Research note 036 — 2026-09-04
Roasting Makes Bitterness, and Deeper Roasting Changes It
Neither green coffee beans nor fresh tea leaves yet taste the way they taste now, before they are roasted.coffee roasting and the roasting (hoji) of hojicha are processes that add heat to raw material and remake it into a different set of substances. So what does that process do to bitterness? Tracing through the sources, the answer does not fit into a single sentence like "the deeper the roast, the more bitter it gets." There is not even a common yardstick for what to call a roast level, and what a darker roast increases is not only the amount of bitterness, but the very lineup of compounds that carry it.
On this page (10 sections)
- What this page establishes
- Roast-level terms have no common yardstick
- Darker roasts are reported to score more bitter
- Roasting remakes compounds
- In hojicha too, compounds are swapped, not erased
- Caffeine falls with roasting while bitterness rises
- The choice to roast light
- How the shelf talks about dark-roast bitterness
- What is not known
- Frequently asked
What this page establishes
- Several sources note that the names for roast level (from light roast to Italian roast) have no industry-wide common standard.
- One study finds that trained assessors' bitterness sensory scores rise significantly the deeper the roast, from light roast to dark roast.
- Multiple sources confirm that as dark roasting intensifies bitterness, chlorogenic acid also decreases while caffeic acid increases — a swap of compounds, not just a change in amount.
- Whether the shift toward light roasting has rewritten consumers' own expectation of bitterness has not yet been confirmed.
Roast-level terms have no common yardstick
The All Japan Coffee Association presents an 8-stage roast-level terminology, from light roast to Italian roast, together with the general rule that the lighter the roast, the stronger the acidity, and the darker the roast, the more bitterness is felt.
Scope: A column article on the All Japan Coffee Association's official website
Show data
process-coffee-roasting —has_standardized_vocabulary_entry_in→ "The 8-stage roast-level terminology listed in the 「知る・楽しむ」 (in translation: "Learn and Enjoy") section of the All Japan Coffee Association's official website — light roast, cinnamon roast, medium roast, high roast, city roast, full city roast, French roast, and Italian roast."Yet from within the industry comes the observation that there is no universal agreement on this kind of naming, and one roaster's "light roast" can even be darker than another roaster's "dark roast."
Scope: Concerns raised about roast-level terminology by the SCA, people connected with the UC Davis Coffee Center, and trade publications
Note: A peer-reviewed journal article (Anokye-Bempah et al. 2025) and multiple SCA official articles and trade-press articles independently raise the same concern. This fact calls into question the very premise that 「地域名のついた焙煎度呼称が実際にその地域の慣行を一意に指す」 (in translation: "a roast-level term carrying a regional name actually refers uniquely to that region's practice").
Show data
process-coffee-roasting —has_unresolved_property→ "There is no universally agreed standard in the roasting industry for what "light roast," "medium roast," and "dark roast" mean; one roaster's "light roast" can even be darker than another roaster's "dark roast." The SCA itself states that terms carrying regional names, such as French, Italian, and Full City, can be "too arcane or flexible to be meaningful.""The color-disk kit officially distributed by the SCA (Specialty Coffee Association) is one tool that tries to line up this naming gap with a physical yardstick. Even talking about roast level, it turns out, is still a work in progress.
Scope: Product description on the SCA's official store
Note: The SCA's own official description of a visual reference tool it distributes. The accompanying manual is authored by Carl Staub. It is positioned not as an academic definition of roast level but as an industry consensus-building tool for roasters and customers to reach a shared understanding of roast level.
Show data
process-coffee-roasting —is_classified_by→ "The "Agtron Roast Color Kit" officially distributed by the SCA (Specialty Coffee Association). It consists of 8 color disks housed in a white vinyl case, numbered in 10% increments from "Very Light" (tile #95) to "Very Dark" (tile #25)."Darker roasts are reported to score more bitter
The SCA's official sensory handbook states that acidity is dominant in light roasts and bitterness is dominant in dark roasts, and adds that the amount of caffeine actually decreases as roasting progresses. Bitterness increases with roasting even though caffeine itself, as a compound, shrinks — this process is not simple addition.
