Short answer. Cross photo x auto, grow the F1 (all photoperiod), then either sib-cross the F1s (F2: ~25% autos) or backcross an F1 to the auto (~50% autos). To keep a photoperiod line's character while adding autoflowering, backcross towards the photo parent, using a "reveal" generation or a DNA marker each time to keep the hidden auto allele.
What's happening
In the populations mapped by Toth et al. (2022), autoflowering is controlled by a recessive allele at Autoflower1 on chromosome 1; F2s fit 1:2:1 genotypes with about 25% autos [1]. Heterozygotes flowered about two weeks earlier in the field but did not autoflower [1]. Not all auto sources have been shown to share this locus, so crossing two different auto lines may not always give autos if they carry different mechanisms (Contested) [1].
Paths compared
| Plan | Generations | Auto share | Photo-parent genome (avg) |
|---|---|---|---|
| F1 then F2 | 2 | ~25% of F2 | ~50% |
| F1 x auto parent (backcross to auto) | 2 | ~50% | ~25% |
| Backcross to photo with reveal steps (F1, BX1, BX1-F2 reveal, BX2, BX2-F2 reveal) | 5 | ~25% in each reveal generation | ~87.5% |
| Backcross to photo with a marker (F1, BX1, BX2, BX3 picking carriers, then F2) | 5 | ~25% in final F2 | ~94% |
Averages from standard backcross genetics [3]; actual recovery varies.
Step by step: marker-free route
- P: Photo line female x stable auto male (or reverse). Record IDs.
- F1: All photoperiod. Pick the best F1s.
- F2: Sib-cross F1s. Grow under 18/6 or longer; autos are the ones that flower at 3-5 weeks [1].
- Select autos with the photo parent's best traits.
- Backcross chosen autos to the photo parent (BX1, all photoperiod, all carriers).
- Reveal: sib-cross BX1s; select autos again.
- Repeat 5-6 until the line looks like the photo parent in the traits you care about.
- Stabilise: grow 50+ seeds from your auto x auto line under long days; any plant that doesn't flower at the expected age is an off-type. Aim for 0 photoperiod off-types in a 50-plant test before release.
Step by step: marker-assisted route
If a validated marker for your auto source exists (check the lab's validation in your germplasm), backcross F1 x photo parent each generation and test seedlings, keeping only carriers [1]. Reveal autos only in the final generation. This saves time but costs lab fees, and markers may not predict every auto source [1].
Growing tips for selection
House position: autoflowers. An auto runs on a strict countdown from germination, so any setback costs it time it can't get back. Plant it straight into its final pot: transplanting is almost a no-no, because of how the auto's taproot reacts (the founder's observation, compared with photoperiods). Train with LST, don't top: topping will never produce what proper low-stress training does on an auto. Grower practice, house-endorsed.
- Sow selection plants straight into their final pots and train with LST only, so every plant gets the same clean run and setbacks don't mask differences.
- Score autos on days to flower and total cycle length under the same long-day schedule (18/6 or 20/4).
- Expect more variation in size and yield than in photoperiod hunts: you can't veg longer to equalise.
- Clones of autos flower on the parent's age clock; select from seed populations.
Numbers that matter
| Test | Plants for 95% chance of ≥1 auto |
|---|---|
| F2 (p = 0.25) | 11 |
| BX to auto (p = 0.5) | 5 |
| Want ≥5 autos in an F2 | 34 |
Are autos better? Better how?
Growers ask it as if there's one answer. There isn't: better at what? The grower's side (yield per area per year, finickiness, "good for beginners") is in photoperiod vs autoflower. The short version: per square foot per year autos are very hard to beat, except perhaps by a dialled-in sea of green or SCROG with a dedicated nursery, because more harvests fit in a year [6]. Here's the breeder's side: why autos lagged on cannabinoids, and why the best have caught up.
The trait itself is one switch. Day-neutral flowering maps to a major locus, Autoflower1, and behaves as a recessive in the populations studied [1]. It's not quite all-or-nothing: in triploid and diploid hybrids, the fewer photoperiod alleles a plant carried, the earlier it flowered, which points to incomplete dominance of the photoperiod allele [5]. One gene doesn't make a plant weak. What came with it did.
Why autos lagged: what came along with the gene. The auto trait came into drug-type breeding from northern and Central Asian populations popularly called ruderalis [8], which were wild or weedy, small, and not selected for resinbeing reviewed. When you bring one trait in from an unimproved donor, you also bring in the chromosome segments around it and, in the early crosses, half the donor's genome. Plant breeders call the first problem linkage drag [7]. Each backcross to an elite photoperiod parent halves the remaining donor genome on average (50 %, 25 %, 12.5 %...) [3], but the segment linked to Autoflower1 on chromosome 1 hangs on longest because you're selecting for it every generation. Early autos were a cross or two away from the donor, and tested like it.
Why the best have caught up. Breeders kept backcrossing autos into elite photoperiod lines, selecting each generation for resin and potency as hard as for the auto trait, and kept recombining away the linked donor segment. A modern elite auto can be mostly elite photoperiod genome carrying a small piece of the donor. In the founder's experience autos still usually test slightly under their photoperiod sisters, but some give them a real run (no controlled matched-line comparison found)being reviewed.
Founder's records. The founder has had an elite autoflower lab-test at 30 % and has pulled 3/4 lb (about 340 g) in under 80 days. One grower, one elite line: the ceiling, not the average.
Why the yield ceiling is lower per plant. An auto can't be vegged longer, so plant size is set by its genetic clock and how cleanly it's grown in the first weeks. Breeders select for vigour in that window and for a cycle length that suits the market (faster lines give more cycles a year; slower ones give bigger plants). A one-week difference in cycle length changes cycles per year by roughly 8–10 % at 10–12-week cycles, which is why days to harvest is a breeding target, not just a convenience.
What that means for breeding. Treat potency and terpenes in an auto project as quantitative traits to select for every generation, not as something the elite photo parent will hand over by itself (see quantitative traits and heritability). Grow big populations, select the males too, and test-grow before release; the same standards as any line (see breeders vs pollen chuckers).
Common mistakes
- Expecting F1 autos.
- Releasing a "stable auto" that throws photoperiod plants (incomplete fixation or a second photoperiod source).
- Crossing two auto lines of unknown origin and getting photoperiod offspring (possible if mechanisms differ) [1].
Evidence
- Recessive single-locus inheritance in mapped populations: Supported [1][2].
- Universality across auto sources: Contested [1].
- Backcross arithmetic: Established [3].
- Incomplete dominance of the photoperiod allele (dosage affects flowering time): Supported [5].
- Linkage drag from an unimproved donor as the reason early autos lagged: Supported as general breeding principle [7]; not measured directly for Autoflower1.
- Modern autos testing slightly under photo sisters, the best matching them: founder's observation; no controlled comparison found.
