Topic hub — 2026-09-19
Cacao bitterness and astringency
Cacao bitterness is usually described on a single scale: the higher the cacao percentage, the more bitter. The measurements collected on this page show that scale breaking several times along the way. Fermentation raised bitterness at some facilities and lowered it at others. Roasting lowered perceived bitterness as the temperature went up. Raising the fat content lowered perceived bitterness without changing the compounds. And the relation between cacao percentage and bitterness itself was visible in one group of products and not clear in another. Here the sample, the compounds, the process and the formulation are kept apart, and only the confirmed claims are laid side by side.
On this page (8 sections)
What this page establishes
- No single compound carries cacao bitterness and astringency. Alongside theobromine and caffeine, epicatechin and procyanidins are identified as bitter contributors, and N-phenylpropenoyl amino acid amides as astringent ones.
- Fermentation does not move bitterness in one direction. Comparing three facilities in Colombia, bitterness and astringency rose at two and fell at one.
- Roasting lowers perceived bitterness as temperature rises: a one standard deviation (42 degC) increase in roasting temperature corresponds to a drop of about 1.56 gLMS units near a mean bitterness of 26.9 (the authors' conversion of a square-root model coefficient).
- The relation "higher cacao percentage, higher bitterness dominance rate" was seen in a group of non-origin chocolates but was not clear in a group of origin chocolates, and the two groups also spanned different cacao ranges.
- Formulation moves bitterness too. Raising total fat from 30% to 40% lowered perceived bitterness, and in cocoa beverages sucrose concentration changed it. Yet the difference between 0.1% and 4.5% fat did not affect bitterness in the beverage.
Which sample are we looking at
The first thing to separate when reading cacao bitterness is what is actually in the mouth. Roasted nibs, cocoa powder, solid chocolate, a cocoa beverage, and an infusion of the bean's shell are all called "cacao," but the object being measured differs in each case.
The shell (husk) is the skin separated from the nib, distinguished from the pod that encloses the whole fruit. Infusing it as a drink has been tested in recent years, and it is treated as a separate sample on this page.
Scope: In this entry, husk refers to the bean shell.
Show data
ingredient-cocoa-bean-shell —is_defined_in_context_as→ (the claim sentence above)Nor do the sensory descriptors recorded at one processing stage necessarily carry through to the finished product. The process-side claim states explicitly that retention of the same characteristic attributes in the final chocolate is not guaranteed. That is why the process measurements and the finished-product measurements are kept in separate sections here.
Scope: Fine-flavor cacao literature reviewed through May 2023
Note: The fermentation, roasting, liquor, and chocolate stages should be modeled separately.
Show data
ingredient-roasted-cocoa-nibs —does_not_guarantee_retention_of→ (the claim sentence above)What carries the bitterness and the astringency
The compounds named as the main carriers of bitterness in roasted nibs are two methylxanthines, theobromine and caffeine.
Scope: Water-based reconstruction of roasted cacao nibs and human omission testing
Note: Specific to the roasted-nib material studied and the water-based model.
Show data
ingredient-roasted-cocoa-nibs —has_key_bitter_organoleptics→ (the claim sentence above)They are not the only ones. For the same sample, epicatechin, catechin, procyanidin B-2, B-5, C-1 and higher procyanidins are identified as contributing to both bitterness and astringency. This is why the two are kept on separate axes rather than merged into one.
Scope: Roasted cacao nibs, sensory-guided fractionation
Show data
ingredient-roasted-cocoa-nibs —has_bitter_and_astringent_contributors→ (the claim sentence above)A different group of compounds appears on the astringency side: N-phenylpropenoyl amino acid amides including clovamide and deoxyclovamide, and the glycopyranosides of quercetin, naringenin, luteolin and apigenin. These do not overlap with the compounds named for bitterness.
Scope: Roasted cacao nibs, taste dilution analysis and half-tongue threshold testing
Note: These compounds are reported as contributing to astringency, not bitterness. Cacao's astringency cannot be reduced to flavan-3-ols alone.
Show data
ingredient-roasted-cocoa-nibs —has_key_astringent_compounds→ (the claim sentence above)In cocoa powder, the compounds with a dose-over-threshold above 1 are epicatechin, epigallocatechin gallate, luteolin, protocatechuic acid, theobromine and caffeine. The amount detected and the amount that contributes to perception are not the same thing; this index is an attempt to bridge that gap.
