A germination test measures a sample, and a sample of 10 seeds cannot distinguish a lot that is 60% viable from one that is 90% viable, because the 95% range around 8 of 10 germinating is about 49% to 94%. Test 50 to 100 counted seeds when the decision matters, and correct the sowing rate by dividing the target stand by the germination fraction, not by adding a flat percentage.
Part of Community Seed Bank Management
A germination test counts how many seeds in a sample produce normal seedlings under favourable conditions, and uses that count to estimate the whole lot.
The Basic Test
The quick test: lay counted seeds on a damp towel, roll or fold it, keep it warm and open, and count sprouted seeds at the expected dates (Germination Test, Seed Viability). The Phase 2 guide gives test temperatures and count days by crop, and counts a seed once the root tip is at least 2 mm long.
Seed-testing laboratories that follow the International Rules for Seed Testing of the International Seed Testing Association (ISTA) use four replicates of 100 seeds, 400 in all, and sort the seeds into normal seedlings, abnormal seedlings, hard seeds, fresh ungerminated seeds and dead seeds. A seed that stays hard or fresh may be dormant and not dead, so a low result on a seed type that is known to be dormant should be retested after the recommended treatment before the lot is written off.
How Sure Is The Result?
A sample of n seeds from a lot of true germination p gives a result with standard error √(p(1 − p) ÷ n), and an approximate 95% range of ±1.96 standard errors. This is the usual normal approximation, which becomes poor when n is small or p is near 0 or 100%. The table gives the approximate range and, for the cases that matter, the Wilson interval, which stays valid at small n.
| Seeds germinating / sample | Result | Approximate ±95% (normal) | Wilson 95% range |
|---|---|---|---|
| 8 / 10 | 80% | ±24.8 points | 49.0% to 94.3%, or about 49% to 94% |
| 16 / 20 | 80% | ±17.5 points | 58.4% to 91.9% |
| 45 / 50 | 90% | ±8.3 points | 78.6% to 95.7% |
| 90 / 100 | 90% | ±5.9 points | 82.6% to 94.5% |
| 360 / 400 | 90% | ±2.9 points | 86.7% to 92.6% |
Worked example for 45 of 50: p = 0.90; standard error = √(0.90 × 0.10 ÷ 50) = √0.0018 = 0.0424, or 4.2 points; ±1.96 × 4.2 = ±8.3 points.
Two consequences. First, with 10 seeds you cannot separate “good” from “marginal”; the same data fit a lot at 50% and a lot at 94%. Second, the range shrinks only with the square root of the sample: to halve the width you need four times as many seeds. Going from 50 to 100 seeds narrows ±8.3 to ±5.9, not to half.
Comparing two tests. A drop from 90% to 84% in 50 seeds each year is inside the range of chance (84% at 42 of 50 has a Wilson range of 71.5% to 91.7%). A decline is shown by a trend across several years, or by a larger sample, not by one pair of small tests.
The Sowing Correction
To get a target stand from a lot with germination fraction g, sow 1 ÷ g times the normal rate.
| Germination | Sow this multiple | Extra seed |
|---|---|---|
| 90% | 1.11 | 11% |
| 80% | 1.25 | 25% |
| 70% | 1.43 | 43% |
| 60% | 1.67 | 67% |
| 50% | 2.00 | 100% |
The seed bank carrier says to sow 20-30% more for lots at 70-89%. The table shows why that range is not safe at the bottom of it: at 70%, sowing 30% more gives 0.70 × 1.30 = 91% of the target stand; sowing 20% more gives 84%. At 89% the correct extra is 12%, so 20-30% over-sows. The Phase 2 viability guide’s own rule (sow 25-40% denser for 70-79%) agrees with the arithmetic.
Field emergence is lower than laboratory germination because of cold soil, crusting and disease. If it matters, measure it: sow a counted batch (say 50 seeds) in the bed, count emerged seedlings, and use that fraction in place of g.
Method For A Seed Bank
- Sample from the whole lot. Mix the lot and draw seed from several places; do not scoop from the top.
- Choose the sample size by the decision. Routine screening of a large lot: 50. A critical or declining variety: 100 or more. A small lot where testing eats a large share of the seed: test what you can afford and say that the answer is rough.
- Run it under proper conditions. Use the crop’s test temperature and count dates (Seed Viability).
- Record the numbers, not only the percentage: the count germinated, the sample size, the dates of the first and the last counts.
- Watch the speed as well as the total. Ageing seed germinates later and less evenly before the final percentage drops. A lot that reaches the same total but later is declining.
- Set action levels. The Phase 2 guide advises growing a fresh lot when an important variety tests below 70%. Even 70 of 100 spans 60.4% to 78.1% (Wilson), so a lot that tests just above the line may truly be below it; act when the range, not only the single figure, reaches the line.
Limits
Everything above assumes the sample is random, the test conditions are right, and the seedlings are counted by one consistent rule. A biased sample (for example, only the largest seeds) overstates germination. Normal-approximation figures should not be used below about 20 seeds or near 0% or 100%.
What’s Next
- Learn next: Grow-Out Trials and Variety Evaluation — renewing lots that fall below the action level
- Combine with: Seed Inventory and Catalog System — recording every test