Short answer. A quantitative trait is one shaped by many genes of small effect plus the environment, so it varies continuously (grams, percent, days). Heritability tells you how much of the variation you see is genetic in your population and your room. When heritability is low, the best-looking plant is often just the luckiest, and its offspring regress towards the average.
What's happening
Observed variation splits into parts [1]:
V_P = V_G + V_E + V_GxE, and V_G = V_A + V_D + V_I
- V_A, additive: the part that reliably passes to offspring.
- V_D, dominance and V_I, epistasis (gene interactions): show in an individual or F1 but do not breed true.
- V_E, environment: pot position, light, watering, pests.
- V_GxE, genotype by environment: cultivar A beats B in coco but not in soil.
Two heritabilities:
- Broad-sense (H²) = V_G / V_P. Useful for clone selection, because a clone copies the whole genotype.
- Narrow-sense (h²) = V_A / V_P. Useful for seed breeding, because only additive effects reliably pass on.
The breeder's equation
R = h² x S [1]
- S (selection differential) = mean of the plants you keep minus the mean of the population.
- R (response) = how much the next generation's mean moves.
Example: a population averages 400 g/m². You keep plants averaging 500 g/m² (S = 100). If h² for yield in that setting is 0.3, expect next generation's mean around 430 g/m², not 500.
What we know for cannabis
| Trait | Heritability (indicative) | Notes |
|---|---|---|
| THC:CBD ratio (chemotype) | Very high | Largely one inheritance unit; see chemotype inheritance |
| Cannabinoid composition ratios | High; stable within a genotype | Absolute amounts varied much more than ratios [2] |
| Absolute THC % | Moderate | Strongly affected by light, harvest timing and lab [2][4][5] |
| Flowering time | Moderate to high, with major genes | Cultivar origin predicted flowering in hemp [4] |
| Yield | Low to moderate | Strongly environment driven [5] |
| Terpene profile (which terpenes) | Moderate to high | Tandem terpene synthase genes; see strain-name article |
| Morphology and phenology | Lower than cannabinoid traits in [2] |
Naim-Feil et al. (2023) concluded that cannabinoid composition should respond more strongly to selection than morphological and phenological traits [2] (Supported). Published cannabis heritability values come from small, single-site studies; treat specific numbers as tentativebeing reviewed.
Lab numbers add noise too: THC results for similar products varied systematically between testing labs in Washington State [3]. A 2-point THC difference between two plants tested at two labs is not evidence of anything.
How to raise your effective heritability
You cannot change the genes' heritability by wishing, but you can cut V_E:
- Grow siblings side by side in the same pots, medium and feed. Randomise positions; edge plants get different light.
- Replicate. Score clones of each candidate (2-4 each) in a second run before choosing.
- Use the same lab, same harvest stage, same sample position (top cola) for chemistry.
- Measure, don't guess. Grams dry per plant, days to finish, weighed and dated in the journal.
- Progeny test. The best evidence a plant is genetically superior is how its offspring perform.
Numbers that matter
| Concept | Formula | Use |
|---|---|---|
| Broad-sense heritability | H² = V_G / V_P | Clone selection |
| Narrow-sense heritability | h² = V_A / V_P | Seed lines |
| Response to selection | R = h² x S | Predict gain |
| Selection intensity | i (e.g. top 10% ≈ 1.75; top 20% ≈ 1.40) | R = i x h² x σ_P |
Selection intensities are standard normal-distribution values [1].
Common mistakes
- Picking the "keeper" from one run of one plant per seed. You are partly selecting the luckiest pot.
- Treating THC % as a fixed property of a cut. It isn't [2][3].
- Selecting on many traits at once with tiny populations. Gains per trait shrink fast.
