers

In a recent Bulletin1 of the American Iris Society it was pointed out that there were known rebloomers among several chromosome "families." These include the diploid Tall Bearded with two sets of twelve chromosomes each, the tetraploid Tall Bearded with four sets of twelve chromosomes each, the forty-chromosome Dwarf Bearded and the largely sterile forty-four-chromosome crosses between the chamaeiris Dwarfs and the tetraploid Talls. Irises with like chromosome counts generally cross the most readily and, of course, most frequently produce fertile offspring.

It would seem logical, assuming like produces like, that the crossing of two reliable reblooming irises of the same chromosome pattern would produce rebloomers. In general this is true, but unfortunately the number of rebloomers that may be expected to result from such a cross is rather low. It is the purpose of this article to comment on this difficulty, present a method of testing differences, and offer some additional observations which might prove helpful.

It is commonly recognized that certain irises will rebloom but for some reason fail to transmit this trait to their progeny. Certain other irises will not rebloom but do produce a certain number of reblooming offspring. In both instances one might say that the phenotype does not correspond to the apparent genotype. (The phenotype is what the iris is as a growing, flowering plant. The genotype is the sum total of all its genetic characteristics.) The obvious answer then, if we want rebloomers, would seem to be the simple one of selecting as parents those irises which have produced the greatest number of reblooming seedlings. Probably most hybridizers interested in developing rebloomers have compiled such lists for their own use from the various iris Check Lists. One such, listing irises most frequently appearing as either parents or grandparents of rebloomers was published in a recent issue of The Bulletin.2

The Rebloomer/Regular Ratio

From the above-mentioned list one will observe that Tiffany (Sass '38), although not itself a rebloomer, is the parent of four rebloomers. On the other hand Martie Everest (Kirk; McDade '35), a reliable old rebloomer, is conspicuously missing. The obvious conclusion would seem to be, if one wants rebloomers, to use Tiffany rather than Martie Everest as a parent. Unfortunately this is precisely the conclusion that some hybridizers have drawn.

Rather than to select parents on the basis of number of reblooming offspring, one should select parents on the basis of relative number of reblooming offspring. Examination of five randomly selected pages from the 1949 Check List showed Tiffany three times as a parent. Assuming this random sample to be representative of the 233 pages of listings would indicate Tiffany to be a parent of 140 registered offspring in the 1949 Check List alone. It has been estimated that only one or two per cent of seedlings grown are registered. However, since Tiffany is recognized as a fine parent, a conservative estimate for this iris might be one registration to twenty seedlings grown.

On this basis, multiplying 140 x 20 gives a grand total of 2800 Tiffany seedlings grown by all hybridizers. Of this total, four were fall bloomers, a rebloomer/regular ratio of 1:699. In contrast, of 18 Martie Everest seedlings planted in my garden in November, 1957, three were fall bloomers, a ratio of 1:5. In the case of Tiffany this is one-tenth of one per cent; for I Martie Everest it is twenty per cent. Viewed in this perspective, Martie Everest is unquestionably a much superior parent for the hybridization of fall bloomers. The point being made is that one should consider the reblooming/regular ratio rather than the number of reblooming seedlings as his criterion for selecting parents.

The Significance of Differences

Iris hybridizers frequently ask questions such as the following: "How many seedlings must I grow to recover such and such trait in the F2 generation?" "What are the chances that such and such is carried as a recessive?" "What percentage of seedlings may be expected to germinate?" "What are the chances of getting a 'take' from such and such cross?" Such questions are answerable only in terms of probability, as, for instance, "Your chances are one in sixteen of recovering the trait." All that this really means is that the mathematical probability of the first seedling possessing the trait in question is one in sixteen. If it does or does not, the chances of the second seedling possessing it are still one in sixteen and so on right down the seedling row. It is therefore possible that none of the seedlings might inherit it. If, however, the ratio is 1:16, we ordinarily conclude that it should be possible to recover the trait by growing sixteen seedlings. This illustration should make it abundantly clear that hybridizers interested in the transmission of certain inherited traits are necessarily involved in a certain amount of probability theory. There simply are no other terms which make sense with which to answer such questions.

