The generation code on a seed packet tells you how much your plants will vary. This part covers the later steps of breeding: stabilising a line from F3 to F6, inbred lines, selfed seed (S1), backcrosses (BX) and polyhybrids. It then shows how many plants and generations a breeding project needs, and how to read a seed listing line by line.
This is Part 2. Part 1: Cannabis Genetics Decoded: The Breeder Vocabulary Every Grower Misreads on Seed Labels
Key Terms From Part 1
- Genotype is the set of gene versions a plant carries. Phenotype is what you see and measure: height, aroma, resin, flowering time. Phenotype comes from the genotype plus the growing conditions.
- Each plant carries two copies of every gene, one from each parent. The place where a gene sits is a locus (plural: loci), and the versions of a gene are alleles.
- A plant with two identical alleles at a locus is homozygous there. A plant with two different alleles is heterozygous.
- Heterozygous parents pass on one allele or the other at random, so their seeds differ. This splitting is called segregation.
- F1 is the first generation from a cross. A true F1 comes from two distinct inbred lines and is uniform. F2 is the offspring of two F1 siblings, and this is where variation opens up.
- Selfing means pollinating a plant with its own pollen.
9. Third to Fifth Generation: Stabilising a Line Step by Step
From the F2 onward, the breeder picks the plants closest to the target and crosses them together. Each round of inbreeding roughly halves the share of loci that are still heterozygous.
| Generation | Expected heterozygosity under selfing | What you see in the room |
|---|---|---|
| F1 | 100% | Uniform, but genetically mixed |
| F2 | 50% | The most visible variation |
| F3 | 25% | Two or three recognisable types |
| F4 | 12.5% | One main type, with some outliers |
| F5 | 6.25% | Mostly uniform |
| F6 and beyond | Under 4% | Uniform enough for commercial use |
These numbers have two limits. First, they apply only to traits the breeder selects for. A breeder who selects for aroma but ignores vigour can lose vigour along the way. Second, halving heterozygosity does not mean reaching the target, because a plant can end up homozygous for the wrong allele.
As a rule of thumb, not a measured constant: F4 is the first generation worth planning a production run around. F5 and later are genuinely predictable.
What you can do: For a production crop, choose seed labelled F4 or later from a breeder who can show selection records. Records show which traits were selected, so you know which traits to expect to be stable. If you buy F2 or F3 seed, plan space and time for selection.
10. Inbred Lines and What “Stable” Really Means
An inbred line is a family of plants bred within itself until it breeds true. When you cross two members, the offspring look like the parents. The cost is that the inbreeding that makes a line predictable also tends to reduce vigour. That is exactly why breeders cross two inbred lines to make a true F1, which brings back vigour and stays uniform.
A finished line has few hidden recessive alleles left, so it works well as a parent for new crosses. A polyhybrid (section 13) is the opposite on both counts. It hides many recessive alleles and gives unpredictable offspring.
What you can do: When you choose a parent for your own crosses, prefer a plant from a line that has bred true in your own test run over a famous polyhybrid cutting. A parent that bred true for you will give fewer surprises in its offspring.
11. Selfed Seed (S1): How It Is Made and What It Costs
S1 is the first selfed generation: one female plant pollinated with its own pollen. To do this, the breeder treats part of the plant so it reverses and makes male flowers. For the methods, see Making Feminized Seeds. The cross has no Y chromosome, so all the seed is female.
However, an S1 is not a clone in seed form. The mother was heterozygous at many loci. Selfing splits those loci up in her seeds, much as an F2 does.
| What growers expect from S1 seed | Is it accurate? |
|---|---|
| All seedlings are female | Yes, unless stress later makes plants reverse |
| All seedlings are identical to the mother | No |
| Recessive traits may appear | Yes, commonly |
| Genetic diversity collapses in one generation | Not necessarily |
In 2020, Punja and Holmes published a study in Frontiers in Plant Science. They examined seed that formed on hermaphroditic flower clusters, which carry both female and male flowers. The seedlings from that seed were all genetically female. Seed from cross-pollination gave roughly one male to one female.
They also found no measurable difference in genetic variation between selfed and cross-pollinated populations after one generation. This contradicts the common claim that one round of selfing sharply narrows a line.
The hermaphrodite risk, stated precisely
Hermaphroditism is not a simple inherited trait, and its causes in commercial grows are still poorly understood. The same 2020 work reported that it appears without any treatment in roughly 5% to 10% of plants under commercial conditions. Earlier research showed that gibberellic acid (a plant hormone), silver compounds and environmental stress can all trigger it.