Scope: Cocoa powder before and after microwave-assisted alkalization
Note: Methylxanthines appear in the same DOT set alongside 4 kinds of phenolics, and are not the sole bitterness driver in this material.
Show data
ingredient-cocoa-powder —has_key_bitter_organoleptics→ (the claim sentence above)Does processing lower the bitterness
Reading fermentation, drying, roasting and alkalization as a single arrow of falling bitterness stops matching the measurements. Each process differs in both direction and size.
Compounds do not move monotonically during fermentation. Theobromine and caffeine concentrations fell to day 5 and rose again by day 6; catechin did the opposite, rising to day 5 and falling by day 6.
Scope: 3 cacao genotypes (CCN 51, ICS 95, TCS 01), 3 temperature-controlled fermentation profiles
Note: A non-monotonic trajectory. Only 3 genotypes were tested, and temperature-controlled fermentation is a different thing from natural fermentation.
Show data
process-cocoa-fermentation —can_change→ (the claim sentence above)Perception is not one-directional either. Comparing day 4 and day 7 cocoa pastes across three facilities in Colombia, bitterness and astringency rose at two (ASA and COC) and fell at one (CMI). The same operation of "fermenting longer" worked in opposite directions by place.
Scope: 3 fermentation boxes per facility, sampled on day 4 and day 7, dried, roasted, and ground paste, a trained panel of 5, a 0-10 intensity scale
Note: Because the clonal composition of the beans, fermentation quantity, turning, and drying conditions differ by facility, this is not treated as a consistent-direction effect of fermentation days alone.
Show data
process-cocoa-fermentation —has_perceived_bitterness_affected_by→ (the claim sentence above)Drying temperature affects bitterness as well. Liquor from beans dried at 80 degC was rated more bitter and more astringent than liquor from the 60 degC side, though this remains a preliminary observation.
Scope: Nacional samples from 2 sites in Ecuador, compared without pod storage. The 60°C group also had longer fermentation.
Show data
process-cocoa-bean-drying —was_observed_with→ (the claim sentence above)Roasting is the process on this page that most clearly lowers bitterness. For perceived bitterness in 100% chocolate, the square-root model gave a coefficient of about -0.153 for a one standard deviation (42 degC) rise in roasting temperature, which the authors convert to a drop of about 1.56 gLMS units near a mean bitterness of 26.9.
Scope: 27 roasting treatments across 3 origins, 145 chocolate consumers, gLMS
Note: The direction is the reverse of the intuition that 「焙煎温度が高いほど苦くなる」 (in translation: "the higher the roasting temperature, the more bitter it becomes"). An exact p-value is not given, but Table 1A marks the temperature term as significant at p≤0.05 (source check by Codex, 2026-10-21).
Show data
process-cocoa-roasting —reduces→ sensation-bitterIn shell infusions, however, roasting works the other way. For shells separated from beans roasted at 100 degC for 20 minutes, 120 degC for 20 minutes and 140 degC for 18 minutes, the mean bitterness rank rose across 17.4, 37.8 and 54.7. Change the sample and the direction of the same process changes too.
Scope: Trinitario cocoa beans from Jember, Indonesia. 2 g of shell infused in 150 mL of hot water for 5 minutes, unsweetened and without milk. Evaluation by 15 panelists after screening.
Note: The mean rank is not a score on an intensity scale. The table does not show individual significant differences between the 3 conditions; this is a Kruskal-Wallis analysis using evaluation data from 6 samples, including 3 black-tea samples. The 140°C condition alone had an 18-minute roasting time, so this is not a comparison of temperature alone.
Show data
ingredient-cocoa-bean-shell —has_sensory_response→ (the claim sentence above)Some processes are carried out with the aim of reducing bitterness. Alkalization (the Dutch process) raises pH from the acidic range of natural cocoa into the neutral-to-alkaline range. The paper cited here describes it, citing a review, as a step that reduces bitterness. Its experiment, however, compared eight alkalized cocoas made for the study and did not directly test a drop in bitterness from untreated powder.