Scope: Chapter 7 (the chemistry of coffee's taste) of the SCA's official Coffee Sensory and Cupping Handbook (2021)
Note: Original text: "Sourness is preponderant in light roasts and bitterness in dark roasts." "bitterness actually increases with the roast level". It explicitly states that caffeine is only one part of coffee's bitterness compounds, and that bitterness increases even though caffeine decreases with roasting ("caffeine ... only accounts for a small percentage of coffee's bitterness").
Show data
process-coffee-roasting —increases→ sensation-bitterMeasured results point the same way. Trained assessors' sensory scores rose significantly as roasting progressed from light to dark, and even at the same roast level, the extraction condition (ultra-high-pressure processing versus conventional cold brew) shifted the score itself. Roast level and extraction condition are two separate dials that move bitterness independently.
Scope: Ethiopian arabica beans; comparison of ultra-high-pressure-processed cold brew and conventional cold brew
Note: As roast degree deepened from Agtron color value 73 (light) → 62 (medium) → 53 (dark), the bitterness score increased significantly (ultra-high-pressure: 2.00±0.10 → 3.35±0.30 → 5.50±0.40; conventional: 3.00±0.28 → 4.30±0.42 → 6.50±0.42; p<0.05 in both). This is an academic source that directly shows, with statistical significance from a human sensory panel, the relationship that deeper roasting means stronger bitterness. The same directional trend was confirmed regardless of extraction method (both of the 2 extraction conditions: ultra-high-pressure and conventional).
Show data
process-coffee-roasting —can_change→ "Sensory bitterness score (rated out of 10 by 12 trained assessors holding Q Arabica Grader certification)"What happens across the moment of drinking also changes with roast level. In medium roasts, bitterness and roast aroma were recorded as the dominant sensations within the time course; in dark roasts, bitterness and burnt aroma were dominant instead. The same word "bitterness" keeps different company depending on which aroma stands next to it.
Scope: Comparison of Temporal Dominance of Sensations (TDS) between medium-roast and dark-roast Arabica
Note: For medium roast, bitterness and roasty aroma dominate; for dark roast, bitterness and burnt aroma dominate. TDS is not an intensity curve.
Show data
process-coffee-roasting —can_change→ "The temporal composition of bitterness, roasty aroma, and burnt aroma that dominate during drinking"Roasting remakes compounds
In Doi Chang-grown arabica from northern Thailand, raising the roast level has been shown to decrease the content of chlorogenic acid, while the content of its degradation product, caffeic acid, increases.
Scope: Comparison of light roast (180°C), medium roast (200-220°C), and dark roast (230°C); total phenolic content quantified by the Folin-Ciocalteu method, and chlorogenic acid and caffeic acid quantified by UPLC
Note: The lightness L* value correlated positively with chlorogenic acid content (r=+0.9101) and negatively with caffeic acid content (r=-0.9360). This is a single lot (n=1) from a single farm, and no human sensory panel was conducted.
Show data
process-coffee-roasting —can_change→ compound-chlorogenic-acidA comparison of northern Thai arabica from a different origin and a different analytical method recorded the same pattern: chlorogenic acid lower in roasted beans than in green beans, and caffeine instead higher in roasted beans. Roasting is not a process that lowers every compound uniformly.
Scope: Quantification by UV-Vis spectrophotometry (330 and 272 nm, 70% ethanol extraction), compared against a standard calibration curve
Note: หนองน้ำดำ (Nong Nam Dam) village is not necessarily the same growing area as Doi Tung/Doi Chang, but it connects to this cluster in the context of arabica grown under a Royal Project in northern Thailand. Only the abstract was checked (the full-text PDF was rasterized and could not be extracted).
Show data
compound-chlorogenic-acid —has_measured_content_in→ ingredient-coffee-brewThe same Doi Chang study also recorded a change, with roast level, in the content of furfuryl alcohol, an aroma compound, and connects this to the context in which a burnt or roasted smell is perceived as bitterness. This study itself, though, did not run a human bitterness sensory panel.
