Fig WGS lineage project · experiments at home

Experiments at home

Two ways to turn “these taste different” or “this one grows differently” into evidence: a blind triangle tasting scored with an exact p-value, and a simple, consistent record of how each plant grows and fruits.

A field guide for growers · no lab needed · the next steps page explains why these are the right questions

The short version

A triangle test asks one question: can tasters tell two figs apart at all? Each taster gets three coded samples — two of one fig, one of the other — and must pick the odd one out. Guessing is right one time in three, and the result is scored against exactly that.
Match everything except the variety: same garden, same harvest morning, same ripeness, same temperature, same cut, no visual cues, random codes and serving order, and a server who does not know which cup is which.
Panel size decides what you can see. With 12 tasters, a difference that half your tasters can genuinely perceive is detected only 63% of the time; with 30 tasters, 98%. A result that is not significant is not proof the figs taste the same.
For growth and fruiting claims, grow several plants of each variety side by side and measure them with the same definitions all season: bud break, first fig, shoot growth, node spacing, fruit mass, Brix and days to ripen.
One plant is not a variety. Compare plant averages, not piles of fruit from one tree.

1 · The blind triangle test

Is there a difference anyone can taste?

A triangle test answers exactly one question: can tasters tell the two figs apart? It does not say which is better, why they differ, or whether the difference comes from the variety rather than the garden, the tree or the day. It is the standard sensory discrimination test used in food science (ISO 4120, ASTM E1885), and it works at a kitchen table.

Each taster receives three samples in coded cups. Two come from one fig, one from the other. The taster must pick the one that is different — even when unsure. If there were no perceptible difference, a taster would pick correctly one time in three by chance. So the question becomes: did more tasters pick correctly than chance explains? The probability that guessing alone would do at least that well is the p-value.

Plan before you taste

Write the question down first — for example, “Can tasters tell Smith from Cessac at full ripeness?” — together with the number of tasters, the significance level (5% is conventional) and your definition of ripe. Deciding afterwards is how a coincidence turns into a finding.
Source the fruit fairly. Ideally both varieties grow in the same garden under the same care. Fruit from two different gardens compares two gardens as much as two varieties; if that is what you have, say so when you report the result.
Find the tasters: 12 at the very least, 24–30 to catch a modest difference (see “How many tasters?”). Tasters should not know which varieties are in the test, and the person who prepares the cups should not be the person who serves them.

Prepare the figs

Harvest both varieties the same morning at the same stage — fully ripe: soft to the touch, the neck softened and bending, the fruit drooping. Discard split, soured or insect-damaged fruit.
Check ripeness with a refractometer if you have one (see Brix, below): if a spare fruit of each variety differs by more than about 1 °Bx, the test will largely measure ripeness, not variety.
Taste the same day, at the same temperature. Room temperature, about 20–22 °C (68–72 °F), for both. If the fruit was chilled, give both varieties at least an hour out of the fridge together.
Remove every visual cue. If the skins or the flesh differ in color, peel the figs and serve the pulp in opaque cups, taste under dim or red light, or blindfold the tasters and have a helper hand them each spoon.
Make the samples uniform. Fig-to-fig variation is real. For a flavor comparison, mash the pulp of at least three fruits of each variety into one pool per variety and serve a level teaspoon per cup. If texture is part of the question, serve equal wedges instead — each cut from a different fruit, so that what tasters detect is the variety and not simply “a different fig”.
Code and balance the cups. Label every cup with a random three-digit number (never 1-2-3 or A-B), and use the six possible arrangements (AAB, ABA, BAA, BBA, BAB, ABB) equally often, so that each fig is the odd one out for half the tasters. The planner below does both.

Session planner

Tasters Fig A Fig B

Run the tasting

Tasters work alone and in silence, left to right, and may go back and retaste.
Every taster must choose one cup, even if guessing. “Can’t tell” is not an allowed answer — the one-in-three arithmetic depends on a forced choice.
Rinse with room-temperature water between cups; keep plain crackers for longer breaks. Figs are sweet, and sweetness tires the palate: one triangle per person per session.
Each taster does the test once. A second triangle from the same person is not independent evidence. With a small group, run another session on another day with fresh fruit and report it separately rather than adding the two together.
Record the code each taster chooses. Only when everyone has answered does the preparer reveal the key.