Scope: A test applying 8 alkalized cocoa powders made for the study (with varied treatment conditions) to sponge cake
Note: External audit N2-2 (2026-10-20): re-read in the full text from the institutional repository (RiuNet, hdl 10251/189867). The statement that alkalization reduces bitterness is an introductory sentence citing a review (Valverde-Garcia et al. 2020); this study itself did not compare bitterness with non-alkalized cocoa. In its sensory evaluation (9 trained panelists, 0 to 5 scale), the more bitter samples were CM_KOH and NM_K2CO3 under mild conditions and CS_K2CO3 and CS_KOH under strong conditions, so bitterness did not decrease as the treatment got stronger. The status is preliminary until the primary basis of the review is checked.
Show data
ingredient-cocoa-powder —can_be_debittered_by→ (the claim sentence above)How it is perceived in the finished product
Solid chocolate and cocoa beverages need to be read separately. Solids first.
The relation "higher cacao percentage, more bitter" holds conditionally. In a study comparing five origin and six non-origin chocolates by temporal dominance of sensations (TDS), the dominance rate of bitterness rose clearly with cacao percentage in the non-origin group, whereas that relation was not clear in the origin group. The two groups also covered different cacao ranges (origin 66.8-80.1%, non-origin 54.5-80.0%), so the difference cannot be put down to origin alone. A dominance rate is the proportion of assessments selecting a sensation as dominant at each time point, not bitterness intensity.
Scope: TDS panel evaluation of 5 origin (66.8-80.1% cc) and 6 non-origin (54.5-80.0% cc) dark chocolates
Note: This is not a source claiming the simple, monotonic relationship that "the higher the cacao percentage, the more bitter it always tastes." The authors suggest effects of variety, region, cultivation and post-harvest handling in addition to cacao percentage; the contribution of each factor cannot be separated in the abstract we checked. A dominance rate is not bitterness intensity. The abstract uses the words origin / non-origin, and whether these mean single-origin / blend has not been confirmed in the full text (source check by Codex, 2026-10-21).
Show data
ingredient-dark-chocolate —has_bitterness_correlated_with→ (the claim sentence above)Between the two commercial products compared, the difference was plain. The median of the 10-point bitterness rating was 2 for a 36% product against 7.45 for an 85% one, a significant difference. The two products also differ in sugar, fats and maker, so the gap cannot be attributed to cacao percentage alone. The formulation, processing and evaluation conditions of each study, and how far they can be compared, are set out in Is chocolate with a higher cocoa percentage more bitter?
Scope: Wagner et al. 2023, Experiment 1; 39 untrained participants; 5 g of each of 2 commercial products presented in duplicate; 10-point category scale.
Note: Because the 36% and 85% products differ in manufacturer and ingredient composition, this is not treated as an effect of cocoa content alone. The 70% product is not included in this sensory experiment 1.
Show data
ingredient-dark-chocolate —has_perceived_bitterness_affected_by→ (the claim sentence above)Cacao percentage is not the only formulation lever. In reconstituted chocolates held at 70% cacao while total fat was varied across 30%, 35% and 40%, higher fat lowered perceived bitterness, cocoa flavour and astringency. Perception moved by changing the medium, not the amount of the compounds.
Scope: A 70% cocoa-percentage reconstituted (cocoa powder + cocoa butter) two-component chocolate, with sensory evaluation by 108 regular chocolate consumers
Note: The result comes from a two-component reconstituted chocolate in which cocoa content and fat were deliberately varied, not from ordinary commercial chocolate with milk or emulsifiers such as lecithin, so extrapolation to commercial products needs care. 2026-10-21 (full-text re-evaluation, batch 1): adding fat meant less cocoa powder, so the effect of fat alone cannot be separated (added to the evidence limitations). The observed drop in bitterness itself matches the text excerpt (bitterness coefficient −7.41±1.88, p=.0001), so the claim stays supported.
Show data
ingredient-dark-chocolate —has_perceived_bitterness_affected_by→ (the claim sentence above)On the beverage side, sugar is what moves. Sucrose concentration (0%, 2.5%, 5%) changed perceived bitterness in a cocoa drink.
Scope: A trained-panel test that factorialized 2 fat levels and sweetener
Note: Bitterness and bitter aftertaste decreased significantly as sucrose increased. This does not mean that the sugar removed the bitter compounds.