Scope: Volatile compound analysis of Doi Chang-grown arabica (HS-SPME-GC/MS, comparing light, medium, and dark roasts)
Note: The body of the paper connects the change in furfuryl alcohol content with roast level to a description, via the Maillard reaction, that 「焦げた・ロースト香が苦味として知覚される(perceived as a bitter taste)」. However, this description is a discussion grounded in other literature, not the study's own human sensory measurement, so it is treated as contextual.
Show data
process-coffee-roasting —is_associated_with→ sensation-bitterCoffee's bitterness has been recorded to draw on a group of compounds generated by roasting itself, beyond caffeine. Part of the lead role in bitterness rises up not from within the green bean, but within the process of roasting.
Scope: Multiple lines of evidence: human sensory testing, receptor assays, and quantitative chemistry
Note: A group of counterexamples to the common belief that 「コーヒーの苦味＝カフェイン」 (in translation: "coffee's bitterness equals caffeine"). The quantitative contribution ratio is not recorded, since the full text was not confirmed.
Show data
ingredient-coffee-brew —has_bitter_contributors_other_than→ compound-caffeineIn hojicha too, compounds are swapped, not erased
Records from a tea experiment station in Kochi Prefecture show that total catechin content fell as roasting deepened, across every tea flush. Some individual catechins, however, first rose from aracha to light roast before falling as roasting deepened further — a hump-shaped movement. The decrease is not uniform.
Scope: Kochi Prefectural Agricultural Technology Center Tea Experiment Station; crude teas of first flush, cut-leaf, and second flush with a 50-second steaming time. Roasted 2 times with an automatic continuous hojicha roaster (exhaust temperature 90 C for the 1st roast, 100 C for the 2nd roast, with drum rotation speed adjusted for each flush and roasting degree).
Note: Direct quantification by HPLC (JASCO X-LC system). Only 1 sample per condition; the number of replicates is not stated.
Show data
process-hojicha-roasting —reduces→ compound-catechinsAmino acids move even more sharply. Roasting sharply lowers the total amount, and also narrows the range of amino acids that remain, so that only a few kinds were left in the most deeply roasted samples.
Scope: Kochi Prefectural Agricultural Technology Center, Tea Industry Experiment Station; the same first-flush, karibancha, and second-flush tea samples as above.
Note: The remarks column of the report classifies arginine, leucine, lysine, and γ-aminobutyric acid under the 「苦味」 (in translation: "bitterness") taste category, but this is a general taste classification for individual amino acids, not a measurement of hojicha's own sensory bitterness. The connection to bitterness perception is unconfirmed.
Show data
process-hojicha-roasting —has_composition→ "Total amino acid content (mg per 100 ml of tea liquor) decreased sharply with roasting: for first-flush tea, aracha 33.52 → light roast 4.10 → medium roast 3.90 → dark roast 3.10; for karibancha, 15.32 → 2.77 → 2.38 → 2.32; for second-flush tea, 4.10 → 2.87 → 2.74 → 1.97. The aracha contained 11 kinds of amino acids for first-flush tea and karibancha, and 8 kinds for second-flush tea; in the light-roast first-flush tea, 8 kinds remained, including theanine, aspartic acid, glutamic acid, and arginine, while in the medium- and dark-roast karibancha and second-flush tea, only 3 kinds remained — γ-aminobutyric acid, glutamic acid, and serine. Among these, γ-aminobutyric acid accounted for about 70% of the total amino acids."At the same time, some aroma compounds absent from aracha were recorded only after roasting. Roasting does not just remove compounds — it is also a process that generates ones that did not exist before.
Scope: Kochi Prefectural Agricultural Technology Center, Tea Industry Experiment Station; aroma compound analysis by GC-MS (extracted by continuous steam distillation).
Note: Reported by a figure (radar chart); no numerical table or number of replicates is given, and no statistical test is shown.
Show data
process-hojicha-roasting —can_generate→ "2-ethyl-3,5-dimethylpyrazine and tetramethylpyrazine, aroma compounds of roast fragrance, were not detected in aracha but were detected after roasting (except for tetramethylpyrazine in second-flush tea). By roast level, the amounts of both compounds were higher the lighter the roast for first-flush tea and karibancha, while second-flush tea showed no difference by roast level. By tea flush, first-flush tea had the most, followed by karibancha, with second-flush tea the least."With the same raw material, the very description of the aroma changed with roast depth. One sample at medium roast was described as having a "bitter aroma, a burnt roast smell," while the same material roasted lighter was described as having a "pleasant roast aroma."