Score it

Count the correct answers, k, out of n tasters. The p-value is the probability that guessing alone would give at least k correct — an exact binomial calculation with a one-in-three chance per taster:

p = Σ_i=kⁿ C(n, i) (1/3)ⁱ (2/3)ⁿ⁻ⁱ

p-value calculator

Tasters (n) Correct (k)

Or read the threshold from the table: the fewest correct answers that reach each significance level for a panel of that size.

Tasters	p < 0.05	p < 0.01	p < 0.001
12	8	9	10
18	10	12	13
24	13	15	16
30	15	17	19
36	18	20	22

Full table, 6–60 tasters

Tasters	p < 0.05	p < 0.01	p < 0.001
6	5	6	—
7	5	6	7
8	6	7	8
9	6	7	8
10	7	8	9
11	7	8	10
12	8	9	10
13	8	9	11
14	9	10	11
15	9	10	12
16	9	11	12
17	10	11	13
18	10	12	13
19	11	12	14
20	11	13	14
21	12	13	15
22	12	14	15
23	12	14	16
24	13	15	16
25	13	15	17
26	14	15	17
27	14	16	18
28	15	16	18
29	15	17	19
30	15	17	19
31	16	18	20
32	16	18	20
33	17	18	21
34	17	19	21
35	17	19	22
36	18	20	22
37	18	20	22
38	19	21	23
39	19	21	23
40	19	21	24
42	20	22	25
45	21	24	26
48	22	25	27
50	23	26	28
54	25	27	30
60	27	30	33

A dash means no score reaches that level with so few tasters. Values are exact binomial tails with a chance rate of one in three, computed when this page is built.

How many tasters?

Not everyone perceives a small difference. Suppose some share of your tasters genuinely can tell the figs apart and the rest guess: the table gives the chance that a panel of each size comes out significant at p < 0.05.

Tasters	20% can tell	30% can tell	50% can tell
12	14%	26%	63%
18	30%	52%	89%
24	30%	55%	93%
30	43%	71%	98%
36	41%	72%	99%
48	60%	88%	>99%
60	65%	92%	>99%

So a dozen friends will usually catch an obvious difference and usually miss a subtle one: if only 30% of tasters can perceive it, 12 tasters detect it 26% of the time and 30 tasters 71%.

If nothing is significant

A non-significant result means the test did not detect a difference — not that there is none. Claiming that two figs taste the same is a different, harder test, and it has to be planned as one: choose the largest share of discriminating tasters you would still call “no meaningful difference” (30% is a common choice), and use enough tasters that a low score becomes strong evidence. With a 5% risk of wrongly concluding “the same”, the most correct answers that still support the claim are:

Tasters	Most correct that still supports “the same”	Chance average
12	3	4
24	8	8
30	11	10
36	13	12
48	19	16
60	25	20
72	30	24
96	42	32

With 12 tasters the claim needs 3 or fewer correct — fewer than guessing would give — so a small panel can almost never show that two figs taste alike. Showing “different” takes a dozen people; showing “the same” takes several dozen.

Common ways a tasting goes wrong

The person who prepared the cups serves them, and gives the answer away without meaning to.
One variety was picked a day earlier, is riper, or is warmer.
Tasters confer, watch each other, or can see the fruit.
“No difference” answers are accepted instead of a forced choice.
Several tastings are run and only the one that “worked” is reported. Decide the comparison in advance and report every session.
One tree’s fruit is treated as the whole variety.

2 · Plant data

Is there a difference in how they grow and fruit?

Growth and fruiting claims — “more vigorous”, “ripens a week earlier”, “bigger fruit”, “sweeter” — can be tested with a tape measure, a kitchen scale, a refractometer and a calendar. The rules that make the numbers mean something matter more than the tools.

Ground rules

Grow them side by side. Same site, same pot size and mix (or the same bed), same watering, feeding and pruning. Alternate the positions (A, B, A, B) so that neither variety gets the sunny end.
More than one plant of each. One plant is one plant, with its own history, root system and virus status. Three or more plants per variety — ideally from independent sources of verifiable provenance — is what lets you talk about the variety.
Write the definitions down before the season starts, and keep them. A “first fig” or “ripe” date is only comparable if it means the same thing every time.
Keep your own chain of custody. Record where every plant came from (seller, date, a photo of the label) and tag the plants so they cannot be swapped.
Blind what you can. Where possible, have someone who does not know which plant is which weigh the fruit and read the refractometer.

What to measure

Bud break (“wake-up”)Spring, every 2–3 daysdate

The first date a leaf has fully unfolded anywhere on the plant.

