日本語版

Research note 046 — 2026-09-04

How Thin a Bitterness Can People Notice — A Map of Thresholds

Research does not answer, with a single number, the 「これくらい」 (in translation: "how dilute") in「これくらい薄ければ苦味は分からない」 (in translation: "if it's this dilute, you can't tell it's bitter"). First, the numbers differ between the stage of noticing that something is there (detection) and the stage of recognizing it's bitter (recognition). Next, different compounds differ by orders of magnitude, and even within the same person, how widely individual variation opens up differs by compound. Further, the premises change depending on whether the measuring side is a small trained panel or general subjects, and the numbers move depending on whether the substance is dissolved in water or in beer or juice. This note lines up the reported numbers along each of these four axes.

What this page establishes

• Bitterness's 「検知閾値」 (in translation: "detection threshold") and 「認知閾値」 (in translation: "recognition threshold") are measured as different numbers even for the same substance.
• For quinine hydrochloride, the relative size of the 2019 recognition threshold and the 2026 detection threshold is reversed from what it should be, and the reason for the roughly 2-order-of-magnitude gap is not yet known.
• Even for the single compound PTC, the spread of individual variation between people (up to 8000-fold) has been measured as larger than the difference in thresholds between compounds (more than 40-fold).
• The bitterness thresholds of limonin and naringin have been measured as changing in number between when dissolved in water and when dissolved in juice.

Detection threshold and recognition threshold are not the same number

The detection threshold, where one notices 「わずかに何か違う」 (in translation: "something is faintly different"), and the recognition threshold, where one can name it 「これは苦味だ」 (in translation: "this is bitterness"), come out as different numbers even for the same substance. For picrocrocin, saffron's principal bitter component, the detection threshold in water was 5.34 mg/L and the recognition threshold was 7.26 mg/L. Under hot presentation at 61±4°C the threshold is reported to fall, and in 40% ethanol the bitterness is reported to be enhanced.

Iso-alpha acids, hops' principal bitter acid, were measured in 14 volunteers by the ASTM 1432 method (three-alternative forced choice) dissolved in unhopped lager beer, and the group-level detection threshold was 7.1 mg/L (range 4.5-11.2 mg/L). The group value is the point at which 50% of panelists correctly identified it, and individual thresholds are read off a curve at 66% correct.

Measuring tetrahydroiso-alpha acids with the same method and the same 14 people, the detection threshold came out lower, at 2.7 mg/L (range 0.7-10.0 mg/L), than iso-alpha acids' 7.1 mg/L (4.5-11.2 mg/L). Comparing the results of the ASTM 1432 method and the ASTM 679 method, no difference is reported except for 1 test day.

For quinine hydrochloride, the order of these two stages itself becomes unsteady. A 2019 study reported the recognition threshold as about 0.011-0.018 mM (-1.737 to -1.980 in log10 value). Yet a 2026 study reports the detection threshold — which should use a looser criterion — as 0.264-0.809 mg/mL (about 0.67-2.0 mM). The detection threshold should be at or below the recognition threshold, but lining up these two shows a shift of roughly 2 orders of magnitude in the higher direction. The 2026 study has n=79/78, and both the concentration series and the spray range differ from the 2019 study; neither paper pins down the reason for the divergence.

The spread between compounds, and the spread between individuals

That thresholds differ in order of magnitude between different compounds is only to be expected, but even within the same compound, the spread between people can exceed that. For TAS2R38, the receptor that divides bitterness sensitivity to PTC (phenylthiocarbamide), 5 worldwide haplotypes formed by 3 SNPs in the coding region explain 55-85% of the variance in PTC sensitivity among unrelated subjects. Mean thresholds spread by more than 40-fold between compounds, distributed from about 50 µM for PTC to about 2000 µM for sinigrin, but individual variation within a single compound reached 40-fold for salicin, goitrin, and sinigrin, and 8000-fold for PTC — the individual variation within PTC alone is larger than the spread between compounds.

PROP is often used as a stand-in for PTC, but its behavior differs even when compared within the same TAS2R38 genotype. In an experiment that stratified 32 subjects by PAV/AVI type and defined the recognition threshold as the 75% correct point on the psychometric function, the difference between PAV/PAV and AVI/AVI exceeded 400-fold for PTC, while for PROP it stayed at 40-fold.