Scope: Ooishida-kuki stem tea from Suizawa, Mie Prefecture; sample descriptions from Table I.
Note: This is an olfactory aroma description (aroma descriptor) by the authors, not a taste evaluation of bitterness as tasted in the mouth. The number of evaluators and the scale are not stated, and the specific aromatic quality that "Bitter aroma" refers to cannot be identified from the text. Its connection to bitterness perception in general is treated cautiously.
Show data
process-hojicha-roasting —has_sensory_attribute→ "A medium-roast sample of Ooishida-kuki stem tea from Suizawa, Mie Prefecture, roasted at 150°C for 12 minutes, was described in the paper's aroma evaluation as "Bitter aroma, burnt roasted aroma." The same raw material roasted at 130°C for 10 minutes as a light-roast sample was described as "Preferable roasted aroma," and the unroasted aracha was "Fresh greenish, toasted laver-like.""And it is not only roast "depth" that moves the reading of bitterness — "method" does too. In a comparison of the same fresh-leaf batch of yellow tea roasted by charcoal, electricity, and drum, the bitterness output value shown by an electronic tongue was higher for charcoal roasting than for the other two methods. Even holding roast level constant, a different way of delivering heat changes the reading of bitterness.
Scope: Chinese large-leaf yellow tea; a comparison of 3 roasting methods from a single fresh-leaf batch.
Note: The e-tongue output values are a dimensionless sensor response and are not generalized to human sensory evaluation or to Japanese green tea/hojicha. Contrary to the popular narrative that charcoal roasting mellows bitterness and astringency, it is noteworthy that in this e-tongue measurement, charcoal roasting showed the highest bitterness output value among the 3 methods.
Show data
process-charcoal-roasting —has_sensory_response→ "Large-leaf yellow tea made from the same fresh-leaf batch (picked June 20, 2023) was processed under charcoal roasting (CR: roasting-basket surface temperature 145±5°C, leaf temperature 125±10°C), electric roasting (ER: dryer temperature 145°C, leaf temperature 125±5°C), and drum roasting (DR: drum inner-wall temperature 145°C, leaf temperature 90±5°C), all for 95 minutes or more, and taste was measured with an electronic tongue (e-tongue). The bitterness output values were CR 3.10±0.12, ER 2.54±0.17, and DR 2.32±0.23, with CR significantly higher than the other 2 methods (at the p=0.05 level, only CR carried a different statistical letter from the others). The astringency output values were highest for DR at 13.07±0.24, followed by CR at 10.43±0.06 and ER at 9.61±0.14, with significant differences between every pair."Caffeine falls with roasting while bitterness rises
As the earlier SCA handbook showed, coffee's bitterness rises with roasting even as caffeine content falls. The hojicha records show barely any decrease in caffeine at light roast, but a 6-13% decrease at dark roast. The process of roasting does not keep assigning caffeine the leading role in bitterness.
Scope: Kochi Prefectural Agricultural Technology Center Tea Experiment Station, 2014; light-roast and deep-roast hojicha made from crude first-flush and cut-leaf tea (roasted 2 times with an automatic continuous hojicha roaster).
Note: Direct quantification by HPLC (Shimadzu).
Show data
process-hojicha-roasting —reduces→ compound-caffeineThe choice to roast light
Some movements have stepped in the opposite direction from intensifying bitterness through dark roasting. The SCA's official handbook records that while one culture softens the strong bitterness of dark-roast, high-robusta coffee with sugar and milk, another culture takes a different approach: lowering coffee's own bitterness level by using light- to medium-roast arabica.
Scope: Chapter 7 of the official SCA "Coffee Sensory and Cupping Handbook" (2021), the passage on differences in bitterness by roast degree and cultivar
Note: Original text: "Other cultures have 'solved' the coffee bitterness problem using a different approach, using light- or medium-roasted C. arabica coffee. The level of bitterness of these brews is much lower than with dark-roasted coffee." The term "specialty coffee" itself does not appear in this passage, but in industry usage, light-roast, arabica-centered specialty coffee falls on this lower-bitterness side.
