Can Line Breeding Stabilize Dahlia Flower Color?

Why “stable color” is not a single breeding outcome

Copyright © 2026 by Steve K. Lloyd
All Rights Reserved

Most dahlia hybridizers have had the thought at least once. You have two dahlias that bloom the same color, the color you have been chasing for years. The idea takes shape on its own: keep the good dahlias and breed them back into the family. Repeat next season and the season after. Tighten the bloodline and the color should settle down. Become dependable. Come true.

The hope is reasonable, and the phrase that usually describes it is “line breeding for stable color.” Both halves of that phrase hide more than they show. Line breeding is not one procedure. It can mean recurrent selection within a family, deliberate self-pollination, or repeated crosses between siblings, and those are not the same act with different names. They move a population in genuinely different directions (Bender, 2022; Horn, 2002; Onozaki & Fujimoto, 2023).

Stable dahlia flower color is just as slippery. It can mean several things that look alike in a row of blooms but are not alike underneath.

Consider what a breeder might actually be after. A population that throws more seedlings in the target color, though not every seedling. A parent that passes its color along more reliably, though some offspring still wander. A color that is faithfully inherited but shifts under a cold autumn. A single plant whose flowers vary from branch to branch or drift as the season goes on. A color that rides through cuttings without showing up identically in every plant it lands in (Lawrence, 1931; Muthamia, 2024; Ohno et al., 2016; Onozaki & Fujimoto, 2023).

Those are not five versions of one problem; they are five different problems. And until you know which one you are solving, you cannot know whether any breeding method has worked, because each one is measured in a different way and answered by different evidence.

Dahlia flower color spans an enormous range of hues and patterns. The photo shows a mix of named cultivars and unnamed seedlings from the author’s garden.

Start with the simplest version: abundance. A breeder may simply want more seedlings landing inside a desired color range. This is a question about the whole population, and it is answered by comparing how common the target color is from one generation to the next.

Dahlia breeding has shown, directly, that repeated selection can move a trait at the population level. A five-generation program selecting for vase life carried mean vase life from 4.4 days to 8.0 days, and lifted the proportion of seedlings lasting at least seven days from 3.8 percent to 70.4 percent. That is a real selection response, and it is worth holding onto as a picture of what breeding progress actually looks like when the target is measured across an entire seedling population rather than spotted in a single standout plant. Keep an eye on how that program worked, too, because its method turns out to matter as much as its result, and we will come back to it (Onozaki & Fujimoto, 2023).

The trait in that study was vase life, not color (Onozaki & Fujimoto, 2023). So it does not tell us how fast color would move, or whether color would move the same way at all. What it does establish is a principle that carries over cleanly: a handful of exceptional seedlings is not evidence that the population has shifted. The frequency has to shift. One remarkable bloom is a lucky draw until the numbers behind it change.

The second meaning is a different thing entirely: predictability from the parent. A breeder may want a plant of a given color to hand that color down more faithfully. But the look of a single flower does not, by itself, tell you what that plant will transmit.

An old and unusual phenotype makes the point well. Early researchers studied a condition they labeled ab.-white, an abnormal white patterning that varied in how strongly it showed. Plants carried capitula, the dense flower heads dahlias produce, with different degrees of the white expression. Parents showing more of it tended to produce more affected offspring, and more severely affected ones. The tendency was real, and also loose. The parent’s appearance carried information without ever becoming a reliable forecast (Lawrence, 1931).

That looseness is the whole point. Improving your odds of a color and fixing that color in every seedling are different achievements, and a program can reach the former without ever arriving at the latter.

The third meaning turns from what a dahlia inherits to whether that inherited color holds up under different growing conditions. A dahlia can inherit the full capacity for a color and still show something different depending on how it is grown. Work on the red cultivar ‘Nessho’ found that colder conditions could push its red toward orange, a visible fade, and that some lines faded more readily than others. The color was inherited. The expression answered to the weather (Muthamia, 2024).

The researchers found the fading associated with a shift in pigment chemistry: faded florets accumulated more flavones, the paler compounds in the pigment pathway, and showed lower red color values than the red florets they were compared against. The faded florets also showed greater expression of a Dahlia flavone synthase gene called DvFNS, which supports its involvement without proving that it alone caused the change. That gives real insight into how this cultivar fades, but it does not establish one universal mechanism for every red dahlia, and it should not be read that way (Muthamia, 2024).