The size of individual variation itself also differs between compounds. In a study that measured the stimulus thresholds of 4 taste substances by Pfaffmann's single-point identification method in 28 brewing-course trainees (19-40 years old) at the National Tax Agency's Brewing Research Institute, mean thresholds were glucose 4.69×10⁻²M, malic acid 1.9×10⁻⁴M, salt 1.26×10⁻²M, and quinine sulfate 7.2×10⁻⁶M (equivalent to 0.0006%). Of the 4 substances, individual variation (coefficient of variation, σ/x̄×100) was largest for quinine sulfate, reaching 136%.

Even just within a group of closely related compounds, the spread opens up widely. Tricyclic and tetracyclic conversion products of iso-alpha acids that arise during beer storage (such as tricyclocohumol) had reported threshold concentrations scattered across a range of 5 to 70 µmol/L.

Even for the conversion products formed by heating hop beta acids under wort-boiling conditions, recognition thresholds spread from 7.9 to 90.3 µmol/L. The lowest, cohulupone (7.9 µmol/L), is described as a short-lasting, iso-alpha-acid-like bitterness, while hydroxytricyclocolupone (14.7 µmol/L) is described as a coarse bitterness that lingers long at the back of the tongue and in the throat.

In a study that fractionated and identified Class III (ammonia caramel) coloring by bitterness, the threshold concentrations of the 4 compounds DDMP, HMF, furfural, and 5-methylfurfural were reported as 14.3, 30.0, 4.4, and 17.5 µmol/L respectively.

Even for the 12 peptides purified and synthesized from aged Gouda cheese, recognition thresholds were distributed from 0.05 to 6.0 mmol/L, and the lowest, the decapeptide YPFPGPIHNS, had a caffeine-like bitter quality at 0.05 mmol/L.

And even at the compound level, the number itself is not necessarily fixed at one. For mozambioside, a bitter compound specific to coffee beans, a 2015 abstract reported the recognition threshold in aqueous solution as 60±10 µM, while a 2025 abstract from a related research lineage reports 132 µM. Even for the same compound and a closely related lab lineage, a gap of roughly 2.2-fold remains.

Trained panels and the general public

Many of the numbers so far come not from the broad general population, but from small trained panels or industry reference-sample-building. In research on glutelin-derived peptides in sake, the 5 purified peptides (peaks 9, 13, 17, 18, 20) had detection thresholds of 0.02-0.18 mM and recognition thresholds of 0.08-0.66 mM (geometric mean of 6 evaluators). An index dividing the actual concentration in sake by the recognition threshold fell in the range of 6 to 60.

In a sake flavor-vocabulary system compiled with the cooperation of bodies including the Japan Sake and Shochu Makers Association, tyrosol is assigned as the standard reference sample for 『苦味』 (in translation: "bitterness"), with the discrimination threshold (detection) set at 1.4 g/L and the standard addition amount at 4.9 g/L. The tyrosol content in sake is 70-90 mg/L, which stays at a level far below this detection threshold.

In the beer flavor-vocabulary system too, item number 1200 for bitterness selects the isohumulone preparation Isolone, and is described as 『他の風味成分が少ない典型的な苦味』 (in translation: "a typical bitterness with few other flavor components") and 『閾値は正規分布を示す』 (in translation: "the threshold shows a normal distribution").

Two actual group-threshold records remain: 7.3 mg/L with 16 tasters in the Difference Rating Test, and 5.7 mg/L with 18 tasters in the A.S.T.M. ascending method using unhopped beer. Even for the same reference substance, the number moves when the method and the panel size change.

The threshold of the same substance moves with the medium

Values dissolved in water, and values in actual drinks or food, are not the same either. Limonin, a citrus bitter compound, has a threshold of 1 mg/L in distilled water. But in pH 3.8 orange juice the maximum discrimination threshold rises to 6.5 mg/kg, and the maximum value of 11 mg/kg in a model solution also occurred at the same pH 3.8.

The naringin threshold in distilled water from the same study was 20 mg/L. A separate debittering study describes an immediate bitterness arising in juice once the concentration exceeds 20 ppm.

Even for bitter gourd's bitter compounds, the medium is a solvent in which the purified compound is dissolved. Charantoside B's recognition threshold is reported as 13.6 ppm in 3% ethanol.