Show data
concept-specialty-coffee —has_perceived_bitterness_associated_with→ "The official SCA handbook states that while some cultures have a tradition of softening the bitterness of strong, dark-roast, high-robusta-ratio coffee with sugar or milk, "other cultures have 'solved' the coffee bitterness problem using a different approach, using light- or medium-roasted C. arabica coffee. The level of bitterness of these brews is much lower.""The Nordic light-roast style is described as having been started concretely by identifiable individual roasters. One roaster recalls that the turning point was noticing that his own roasted beans tasted burnt compared with the coffee samples brought in by exporters.
Scope: Interviews (2023) with specialty coffee roasters in Norway and Denmark
Note: A 2023 interview article from the trade publication Perfect Daily Grind. Note that while the roasting style itself carrying the regional name "Nordic" traces back to specific roasters and a specific period, the person who established the style himself voices disagreement with that very regional-name label. Whether a term carrying a regional name uniquely refers to that region's practice is not generalized here.
Show data
process-coffee-roasting —has_historical_origin→ "The so-called "Nordic light roast" style traces back to specific, named individual roasters — Tim Wendelboe, who opened a roastery in Oslo, and Klaus Thomsen, co-founder of Copenhagen's Coffee Collective, among them. Wendelboe recounts that the turning point was noticing that his own roasted beans tasted burnt compared with the coffee samples brought in by exporters. Thomsen himself, while saying he does not like the term "Nordic roasting," acknowledges that this approach itself — roasting as light as possible and holding back coffee's inherent development — carries major worldwide influence."Still, the choice to roast lighter does not necessarily rewrite what is inside a consumer's head. In a Brazilian consumer survey, bitterness came up as the basic taste expected of specialty coffee, second only to sweetness. The researchers add that this is despite this kind of coffee having lower bitterness than commodity coffee.
Scope: A digital questionnaire survey of Brazilian consumers (n=1,005) on the basic tastes expected in specialty coffee
Note: Original text: "Bitter is the second basic taste they expect to find in specialty coffee (41.6%), even though this type of coffee has a lower bitterness compared to commodity coffee." An observation of a gap between consumer expectation (expecting bitterness) and the industry's actual bitterness level (lower than commodity coffee). Material that complicates the simplification that differentiated coffee = putting bitterness front and center.
Show data
concept-specialty-coffee —has_sensory_attribute→ "In a survey of 1,005 Brazilian consumers, "bitter" (41.6%) was the 2nd most cited basic taste expected in specialty coffee, after "sweet" (71.4%). The researchers add that this is "even though this type of coffee has a lower bitterness compared to commodity coffee.""How the shelf talks about dark-roast bitterness
Sometimes a product's own words put dark-roast bitterness right on the label. The product page for a top-tier unsweetened black canned coffee carries the headline "a refined bitterness and lingering note," tying it directly to the roasting method.
Scope: UCC's official online store; Japan's mass-distribution RTD canned-coffee market, unsweetened-black segment
Note: Original headline: 「上質な苦みと余韻」 (in translation: "a refined bitterness and lingering note"). Original body copy states that, through its particular roasting method and 2-temperature extraction (high and low), it draws out a 「上質な苦み」 (in translation: "refined bitterness") and 「余韻」 (in translation: "lingering note") free of off-flavors. This is an example on the mass-distribution side (canned coffee) that foregrounds bitterness rather than removing it, and it is material that does not allow a simplified H10 reformulation.
Show data
concept-rtd-canned-coffee —has_sensory_attribute→ "The product page on UCC Ueshima Coffee Co.'s official online store for the mass-distributed unsweetened black canned coffee 『UCC BLACK無糖 黒の余韻』 (in translation: "UCC BLACK Unsweetened: The Lingering Note of Black") carries the headline 「上質な苦みと余韻」 (in translation: "a refined bitterness and lingering note"), and explains in the body copy that, through its particular roasting method and 2-temperature extraction (high and low), it draws out a 「上質な苦み」 (in translation: "refined bitterness") and 「余韻」 (in translation: "lingering note") free of off-flavors."The same pattern shows up at another brand. Its morning-only unsweetened black line foregrounds "a crisp, bitter taste," while the same brand's slightly-sweet line makes no mention of bitterness at all in its description.