The real risk lies in selection. A plant that reverses easily under treatment has a low threshold for reversal. Selfing passes that low threshold on to its seeds. So never breed from a plant that produced pollen without treatment.
Translator’s note: Thai heat, rainy-season humidity and power cuts all add stress, so you may see untreated pollen more often than growers in cooler countries. The rule still holds. A plant that threw pollen on its own should not become a parent, even if it is your best plant otherwise. Clean its pollen sacs out of the room, because its pollen can still seed nearby females.
What you can do: Keep a written list of every plant that made pollen without treatment, and leave those plants out of all breeding. When you grow S1 seed, check the plants for pollen sacs regularly through flowering. That way you catch reversal before it seeds the crop.
12. Backcrossing (BX): Turning a Cutting-Only Mother Into Seed
A backcross solves one problem. You have an exceptional plant that exists only as a cutting, and you want seed that carries it. First you cross it with another plant. Then you cross the offspring back to the original plant, called the recurrent parent, again and again.
| Step | Label | Expected share of the genome from the recurrent parent |
|---|---|---|
| First cross | F1 | 50% |
| First backcross | BX or BX1 | 75% |
| Second backcross | BX2 | 87.5% |
| Third backcross | BX3 | 93.75% |
| Fourth backcross | BX4 | 96.875% |
These percentages are averages across the whole genome, not a guarantee for any single seedling. More importantly, a heterozygous recurrent parent can never be recovered perfectly in seed. A heterozygous genotype always splits again. BX3 seed of a famous cutting is a population that resembles it closely on average, not a pack of copies.
For a grower, a backcross gives seed that carries most of a valued cutting. Seed is easier to store than a living mother plant. If you want the exact plant, keep it as a mother (Mother Plants & Cloning) or in tissue culture, which means keeping plant tissue alive in sterile lab containers.
Translator’s note: Famous cuttings in Thailand and Southeast Asia usually pass from farm to farm. Buying a clone from another farm carries three risks: pests such as mites, hop latent viroid (HLVd), and getting the wrong cultivar. Hop latent viroid is a tiny infectious agent that stunts cannabis plants. Keep a new clone in a separate area until you have checked it. Ask a lab whether it can test the clone for the viroid.
What you can do: Start with a cutting you already own and have checked. Make one cross and one backcross on a small batch. Then grow a few BX1 plants beside the mother and compare them. This shows how close the seed gets before you spend more generations on it.
13. Polyhybrids: Why Many Strain Names Predict Little
A polyhybrid is a hybrid whose parents were also hybrids, going back several generations with no inbreeding in between. Almost every modern cultivar with a combined name is a polyhybrid. So the name stops predicting anything. Two breeders can make a cross with the same name from different parent plants. Their two seed populations may then have almost nothing in common.
In 2021, Watts and colleagues published a study in Nature Plants. They tested cannabis samples at over 116,000 genetic markers, which are known points on the DNA that can be compared between plants. They also ran chemical analysis. In total, 297 samples were analysed chemically, and 137 had both genetic and chemical data.
Genome-wide genetic structure explained only about 37% of the variation in the Sativa and Indica labels given to the samples. The researchers concluded that these labels do not reliably reflect how closely plants are related.
What the labels partly tracked was terpene chemistry, meaning the aroma compounds. Myrcene, a common aroma compound, was linked with the Indica label. It explained about 21% of the variation in labelling, one of the strongest single links in the study. This pattern traced back to clusters of genes that make terpenes (terpene synthase gene arrays).
So a polyhybrid name is a moderate guide to the lineage the breeder intended. It is a weak guide to aroma family. It tells you nothing about uniformity, or about how related the plant is to another plant with the same name.
What you can do: Judge seed by its generation code and the breeder’s records, not by its name. If you buy two packs with the same name from different breeders, grow them as separate trials and record each one on its own. Do not assume they are the same plant.
14. How Many Plants and How Many Generations You Need
Population size decides whether the genotype you want is in your room at all. The figures below are probabilities, not measured results. They assume independent loci and an allele that simply dominates at each locus, which is the best case.
| Independent loci to fix | Chance per seedling | Seedlings for about a 95% chance of at least one |
|---|---|---|
| 1 | 1 in 4 | 11 |
| 2 | 1 in 16 | 47 |
| 3 | 1 in 64 | 191 |
| 4 | 1 in 256 | 766 |
Generation count decides whether a trait stays fixed in the line. Garcia-de Heer and colleagues (2026, Horticulture Research) ran a single-seed descent study, selfing one seed per plant each generation. They needed 18 months or more indoors to reach the sixth selfed generation. Outdoors, with one generation per season, the same project would take about six years.