First fig (fruit set)Every 2–3 days after bud breakdate

The first date a new main-crop fig is visible in a leaf axil at pea size (about 5 mm). Record breba figs separately: they form on last year’s wood.

Days from wake-up to fruit setComputeddays

First-fig date minus bud-break date.

Linear growth rateWeeklycm/week

Tag three comparable shoots per plant (for example the three strongest upright shoots). Measure each with a flexible tape from where this season’s growth starts — the end of last year’s wood — to the tip. Rate = change in length ÷ days between measurements × 7.

Node spacing (internode length)Once shoot growth stopscm

On the same shoots, measure from the 5th to the 15th node of this season’s growth and divide by 10. That skips the crowded base and the unfinished tip; on a shorter shoot, use the longest run you can and note which nodes.

Fruit massEvery harvestg

Weigh each ripe fig on the day it is picked, stem trimmed flush, on a scale that reads to 0.1 g. Note split or damaged fruit.

Fruit BrixEvery harvest°Bx

Zero a temperature-compensated refractometer with distilled water at the start of each session. Squeeze a few drops of juice from the pulp (not the skin), for example through a garlic press, onto the prism; wipe it clean between fruits. Take two readings per fig and average them. Pick in the morning, and note recent rain or watering, which dilutes sugar.

Days to ripenPer figdays

Tag each fig with a dated tie when it reaches pea size. Days to ripen = harvest date minus tag date, using your written definition of ripe.

Also record the context: location and climate zone; in the ground or in a container (and the pot volume); own-root or grafted; plant age; weekly high and low temperatures, or a nearby weather station; rain and irrigation; feeding; pruning dates; pests, disease, fruit drop and splitting.

Every “days to” figure is mostly a thermometer. Compare varieties within one garden and one season. To compare across gardens or years, convert days to growing degree days: for each day, add the day’s average temperature above a base — max(0, (high + low) / 2 − 10 °C), or − 50 °F in Fahrenheit — and sum from the start date. A base of 10 °C (50 °F) is a common choice.

Keep the data in one sheet

One row per measurement, so nothing needs re-typing later. The template has these columns and a few example rows:

date, plant_id, variety, source, measurement, item_id, value, unit, notes

Download the template (CSV)

Compare plants, not fruit

Summarize each plant first — its mean Brix, mean fruit mass, median days to ripen, bud-break date — and then compare the plant summaries between varieties. Twenty figs from one tree are not twenty replicates of a variety; they are twenty measurements of one tree. With three or more plants per variety, a two-sample comparison of the plant means (a Welch t-test, or a Mann–Whitney test for small or lopsided data) is the right tool. With fewer, report the numbers and do not claim a difference.

Whichever experiment you run, keep the raw sheets. They are what lets anyone else check the result.

Tasters	p < 0.05	p < 0.01	p < 0.001
6	5	6	—
7	5	6	7
8	6	7	8
9	6	7	8
10	7	8	9
11	7	8	10
12	8	9	10
13	8	9	11
14	9	10	11
15	9	10	12
16	9	11	12
17	10	11	13
18	10	12	13
19	11	12	14
20	11	13	14
21	12	13	15
22	12	14	15
23	12	14	16
24	13	15	16
25	13	15	17
26	14	15	17
27	14	16	18
28	15	16	18
29	15	17	19
30	15	17	19
31	16	18	20
32	16	18	20
33	17	18	21
34	17	19	21
35	17	19	22
36	18	20	22
37	18	20	22
38	19	21	23
39	19	21	23
40	19	21	24
42	20	22	25
45	21	24	26
48	22	25	27
50	23	26	28
54	25	27	30
60	27	30	33

Tasters	p < 0.05	p < 0.01	p < 0.001
6	5	6	—
7	5	6	7
8	6	7	8
9	6	7	8
10	7	8	9
11	7	8	10
12	8	9	10
13	8	9	11
14	9	10	11
15	9	10	12
16	9	11	12
17	10	11	13
18	10	12	13
19	11	12	14
20	11	13	14
21	12	13	15
22	12	14	15
23	12	14	16
24	13	15	16
25	13	15	17
26	14	15	17
27	14	16	18
28	15	16	18
29	15	17	19
30	15	17	19
31	16	18	20
32	16	18	20
33	17	18	21
34	17	19	21
35	17	19	22
36	18	20	22
37	18	20	22
38	19	21	23
39	19	21	23
40	19	21	24
42	20	22	25
45	21	24	26
48	22	25	27
50	23	26	28
54	25	27	30
60	27	30	33