Show data
ingredient-chocolate-milk —has_perceived_bitterness_affected_by→ (the claim sentence above)Fat, which worked in the solid, did not work in the beverage: the main effect of 0.1% against 4.5% fat on bitterness was not significant. The same operation of "adding fat" gave different results in solid and liquid.
Scope: The same factorial chocolate-milk test
Note: There was no significant main effect of fat concentration on bitterness. This is not proof that fat in general has no effect on bitterness.
Show data
ingredient-chocolate-milk —has_perceived_bitterness_unchanged_by→ (the claim sentence above)In a beverage prototype, the aftertaste differed. Using the same amount of powders from different brands, buffalo milk chocolate milk made with alkalized cocoa powder scored higher for bitter aftertaste intensity with a trained panel than the natural cocoa powder version (2.9 against 2.4 on a 15 cm scale). The two powders also differed in fat, ash and pH, so this cannot be attributed to alkalization alone; the two commercial products scored the same, 1.8.
Scope: Juvinal et al. 2023; 2 prototypes containing 1% cocoa powder, 7.5% sugar, and so on; 8 trained panelists; 15 cm line scale; 4 replicates.
Note: Because the cocoa powders differ not only in treatment but also in brand, the difference is not treated as a sole causal effect of alkali treatment. This is not extended to solid milk chocolate or milk in general.
Show data
ingredient-chocolate-milk —has_sensory_attribute→ (the claim sentence above)What was measured and did not move
Measurements that produced no difference are given the same weight here as those that did.
For two 70% products made from microwave-roasted and convection-roasted beans, a CATA with 130 consumers found no significant difference in the citation frequency of 12 in-mouth flavour terms including bitter. This did not measure bitterness intensity, and a separate triangle test did tell the two apart (27 correct of 51 judgements, p = 0.003).
Scope: Schouteten et al. 2023; Ecuadorian mixed-variety beans; 600 W for 35 minutes versus 130°C for 30 minutes; 2 70% chocolate bars; 130 adult Belgian consumers.
Note: CATA is a method of selecting applicable terms, not an intensity scale. Not only the roasting method but also the equipment, batch size, and time/temperature conditions differ together as a set. This is not a proof of equivalence between the 2 conditions.
Show data
ingredient-dark-chocolate —has_bitter_cata_frequency_unchanged_by→ (the claim sentence above)In a test combining chilli and sweeteners in unsweetened spicy dark chocolate, the maximum bitterness intensity measured by time-intensity showed no significant difference in nearly all nine combinations. Adding pungency did not move the peak of bitterness.
Scope: Layer: food. Unsweetened spicy dark chocolate, TI method with a trained panel (13 members assigned to bitterness), consisting of a 5-second wait, 15 seconds holding in the mouth, and post-swallow bitterness rated on a 0-10 scale over 120 seconds.
Note: The intensity and duration of spiciness (pungency) increased with rising chili concentration, but the Imax of bitterness did not move statistically across most of the concentration combinations. An example, at the food layer, of pungency and bitterness behaving independently.
Show data
ingredient-dark-chocolate —has_perceived_bitterness_unchanged_by→ (the claim sentence above)Origin produced a difference too. Mean bitterness of 100% chocolate on the gLMS was 28.1 for Peru against 26.0 for Ghana and 26.8 for Madagascar, a statistically significant difference. The study of the commercial 36% and 85% products used a different scale (the median of a 10-point category scale), so the sizes of the two differences cannot be compared directly.
Scope: 3 origins of unsweetened 100% chocolate, 145 consumers
Note: The origin effect coexists with roasting effects in the same experimental design. In the text read, both were not separated into an interaction term.
Show data
ingredient-roasted-cocoa-nibs —has_perceived_bitterness_affected_by→ (the claim sentence above)Questions people ask
Is chocolate with a higher cacao percentage more bitter?
Of the two commercial products compared, the 85% one was rated more bitter than the 36% one (median 7.45 for the 85% product and 2 for the 36% product, significant). But they also differ in sugar and fats, so this is not the effect of cacao percentage alone. In a study comparing five origin and six non-origin chocolates by TDS, the dominance rate of bitterness rose clearly with cacao percentage in the non-origin group, whereas that relation was not clear in the origin group (the groups also covered different cacao ranges). The conditions of each study, and how far they can be compared, are set out in Is chocolate with a higher cocoa percentage more bitter?