Red is only one of the many color expressions found in cultivated dahlias. Photo of an unnamed second-year seedling in the author’s garden.

The practical consequence is what matters here. This kind of instability is invisible to a study of inheritance alone. A line bred for a color that holds through the cold has to be grown in the cold to prove it (Muthamia, 2024). A red that looks steady in one season has told you nothing yet about the season that tests it.

The fourth meaning turns inward, to a single plant. Some dahlia colors are not even consistent across one individual, differing between flowers, between branches, or as the season turns. The ab.-white material behaved this way, its expression shifting from one flower head to another on the same plant even as it also passed through seed to the next generation. Later descendants in that study lost the most extreme grades that had shown up before, yet the odd affected petal still surfaced here and there. The phenotype had softened without being gone (Lawrence, 1931).

The bicolor cultivar ‘Yuino’ shows within-plant variation of another kind. In its mixed flower heads, red petals turned up mainly in sectors or around the outer edge rather than scattered at random, and the plants producing more red also carried more flavonoid pigment in their leaves. The pattern hinted that different shoots on the same plant leaned differently toward red, though the study stopped short of pinning down a single mechanism behind it (Ohno et al., 2016).

When a color behaves this way, one bloom cannot speak for the plant. An honest record may take several flowers, more than one branch, and repeated looks across the part of the season that matters (Lawrence, 1931; Ohno et al., 2016).

The fifth meaning is about what survives propagation. Dahlias are kept going through tubers and cuttings, which lets a breeder hold onto a promising genotype without first grinding it toward uniformity through seed. That is one of the real gifts of a clonally propagated crop. It lets the breeder preserve the genotype that made the plant worth choosing (Horn, 2002; Quagliotti, 1962).

Preserving the genotype, though, is not the same as preserving the look, once the color itself is unsteady. In ‘Yuino’, selection for a stronger tendency toward red did carry forward through cuttings and tissue culture. And still the propagated plants, and the families traced back to different mother plants, kept differing in how much red they actually produced. The tendency held, but the uniformity did not follow it. Clonal maintenance and clonal consistency turn out to be two outcomes, not one, and the only way to know you have the second is to grow the descendants out and look (Ohno et al., 2016).

These five goals overlap in practice, but not one of them stands in for another. More seedlings in the target color is not uniformity. A preserved clone is not a guarantee that every plant off it will look the same. Five questions, five kinds of evidence, five different ways to succeed or fall short.

Bicolor dahlias can display strikingly uneven distributions of color across a flower head. Photo of an unnamed seedling in the author’s garden.

None of this means dahlia color lacks inherited structure. It plainly has one. Controlled crosses made early in the twentieth century showed flower color segregating in clear patterns, and tied those patterns to different combinations of pigments: flavones, the ivory-toned flavones, and the anthocyanins responsible for reds and purples (Bate-Smith et al., 1955; Lawrence & Scott-Moncrieff, 1935).

The researchers of that era captured the inheritance with a set of factors they labeled Y, I, A, B, and H. These were inferred genetic factors, worked out from how traits sorted in the offspring, not genes identified at the molecular level. Their effects turned on more than mere presence or absence. Dominance mattered, and so did dosage, the number of copies present, with different combinations steering a flower toward yellow flavones, ivory flavones, anthocyanins, or blends of them (Lawrence & Scott-Moncrieff, 1935).

The same work suggested the pigment systems interacted. In some of the material, stronger yellow-flavone expression came with weaker anthocyanin or ivory expression, while the reverse pull looked fainter. Later chemical studies supported the broader idea that inherited color classes could be matched to distinct pigment chemistry. Even flowers that read as plain white or ivory did not form a single chemical type. Researchers turned up different combinations of compounds among progeny that looked nearly identical. A similar face could sit over a different chemistry (Bate-Smith et al., 1955; Lawrence & Scott-Moncrieff, 1935).

To explain the interactions, the early authors reached for shared precursors, oxidation steps, enzyme control, and competition among pigment pathways. Those were sound hypotheses for their moment, but they were never confirmed as molecular mechanisms (Bate-Smith et al., 1955; Lawrence & Scott-Moncrieff, 1935). The distinction still holds today. The observations remain useful. The proposed mechanisms remain provisional, and should be described that way.