As planning rules of thumb, not measured constants:
- Finding one keeper cutting takes 10 to 30 plants and one generation.
- Fixing one simple trait takes 50 to 200 plants over four to six generations.
- Fixing a whole plant type takes 200 or more plants over six or more generations.
Translator’s note: Day length in Thailand changes much less through the year than in temperate countries. Outdoors, photoperiod plants can flower early at Thai latitudes, which can upset a breeding plan that follows the season. Indoors, you control the light, so you can run generations back to back. Plants carrying seed must stay on the stem until the seed ripens, so in the rainy season watch them closely for mould.
What you can do: Before you start a project, write down four things. These are the number of traits you want to fix, the seedlings per generation from the table, the number of generations, and whether your space holds that many plants. If it does not, cut the goal to one trait or share the work with other members.
15. Reading a Seed Listing Line by Line
| Listing text | What it means | What to expect in the room |
|---|---|---|
| “First cross, untested” | Two parents crossed once | Lots of variation, no data |
| “F2, pheno hunt recommended” | An honest, segregating population | A wide spread; grow many plants |
| “F5, stable” | Several rounds of selection | Mostly uniform; confirm with a test run |
| “Inbred line, landrace-derived” | Breeds true | Predictable, possibly less vigour |
| “S1 of a named cutting” | Selfed | All female, variable; watch for reversal |
| “BX3 of a named cutting” | Three backcrosses | Close to the original on average |
| “Autoflower F1” | Probably a first cross | Check that every plant really autoflowers |
A landrace is a traditional local variety kept by farmers in one region for a long time. On landraces in this region, see Southeast Asian Landrace Cannabis.
Anyone can test the last row. Autoflowering is caused by a recessive allele at one locus. So a properly made autoflower batch should give almost entirely autoflowering plants. If a noticeable share of plants waits for the light schedule to change before flowering, the batch was still segregating and was mislabelled.
Finally, germinate a small test batch before you fill a whole room. Six plants tell you more about uniformity than any description. The test costs one growing cycle instead of a year.
Translator’s note: Run the autoflower test indoors under long days, the light schedule you would use for vegetative growth. Outdoors in Thailand, photoperiod plants may flower early under natural day length. They can then look like autoflowers and hide a mislabelled batch. Run the six-plant test in the same room and season as the planned crop, because a cultivar can behave differently at 30–35 °C. See Choosing Genetics for Rooms Above 30 °C and 85% Humidity.
What you can do: Copy the table above and fill in one row for every pack you buy. After the six-plant test, add what you actually saw. Over time this record shows which sellers’ labels you can trust.
Share Your Numbers
Published studies do not cover the conditions many members grow in. Those conditions include tropical outdoor seasons, monsoon humidity and low latitudes where photoperiod plants behave differently from temperate data. Your records fill that gap.
If you have grown out an F2 and counted the phenotypes, post the numbers. If a selfed line produced reversals at a rate you can count, share the rate with the treatment method and the room conditions. Post your seedling counts, segregation ratios, failures and photos below this topic or in Breeder’s Lab.
Translator’s note: In Thailand, cannabis flower is a controlled herb (ช่อดอกกัญชา) under the Notification of the Ministry of Public Health Re: Controlled Herbs (Cannabis), B.E. 2568, in force since 26 June 2025. Advertising and online sale of flower are prohibited. Share data and photos for learning, and do not offer flower for sale in your posts.
References
- Punja and Holmes (2020). Frontiers in Plant Science. Seed from hermaphroditic flower clusters, sex ratio, and genetic variation after one generation of selfing.
- Watts and colleagues (2021). Nature Plants. Genotyping at over 116,000 markers compared with Sativa/Indica labels and terpene chemistry.
- Garcia-de Heer and colleagues (2026). Horticulture Research. Single-seed descent and F1 hybrids, proof of concept.
- Toth and colleagues (2022). Frontiers in Plant Science. The recessive autoflowering locus Autoflower1.
This article is part of the Asiannabis Community educational series on cannabis genetics and breeding. The content is for education only, in places where cannabis cultivation is legally permitted.