Fortunately some of the most important tests needed to answer these and similar questions intelligently are clear and simple enough to be understood and applied by any of us. We do not need to be able to derive the test in order to use it. Like the saying, we don't need to be able to lay an egg to be able to tell whether it is good. This is true with the test for the significance of the difference between two proportions. To illustrate, assume that one is interested in developing tangerine-bearded black irises. He begins two lines of hybridizing—both between original crosses of different blacks and tangerine-bearded pinks. After several generations of seedlings, there appear in the seedling patch two black seedlings with tangerine beards from one line (line A), and five black seedlings with tangerine beards from the other (line B). Line A totals 983 seedlings and line B 1126. The question now is, should one hybridize only with line B since it produced the reatest number of seedlings? Or, to put it another way, is there any real difference between these two proportions, or is it a chance difference, with the proportions likely to be reversed the next time these same crosses are grown? Fortunately this kind of question can be answered quite easily in terms of probability simply by testing the significance of the difference between the two proportions. It requires only a knowledge of eighth grade multiplication and division plus possession of a table of square roots which may be found in the appendix of any algebra textbook. The computation3 may be set up as follows.

A test quotient this small means that, so far as these two lines are concerned, they are practically identical. One might use either of them with full confidence that he has lost nothing by so doing. If the quotient between any two proportions tested in similar fashion is 2.0 or larger, one may suspect that there is a difference not due to chance alone. A quotient of this magnitude means that there are about five chances out of a hundred that there is no difference between the two groups of seedlings. If the quotient is as high as 3.0 or more, one may be practically certain that his two lines are different because the likelihood of getting a difference this great by chance alone is one out of a hundred.

The numbers of seedlings used in the illustration above were intentionally made large in order to represent a realistic problem; the computations would appear to be much easier, of course, if small numbers had been used.

This simple test with its quotient limits of 2.0 (representing the five per cent level of confidence) and 3.0 (representing the one per cent level of confidence) may be applied to any two proportions of seedlings in exactly this manner and with the results to be interpreted similarly. Obviously there are more sophisticated techniques4 by means of which several variables can be examined simultaneously, but these require some knowledge of statistical theory. They are not, however, in any way more valid than the one illustrated above.

Horticultural Differences

It seems possible from observation to identify a number of horticulturally different varieties of fall-blooming irises, some of which are valuable in hybridizing for rebloomers, but others of which can lead to dead ends.

The first of these is an extremely slow maturing iris which does not bloom in the spring and sometimes not in the fall. Autumn Splendour and White Alone are examples in this climate. The former, when well grown, blooms regularly but only in the fall. The latter blooms irregularly, apparently at whatever time the rhizome has sufficiently matured which is never more frequently than once a season. Gay Hussar, Whisperwood, and Blue Valley have all performed in this manner for me. While such irises serve admirably in lengthening the blooming season they do not represent the kind of plant useful to hybridizers of rebloomers.

A second type, also not classifiable as a rebloomer in the strict sense, but nonetheless an iris which might prove useful to hybridizers is an extremely vigorous sort, possessing the tendency for rapid vegetative increase and new fall growth. It blooms in the fall far down in the fan, often with but a portion of the flower projecting from between the leaves. Buttercup Lane is a stellar example. Some of David Hall's unnamed pink seedlings will also behave in this way in this climate. They are perfectly hardy and necessarily so since nearly every rhizome will have a new bud and soft growth when freezing weather arrives. Irises like these are unusually vigorous and, if otherwise desirable, might well be incorporated into the breeding program.

A third type seldom if ever blooms in this climate. Actually it does break dormancy in the late fall. Its buds have developed just enough to be killed by the first hard freeze with the result that the center stalk is blasted and the vigor distributed to the side increases, provided it survives the winter. These increases, however, do not mature with sufficient rapidity for regular spring bloom and consequently the procedure is repeated the following fall. This seems to be what happens to many of the slower west-coast and southern varieties (so-called winter-bloomers) when these are tried in more rigorous climates. It is possible that the failures of numerous seedlings to bloom in the beds of hybridizers of regulars may be attributed to this condition. Dr. G. Percy Brown, Dean of the reblooming hybridizers, reports that it happens frequently in Massachussetts. Varieties which have for this reason failed to bloom in my garden include Bengal Princess, Rose Pearl, Ball Gown, Tournament Queen, Bronzino, Royal Band, and Fall Velvet Improved.