This inherited architecture explains why color segregates in patterns a breeder can recognize. It does not reach beyond that. It says nothing about how repeated inbreeding would act on the dahlia genome as a whole, because dosage at a few color factors is a different matter from homozygosity, the extent to which corresponding genetic copies are alike rather than different across the entire genome (Anderson et al., 1992; Horn, 2002; Lawrence & Scott-Moncrieff, 1935; Sparnaaij, 1979). And it makes no promise that a color will stay put across environments, branches, seasons, or the plants raised from cuttings. Inherited structure answers part of the breeder’s question. It was never going to answer all of it.

Here is where the vase-life program comes back. Every generation was measured, the best progeny were selected, the selected plants became the next round of parents, and across five generations the population moved hard toward the target. Repeated selection is a powerful tool, and that program shows it working (Onozaki & Fujimoto, 2023).

It is tempting to treat that experiment as a model for line breeding. The more accurate description is repeated directional selection. The breeders chose plants on measured performance, generation after generation. What they did not do was self-pollinate the same plant repeatedly, cross sibling to sibling, drive up genome-wide homozygosity, or build a formal inbred line (Onozaki & Fujimoto, 2023).

A seedling population gives a breeder something a single exceptional flower cannot: a distribution to select from. Photo of unnamed first-year seedlings in the author’s garden.

Those distinctions carry real genetic weight. Recurrent phenotypic selection keeps favoring individuals that show a desired trait. Self-pollination makes one plant serve as both parents. Full-sib mating crosses related siblings. Clonal selection sets a chosen genotype aside and keeps it, with no requirement that it breed true from seed. All four can involve breeding related plants over and over, and none of them are genetically interchangeable. More relatedness does not by itself mean a population has gone homozygous. A narrowed family does not by itself fix the visible color. A preserved clone does not need to breed true from seed to be worth keeping (Bender, 2022; Horn, 2002; Onozaki & Fujimoto, 2023; Quagliotti, 1962).

Biology sets its own terms as well. Dahlia disc florets, the small tubular flowers at the center of the head, are protandrous, releasing pollen before the female part of the same floret turns receptive. That timing cuts down on selfing within a floret, though it does not make selfing impossible. Highly double flowers, prized for the show bench, carry fewer disc florets, which leaves fewer fertile florets to cross with, without rendering every double cultivar sterile (Horn, 2002; Quagliotti, 1962; Srinivas, 1993).

A breeding method has to fit all of this: the reproductive system, the trait in hand, the variation available, and the way the finished cultivar will be kept. A bold, easily seen trait with plenty of heritable variation may answer well to straightforward selection. A trait that swings with the environment may demand testing across conditions. A trait riding on one major inherited factor may call for a different design than one built from many small components (Bender, 2022; Horn, 2002; Muthamia, 2024; Quagliotti, 1962; Srinivas, 1993).

So the direct dahlia evidence supports one thing firmly and refuses another. Repeated measurement and selection can shift a population for a defined trait (Onozaki & Fujimoto, 2023). That much is settled. What strict selfing or repeated sibling mating would do to dahlia flower color is not settled by the evidence assembled here, because that experiment is not represented in it.*

Evidence boundary

* Conclusions about what has or has not been demonstrated refer to the 13 sources assembled and admitted to the closed evidence set for this article. They are not claims that no other relevant research exists elsewhere in the scientific literature.

The small tubular flowers at the center of a dahlia flower head are disc florets, which are involved in sexual reproduction. Photo of an unnamed first-year seedling in the author’s garden.

With the direct dahlia experiment missing, studies in other polyploid ornamentals can at least sketch the outcomes worth considering. What they cannot do is tell us which outcome dahlias will actually take (Anderson et al., 1992; Horn, 2002; Sparnaaij, 1979).

Repeated selfing in chrysanthemum ran into several separate walls. Some genotypes set selfed seed more readily than others, and among the seedlings that resulted, losses came at germination and again later as plants failed to reach flowering, with the severity swinging from family to family. A pedigree estimate of inbreeding did not fully track the observed biological response. The lesson for a dahlia breeder is about scope, not verdict: a real dahlia experiment would have to measure far more than color, watching seed set, germination, survival, vigor, and fertility, any of which could decide whether a line even continues. Chrysanthemum does not show that dahlias would lose ground the same way, or falter at the same stage (Anderson et al., 1992).