Scope: Manufacturer's official product page (retrieved 2026-09-04)
Note: An example contrasting unsweetened black against lightly sweetened within the same brand.
Show data
concept-rtd-canned-coffee —is_documented_as_having→ "Asahi Soft Drinks' WONDA brand, in the description of its unsweetened black "Morning Shot" line for the morning, explicitly holds up bitterness as a value word: "a crisp, bitter taste to get you cheerfully into work." Meanwhile, the description of its lightly sweetened "Kin no Bitō" (Gold, lightly sweetened) line makes no mention of bitterness at all, holding up "just the right sweetness" instead."On the other hand, there are brand lineups that use not a single instance of bitterness vocabulary across their entire current lineup. 5 products from one canned-coffee brand are described using only the words for body/richness, aroma, crispness, milk feel, and sweetness. The story of how dark roasting creates bitterness, and the story of how the shelf names that bitterness, are two different stories.
Scope: Manufacturer's official product page (retrieved 2026-09-04)
Note: No bitterness vocabulary either, including 「香るブラック」 (in translation: "Kaoru Black"), the unsweetened-black-equivalent product.
Show data
concept-rtd-canned-coffee —is_documented_as_having→ "Among the current lineup listed on Coca-Cola Japan's official Georgia product-list page, all 5 products — Emerald Mountain Blend, Emerald Mountain Blend Shifuku no Bitō, Kaoru Black, Kaoru Bitō, and The Blend — use not a single instance of 「苦味」 (in translation: "bitterness") vocabulary, and are instead standardized on the words 「コク」「香り」「すっきり」「ミルク感」「甘さ」 (in translation: "body/richness," "aroma," "crisp/clean," "milk feel," and "sweetness")."What is not known
- Sources that directly measured the correspondence between roast level and bitterness sensory score are limited to experiments with a single origin and a single lot, so generalizing to coffee as a whole cannot be confirmed here.
- Sources confirming the decrease in chlorogenic acid and the increase in caffeic acid are all compound quantifications and have not been connected to a human bitterness sensory panel.
- Why the shift toward light roasting does not also rewrite consumers' expectations (their association of bitterness) does not have an answer within the sources confirmed here.
Frequently asked
Does coffee get more bitter the darker it is roasted?
One study reports that trained assessors' sensory scores rose significantly as roasting progressed from light roast to dark roast. The SCA's official sensory handbook likewise organizes it as acidity being dominant in light roasts and bitterness dominant in dark roasts. However, this sensory experiment is a result for a single origin, a single lot, and cold brew, and whether the same degree of increase in bitterness holds for a different bean origin or extraction method is not confirmed here.
Does coffee's bitterness get stronger because of its caffeine content?
The SCA's official handbook states that the amount of caffeine actually decreases as roasting progresses, so bitterness intensity and caffeine content do not move in the same direction. Multiple sources report that quinides and other roasting-generated compound groups, beyond caffeine, contribute to bitterness, suggesting that coffee's bitterness cannot be reduced to a single compound.
Does hojicha get more bitter the more deeply it is roasted?
Records from a Kochi Prefecture tea experiment station report that total catechin content decreased with roast depth across first-flush, karibancha, and second-flush teas (for example, from 34.89 in first-flush aracha to 6.98 at dark roast). Meanwhile, one sample's aroma description is also recorded as changing from a "bitter aroma, a burnt roast smell" at medium roast (150°C, 12 minutes) to a "pleasant roast aroma" at a lighter roast (130°C, 10 minutes). This aroma description, however, is a single-material case and is not accompanied by a human bitterness sensory panel.
Related items:coffee roasting / chlorogenic acid / caffeine / hojicha roasting / specialty coffee / RTD canned coffee
Related research note: Standards Certify Bitterness, and the Shelf Chooses It
Source ledger: Source list
How to cite this page
Aoki, M. (NIGAMI LAB) (2026). Roasting Makes and Changes Bitterness. https://nigamilab.com/en/notes/roasting-makes-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