A fourth type blooms both spring and fall but actually should be classified as a once-bloomer since each rhizome blooms but once with the increases not maturing during the one growing season. The most mature rhizomes of the clump bloom in the spring; the less mature bloom in the fall. The increase of the spring blooming rhizomes perform the following spring while the increase of the fall blooming portion of the plant bloom the following fall. These irises are difficult to identify because, while many may be suspect, few can be "caught in the act." Often the spring bloom will be sparse, the clump seeming to produce only about half of what normally might be expected.

A fifth type, a genuine rebloomer, blooms early in the spring, often with the Standard Dwarfs. By the end of the June blooming season it has gone into a dormant stage and is ready for transplanting. During the middle of the summer it remains dormant and is often rather unsightly. Early in August it comes to life, rapidly shooting up fans and bloomstalks which perform until cut down by the first hard freeze. Polar King and Golden Harvest are excellent examples.

A sixth type, likewise a genuine rebloomer, blooms with the earliest, matures side increases with great rapidity, has a second full reblooming period in the middle of the summer, again matures its side increases and goes into a complete third cycle in September and October. Autumn Elf is the best example.

A seventh type, and one which most nearly approaches the everbloomer that we have, is the iris which constantly matures increase, sending up bloomstalks whenever the rhizome has matured sufficiently. Autumn Queen has performed in this manner every day from April until November in my garden. Martie Everest and Equinox usually have a short dormant period immediately following their first bloom, begin a second cycle in July and continue until the first freeze. Summer Surprise, Autumn Afternoon, and Fall Primrose do likewise, but being a bit larger take a bit longer to mature their bloomstalks.

Nearly all of the fall-blooming sorts must be transplanted to new soil regularly in order to perform satisfactorily. The exceptions have proved to be Polar King, Martie Everest, and Equinox, each of which has bloomed regularly for me without transplanting, spring and fall, for a period of ten years.

The following generalizations seem to apply.

  1. Avoid using as parents any irises showing a tendency to soft rot. A rebloomer with a beautiful flower but lacking a hardy plant is compost.
  2. Limit crosses to one chromosome family—24, 40, or 48.
  3. Cross only to rebloomers proved for your climate.
  4. Cross the best with the best and save the best.
  5. Avoid crossing plants with the same serious weakness.
  6. A non-reblooming seedling should be "selfed," crossed back to a reblooming parent, or to another strong rebloomer.
  7. "Selfing" rebloomers sometimes produces rebloomers, but also frequently results in narrow parts, poor form, color or substance, or lack of vigor. Lack of vigor often means loss of the reblooming tendency.


Correspondence, both direct and in robins, with others interested in developing improved rebloomers has revealed a surprising amount of success in totally unexpected directions. At least six seedlings have been grown from the supposedly sterile Autumn Queen. One person has reported success in setting pods and growing seedlings from unsettable October Blaze. One person reports several nice reblooming seedlings from crosses with Sangreal. Two different outcrosses of rebloomers to Pink Cameo have yielded reblooming seedlings. "Breaks" of this nature certainly suggest caution with regard to the formulation of any absolute rules. Apparently the presumed impossible is actually being accomplished all the time by reblooming pollen daubers blithely unaware that what they are doing just can't be done. Or perhaps they do know but are simply following the old motto, "The difficult we do immediately, the impossible takes a little longer."


  1. Smith, Raymond G., AIS No. 155: 27-34, 1959.
  2. Flight Lines, AIS No. 154: 74-75, 1959.
  3. The T test of any statistic is that statistic divided by its standard error. The formula for the standard error of a difference between proportions may be found in Peatman, John G., Descriptive and Sampling Statistics (New York: 1947) Harper and Brothers, p. 558.
  4. See Rao, C. Radhakrishna, Advanced Statistical Methods in Biometric Research (New York: 1952) John Wiley.