Freesia, another ornamental studied for the same question, painted a different picture. Some tetraploid progenies, whether selfed or crossed between siblings, came out more uniform than comparable diploid ones, and severe inbreeding trouble arrived later than a simple diploid expectation would predict. Researchers floated retained heterozygosity, tetrasomic buffering, and preferential chromosome pairing as possible reasons, though these were comparative interpretations rather than mechanisms shown in dahlias, and the relative uniformity on its own did not prove genetic homozygosity underneath. The same work also noted what polyploid breeding can cost: reduced fertility, slower growth, shorter stems, lower yield in some material (Sparnaaij, 1979).

Put the two side by side and the one firm conclusion is that no single rule governs them. Polyploid ornamentals do not always collapse under related mating, and they do not always shrug it off (Anderson et al., 1992; Horn, 2002; Sparnaaij, 1979). Two results pulling in opposite directions give no license to split the difference, drop dahlias in the middle, and call that a prediction. The honest use of this evidence is the narrower one. It tells a breeder what a dahlia experiment should measure, and stays silent on what that experiment would find.

The firmest thing that can be said is that repeated selection can shift the distribution of a measured dahlia trait. When a target is defined and the population is measured, repeated selection can reveal whether its frequency or average changes across generations. That is real, and it is a genuine tool in a breeder’s hands (Onozaki & Fujimoto, 2023).

What the direct evidence does not show is that repeated selfing or full-sib mating will fix a dahlia flower color. No experiment in the evidence assembled for this article follows a dahlia color line through those procedures while tracking genomic change, progeny uniformity, fertility, vigor, seed production, and the effects of inbreeding.¹

That gap cuts both ways, and it is worth being exact about it. The absence does not deliver a negative verdict. It does not show that strict line breeding cannot sharpen color predictability. It equally withholds a positive one.¹ The comparative evidence gives reasons to weigh possible gains against possible biological costs, and then declines to settle the dahlia case either way (Anderson et al., 1992; Sparnaaij, 1979).

What the analysis does deliver is a problem finally stated clearly. Population frequency can be told apart from full uniformity. Inherited transmission can be separated from consistency across environments, within a plant, and through cuttings. Recurrent selection can be distinguished from deliberate selfing or sibling mating. And the exact measurements a direct experiment would need can now be named. That is real ground gained, even though the headline question stays open.

Because it does stay open. The effect of strict selfing or repeated full-sib mating on dahlia color predictability is unresolved, and so is the practical bargain underneath it.¹ We do not yet know whether more predictability could be won without paying too much in fertility, vigor, or seed set. That is not a finding that line breeding fails. It is the recognition that “line breeding stabilizes color” was never one question. It was several, and they have not yet been tested together.

White and ivory are familiar visible categories in dahlias, even though appearance alone does not reveal the underlying pigment chemistry. Photo of “River’s Snow Cone” in the author’s garden.

The better question is not whether dahlia bloom color can be stabilized in the abstract. It is which kind of predictability you are actually after, because the answer decides what you measure and what would count as success.

For more seedlings inside a color range, watch the frequency of that color across the progeny rather than celebrating the one dazzling seedling in the row. For a more dependable parent, set the parent’s own color against the spread of colors in its offspring, and value the plant that keeps throwing a higher share of what you want even when it never gives you a uniform tray (Lawrence, 1931; Onozaki & Fujimoto, 2023).

The last three goals each ask for a particular kind of looking. If you want a color that holds under changing conditions, grow the line under the conditions likely to reveal the suspected shift. For a line selected against cold-induced fading, that means testing it in the colder conditions that bring the fade out (Muthamia, 2024). If you want steadiness within a plant, look at several flowers, more than one branch, and the stretch of season where the color tends to move, which matters most for the phenotypes that have already shown they can wander (Lawrence, 1931; Ohno et al., 2016). If you want a color that survives propagation, grow out a number of plants from cuttings and compare them, because holding the genotype is not the same as proving that every plant off it wears the color the same way (Ohno et al., 2016).