Does roasting make cacao more bitter, or drive bitterness off?
It depends on the sample. In 100% chocolate, a one standard deviation (42 degC) higher roasting temperature corresponded to a drop of about 1.56 gLMS units near a mean bitterness of 26.9 (the authors' conversion), so bitterness fell as temperature rose. In infusions of the bean shell the opposite held: as roasting temperature rose across 100, 120 and 140 degC, the mean bitterness rank rose across 17.4, 37.8 and 54.7. The same word "roasting" covers operations whose direction depends on what reaches the mouth.
Does fermentation reduce bitterness?
It does not settle in one direction. Comparing day 4 and day 7 cocoa pastes across three facilities in Colombia, bitterness and astringency rose at two and fell at one. The compounds are not monotonic either: theobromine and caffeine fell to day 5 and rose by day 6, while catechin moved the other way. The reading that longer fermentation drives bitterness off is not supported within the scope of this page.
Do milk and sugar make the bitterness disappear?
They move it downward rather than removing it. In a cocoa beverage, sucrose concentration (0%, 2.5%, 5%) changed perceived bitterness. In solid chocolate held at 70% cacao, raising total fat from 30% to 40% lowered perceived bitterness, cocoa flavour and astringency. But in the beverage the difference between 0.1% and 4.5% fat did not affect bitterness, so what works in the solid does not necessarily work in the liquid.
What is not known
- What differs between the facilities where fermentation raised bitterness and the one where it fell is not known within the scope of this page. The observation that facilities differed is on record; a source separating the cause is not.
- Why roasting lowers bitterness in solid chocolate and raises it in shell infusions has not been measured. Beyond the fact that the samples differ, no explanation can be placed here.
- Alkalization is described as a step that reduces bitterness, yet in a buffalo milk chocolate milk prototype the version with alkalized powder had the stronger bitter aftertaste. The two powders differed in brand and composition, so whether this is an effect of the treatment or of the powders is not known.
- Why the relation between cacao percentage and bitterness was not clear for the origin chocolates cannot be explained within the scope of this page. Whether it reflects the different cacao ranges, or variety, region, cultivation and post-harvest handling, is not separated in the abstract we checked.
- The observation that drying temperature affects bitterness remains at a preliminary level and goes no further than a comparison of two levels, 80 and 60 degC.
Read in the encyclopedia
- Roasted cocoa nibs — the sample at the centre of the compound analyses, where the bitter and astringent carriers are identified
- Cocoa powder — the sample where contributors are narrowed by dose-over-threshold, and the object of alkalization
- Cocoa bean shell (husk) — the bean's skin, distinguished from the pod; particle size and roasting conditions change the bitterness rank
- Dark chocolate — where cacao percentage, fat and formulation move perceived bitterness
- Chocolate milk — tests of sugar, fat and sweetener replacement in a liquid, set against the solid
- Cocoa fermentation — the process whose direction changes by facility
- Cocoa bean drying — distinguished from roasting; temperature moves bitterness and astringency
- Cocoa roasting — the process where perceived bitterness falls as temperature rises
All entries in this hub (10)
Every encyclopedia entry that belongs to this hub, grouped by category. Entries without an English version open in Japanese.
ingredientcocoa bean shell (husk) · dark chocolate · chocolate milk · roasted cocoa nibs · cocoa powder · milk chocolate
processcocoa bean drying · cocoa fermentation · cocoa roasting · cocoa conching
Go deeper in the research notes
- When cacao was bitter — tracing cacao before it was sweetened, from the records
- The boundary between roast aroma and bitterness — where to separate what roasting creates
- Can cooking create bitterness? — placing the cacao processes in the four quadrants of adding, removing, designing and suppressing
- A world map of debittering — placing alkalization among the world's techniques for taking bitterness out
Source ledger
The primary sources and limitations for the claims on this page can be checked from the source links inside each card. The full list is on the Japanese source ledger.
How to cite this page
Aoki, M. (NIGAMI LAB) (2026). Cacao bitterness and astringency — where does it change?. https://nigamilab.com/en/topics/cacao-bitterness/
Each claim card has a stable anchor (#clm-…). The dataset is CC BY 4.0; see the data page. Dataset DOI: 10.5281/zenodo.22739101