English 
Stability testing turns a hopeful claim into something you can show. Anyone can call a line stable. However, only a proper test tells you whether the next packet of seeds will behave like the last one.
What stability actually means
A stable line produces plants that resemble each other. It does not mean identical plants. Instead, it means the differences between siblings are small enough that a grower can plan around them. Therefore stability is a matter of degree.
Start with a sample
Ten seeds tell you very little. Twenty give a rough picture. Thirty is where patterns become reliable. Consequently, most breeders run at least twenty plants when they want an honest answer about a line.
Grow them together
Every plant in the test needs the same pot size, the same medium and the same feed. Rotate positions weekly. Otherwise you measure the light map of your room rather than the genetics of your seeds.
Flowering time first
Flowering time is the easiest trait to measure and the most useful to growers. Record the day each plant finishes. Then calculate the range. A spread of five days is tight. A spread of twenty days is not.
Height and structure
Measure final height in centimetres and count internodes. Note whether plants branch widely or grow as single colas. A well selected line usually shows one dominant structure with only minor variation around it.
Leaf and node patterns
Leaflet count, blade width and node spacing all vary between siblings. Photograph the fifth node of each plant at the same age. As a result, you build a visual record that is far more honest than memory.
Using the standard deviation
Average values hide the story. The spread tells it. Calculate the standard deviation for flowering day and height. Then compare it between generations. A shrinking number is the clearest proof that selection is working.
Counting off types
An off type is a plant that clearly does not belong. Count them and express the number as a percentage. Below five percent is good for a working line. Above twenty percent means the line is not ready.
Testing for hermaphroditism
Stress reveals what calm conditions hide. Check every plant for stray pollen sacs at week six and week eight. Any plant that throws them gets recorded, since this trait carries strongly into the next generation.
Germination as a health check
Count how many seeds sprout within four days. Ninety percent or better suggests healthy genetics and proper storage. A drop towards seventy hints at inbreeding depression or poor seed handling.
Repeat in a second environment
One room proves very little. Send seeds to a grower using different lights and a different medium. A widely grown variety makes a useful control plant in both trials.
Aroma and flavour come last
Scent shifts with feeding, temperature and cure. Judge it twice: once on the living plant and once after drying. Even then, treat aroma notes as softer evidence than dates and measurements.
Recording the data
Use a spreadsheet with one row per plant. Include sow date, flip date, finish date, height, node count and any off type note. Consequently, you can sort and compare instead of relying on impressions.
Reading the results honestly
Numbers only help if you accept what they say. A line with a twenty day flowering spread is variable, however good the best plant looked. A consistent reference line gives you something to compare against.
When to test again
Test after every second generation, not every one. That keeps the workload realistic. Meanwhile, keep seed from each tested generation so you can return to an earlier point if the numbers get worse.
Deciding what counts as a pass
Set your thresholds before the trial starts, not after you see the plants. Decide, for example, that a flowering spread under seven days passes and anything wider fails. Write those limits in the notebook first. Otherwise every result becomes negotiable, and a line you like quietly gets graded more kindly than one you do not. Fixed thresholds keep the exercise honest.
Blind scoring where possible
Label pots with numbers rather than parent names. Then score the plants without knowing which cross they came from. As a result, expectations stop steering your judgement. Reveal the labels only once the measurements are written down.
Photographing the trial
Take one photo per plant each week from the same distance and angle. Use a plain background and include the pot number in frame. Consequently, you can revisit the trial months later and still compare plants fairly.
Common mistakes
The first mistake is a sample too small to mean anything. The second is comparing plants grown weeks apart. The third is scoring aroma before the cure. Above all, avoid quietly dropping a weak plant from the record because it spoils the average.
Sharing results with buyers
A short summary builds far more trust than adjectives. Publish the sample size, the flowering range, the average height and the off type percentage. Then let growers judge for themselves. Consequently, buyers know what to expect and complaints drop, because nobody was promised uniformity that the seed cannot deliver.
Storing the tested seed
Keep the remaining seed from every tested generation cold, dry and dark. Label each jar with the generation code and the date of the trial. Therefore, if a later generation performs worse, you can return to the batch that tested well instead of starting again.
Testing autoflowers
Autoflowers compress everything into ten weeks. Measure days from sowing rather than from a flip, since there is no light change to anchor the timeline. Meanwhile, watch for plants that finish very early, because those often carry a yield penalty.
Who should run the trial
Testers do not need to be experts, though they do need to be consistent. Give each one the same short recording sheet and the same instructions. Ask for dates and measurements rather than opinions. As a result, three ordinary growers produce data you can actually combine, while three enthusiastic reviewers produce three incompatible stories.
Include a reference variety in every wave
Put the same known variety into every wave of the trial. It works as a ruler: if the January wave underperformed, the reference shows that conditions were to blame rather than the new genetics. Therefore without that plant the waves cannot be compared at all. Consequently the cost of one extra pot pays for itself many times over.
Measure with what you already own
A tape measure, a kitchen scale and a notebook are enough. By contrast, measuring from memory is worth nothing. Therefore record final height, internode spacing, harvest day and dry weight for every plant. In fact those four numbers answer most of the questions people ask about a line.
How many waves before you trust it
One wave shows a tendency, two confirm it and three settle the matter. Seasonal variation is large, so a single wave misleads. Therefore plan the trial across the year rather than within one harvest.
Publishing what you measured
A short factual summary builds more trust than any adjective. State the sample size, the flowering range, the average height and the off type percentage. Consequently buyers know what to expect and complaints drop, because nobody was promised uniformity the seed cannot deliver. Therefore transparency costs nothing and protects the reputation of the line.
Legal note
Seeds are sold as collectable items. Cultivation rules differ between countries and change over time. Therefore check the current law where you live before planning any project. This article is informational only.
Frequently asked questions
How many plants make a valid test?
Twenty is a practical minimum and thirty is better. Fewer than ten gives results you cannot rely on.
Can I test stability from one grow?
Partly. One grow shows the spread, yet a second environment reveals whether that spread holds elsewhere.
Is a small spread always good?
Usually, though a uniform line can still be a mediocre one. Stability and quality are separate questions.