A full line-development experiment would ask for more still. It would track the target color across generations while also recording seed set, germination, survival, vigor, fertility, the distribution of the phenotype, and the relevant genomic change, the whole account of whether predictability was really building and what it cost to build it (Anderson et al., 1992; Horn, 2002; Lawrence, 1931; Muthamia, 2024; Ohno et al., 2016; Onozaki & Fujimoto, 2023; Sparnaaij, 1979).

The original question keeps its value. It simply works better once it is divided. The point is not whether dahlia color can be made stable in one broad sense. It is which kind of predictability you mean to increase, which breeding procedure you are actually using, and whether you have measured the result rather than assumed it. That is the honest end of the exercise: a confounding question, asked in good faith, that turned out to contain five meanings of stability and one central line-breeding experiment that is not represented in the evidence assembled here.* If you want to see how that answer was assembled, the companion posts trace every step.

The conclusions in this article were developed from a closed evidence set assembled specifically for the question of whether line breeding can stabilize dahlia flower color. Statements about what has or has not been demonstrated refer to this evidence set. They should not be read as claims that no other relevant research exists elsewhere in the scientific literature.

Anderson, N. O., Ascher, P. D., & Widmer, R. E. (1992). Inbreeding depression in garden and glasshouse chrysanthemums: Germination and survivorship. Euphytica, 62(3), 155–169.

Bate-Smith, E. C., Swain, T., & Nördstrom, C. G. (1955). Chemistry and inheritance of flower colour in the Dahlia. Nature, 176(4491), 1016–1018.

Horn, W. (2002). Breeding methods and breeding research. In A. Vainstein (Ed.), Breeding for ornamentals: Classical and molecular approaches (pp. 47–83). Springer Netherlands.

Lawrence, W. J. C. (1931). Mutation or segregation in the octoploid Dahlia variabilis. Journal of Genetics, 24(3), 307–324.

Lawrence, W. J. C., & Scott-Moncrieff, R. (1935). The genetics and chemistry of flower colour in Dahlia: A new theory of specific pigmentation. Journal of Genetics, 30(2), 155–226.

Muthamia, E. K. (2024). Selection of non-fading lines for high quality cut-flower production in Dahlia variabilis ‘Nessho’ by elucidating mechanisms of low temperature induced flower color fading [Doctoral thesis, Okayama University].

Ohno, S., Hori, W., Hosokawa, M., Tatsuzawa, F., & Doi, M. (2016). Petal color is associated with leaf flavonoid accumulation in a labile bicolor flowering dahlia (Dahlia variabilis) ‘Yuino’. The Horticulture Journal, 85(2), 177–186.

Onozaki, T., & Fujimoto, T. (2023). Breeding long vase life by crossing and selection for five generations in dahlia (Dahlia variabilis) cut flowers, and selection of fourth-generation line 003-15 with ultra-long vase life. The Horticulture Journal, 92(3), 308–322.

Quagliotti, L. (1962). Aspetti del miglioramento genetico della Dalia (gen. «Dahlia») [Aspects of the genetic improvement of the dahlia]. Rivista di Ortoflorofrutticoltura Italiana, 46(4), 370–378.

Sparnaaij, L. D. (1979). Polyploidy in flower breeding. HortScience, 14(4), 496–499.

These works provide broader context for dahlia breeding, the selectability of floral traits, and the biological mechanisms that can make flower patterns stable or unstable. They support the article’s framing but do not supply its central experimental answer.

Bender, K. R. (2022). State of the art of dahlia breeding [Master’s creative component, Iowa State University]. ISU Digital Repository.

Hosokawa, M. (2014). 花の模様形成メカニズム [Mechanisms for flower color patterns]. Bulletin of the Experimental Farm, Kyoto University, 23, 7–12.

Srinivas, P. T. (1993). Genetic variability in Dahlia (Dahlia variabilis) [Doctoral dissertation, University of Agricultural Sciences, Bangalore].

This article was created collaboratively by the author, a dahlia grower and educator, and three AI language models.

The author directed the structure, tone, scope, and emphasis of the piece; supplied all scientific sources; and retained full editorial control over the final text. The AIs assisted with summarizing complex technical material, suggesting phrasing, and organizing relationships among peer-reviewed sources provided by the author. It did not independently select sources or introduce unsupported claims.

All content was carefully reviewed, edited, and refined by the author to ensure scientific accuracy, clarity, and alignment with the Dahlia Doctor approach to evidence-based horticultural education.