Momordicine IV, too, as a purified compound in the same 3% ethanol, was identified as a major contributor showing strong bitterness. Both are values for a purified compound dissolved in a solvent, not the intensity of the raw fruit itself.

The difference between a pure standard solution and actual food is also confirmed in uruka, 「うるか」 (in translation: "uruka"), a fermented ayu-entrails food. In a standard cholic acid solution, all 6 testers perceived bitterness at 8×10⁻⁴M, and sweetness also began to be perceived at 32×10⁻⁴M. In a standard taurocholic acid solution, bitterness was at 4×10⁻⁴M and sweetness at 16×10⁻⁴M.

But in the total bile-acid fraction derived from 「うるか」 (in translation: "uruka"), mixing the free form and the taurine-conjugated form at a molar ratio of 17.5:5.5, the same 6 testers needed 8×10⁻⁴M in cholic-acid-equivalent concentration before all of them perceived bitterness — a higher value than the 4×10⁻⁴M discrimination threshold of a standard mixed solution at the same composition ratio. The sweetness that was perceived in the standard solution was also not perceived, within the tested range, in the uruka-derived fraction.

An honest note

• Many of the numbers remain at the level of an abstract-level check, and include some for which the panel size and measurement procedure have not been cross-checked against the full text (see each card's review-status label).
• The definition of the detection threshold and recognition threshold (what percentage correct counts as the threshold) differs from study to study, and this difference in definition can itself be a cause of numerical divergence. The gap between the 2019 and 2026 studies on quinine hydrochloride remains with its reason unidentified in either paper.
• This note does not construct a ranking of bitterness between compounds. Even where thresholds differ by orders of magnitude, that does not necessarily mean a ranking of contribution or intensity within a food or drink.

Frequently asked

Why does the same bitterness feel different from person to person?

For some compounds, the difference in how people perceive them is reported to be very large. For the bitter substance PTC (phenylthiocarbamide), individual variation for the same compound has been reported to reach up to 8000-fold, which is larger than the difference in thresholds between compounds (on the order of more than 40-fold). Part of this individual variation is reported to be explainable by the genetic type of the receptor TAS2R38: 5 worldwide haplotypes formed by 3 SNPs are reported to explain 55-85% of the variation in sensitivity among unrelated subjects. However, the strength of this association differs by compound, and for PROP, another bitter substance, the split by genotype was not as clear-cut as for PTC.

Is bitterness sensitivity the same in water as in food?

It is not the same. Limonin, a citrus bitter compound, reached its threshold at 1 mg/L in distilled water, whereas in pH 3.8 orange juice the maximum discrimination threshold rose to 6.5 mg/kg. Naringin's threshold in distilled water is also 20 mg/L, but there is also a description in which an immediate bitterness arises in juice once the concentration exceeds 20 ppm. For uruka, 「うるか」 (in translation: "uruka"), a fermented ayu-entrails food, it is likewise reported that bitterness was discriminated only at a higher concentration in the fraction derived from the actual food than in the standard bile-acid solution. When the solvent or the food's composition changes, the concentration at which the same compound's bitterness is noticed changes too.

What is not known

• The reason the detection threshold and recognition threshold for quinine hydrochloride appear reversed — whether it is the concentration series, the spray volume, or differences in the study population that matters — has not yet been identified.
• The reason mozambioside's threshold differs by about 2.2-fold between the 2015 and 2025 reports — whether it is the panel, the purity, or the method — cannot be determined from either abstract.
• The relationship between differences in threshold order of magnitude across compounds and how much each compound actually contributes to bitterness within a food has not been systematically confirmed within the scope of this material.
• Whether the reason sweetness was not perceived in uruka's total bile-acid fraction lies in the ratio of the free form to the taurine-conjugated form, or in other components, is not yet known.
• For many compounds, the panel size and procedure for the threshold are recorded only in the abstract, and have not been confirmed against the full text.

Related:TAS2R38 ／quinine ／iso-α-acids (isohumulones) ／sake ／uruka (salted, fermented ayu entrails)
Related research notes:「苦い」 (in translation: "bitter") Is Not a Single Taste ／Does any seasoning remove bitterness? — Sugar and salt ／Where Hop Bitterness Comes From, and Where It Is Lost ／Twenty-Five Bitter Taste Receptors: What Do They Divide Between Them?
Source ledger: Source list ／ Topic hub: Beer bitterness and IBU