Topic hub — 2026-10-20
Arabica or robusta: which is more bitter?
Robusta has a reputation as the bitter one. What do studies in which people actually tasted the coffee say? This page lines up the included studies together with their conditions: how many beans and from where, how they were roasted and brewed, and who rated them on what scale. It then separates what was established by changing a condition within one experiment, what differences between studies might explain, and what this comparison alone cannot separate.
On this page (10 sections)
- What this page establishes
- What the included studies allow us to say
- Bitterness ratings at a glance
- Conditions and results, study by study
- What was established within an experiment
- Candidate explanations for the differences
- What is not known — what this comparison alone cannot separate
- How studies and rows were counted
- What we looked at
- Read in the encyclopedia
What this page establishes
- In the study that matched roasting and compared bitterness intensity with statistical tests, robusta was more bitter when averaged over conditions; but within that same study, robusta did not come out ahead in some brewing combinations.
- In another study, the same pair of beans swapped which one was more bitter when the brewing method changed.
- What one study confirmed with tests as stronger in robusta regardless of brewing was astringency, not bitterness.
- Robusta is higher in some bitterness-related compounds and arabica in others, so compound amounts alone do not decide the direction of bitterness.
What the included studies allow us to say
- Among the included studies, 1 (Ugandan coffees made as cold brew) compared bitterness intensity under controlled conditions with statistical tests, and averaged over its conditions robusta was significantly more bitter. But within that study robusta did not come out ahead in some brewing combinations, and in another study using hot brewing, the same pair of beans swapped which one was more bitter when the brewing method changed. [rows: Seninde 2018 average; Seninde 2018 combinations; Šeremet 2022 Turkish; Šeremet 2022 filter]
- So “robusta is more bitter” is a tendency that can appear once the beans and the brewing are fixed, not something true of every cup. In the hot-brew study of 6 single-origin beans, what was confirmed with statistical tests, for every brewing method, as stronger in robusta was astringency, not bitterness. [rows: Šeremet 2022 astringency]
- On the common claim that robusta is more bitter because it has more caffeine: robusta did have more caffeine in the included studies too (19.9 vs 15.7 mg/g in green beans; 167–175 vs 121–123 mg/100 mL in cold brew). Yet as Turkish coffee, robusta RC_6, whose brew had more caffeine (1.71 mg/mL) than arabica RC_3 (1.10 mg/mL), was rated significantly less bitter than RC_3. With these samples and brewing methods, more caffeine did not necessarily mean more bitter. [rows: Caporaso 2018 caffeine; Portela 2020 caffeine; Šeremet 2022 caffeine in brew; Šeremet 2022 Turkish]
Bitterness ratings at a glance
Only the results in which people rated bitterness. "Direction" is the larger of the two values as reported; very small gaps that the paper does not treat as a difference are shown as "about the same". "Test" says whether the paper checked the arabica–robusta difference statistically. Select a study name to jump to its detailed table.
Results in which people rated bitterness only. Values are as reported; scales and methods differ between studies, so compare within a row, not across rows.
| Study | Condition | Measured as | Arabica | Robusta | Direction / test |
|---|---|---|---|---|---|
| Vezzulli 2022 | Espresso (1:2) | median on a 0–9 scale (6 panelists) | 4.0 | 4.7 | Robusta higher No test of the difference |
| Vezzulli 2022 | Filter (home drip machine; 50 g/L) | median on a 0–9 scale (6 panelists) | 1.2 | 3.5 | Robusta higher No test of the difference |
| Vezzulli 2022 | Moka (76 g/L) | median on a 0–9 scale (6 panelists) | 3.4 | 3.3 | About the same No test of the difference |
| Vezzulli 2022 | Neapolitan pot (72 g/L) | median on a 0–9 scale (6 panelists) | 3.1 | 2.9 | About the same No test of the difference |
| Šeremet 2022 | Espresso (1 g to 20 mL) | mean on a 1–10 scale (20 panelists) | RC_3 8.5 and RC_4 8.5 (decaf RC_1 7.5 and RC_2 9) | RC_5 9 and RC_6 9.5 | Robusta higher Tested, not significant |
| Šeremet 2022 | Turkish (1 g to 10 mL, boiled) | mean on a 1–10 scale (20 panelists) | RC_3 6.5 and RC_4 6 (decaf RC_1 6.5 and RC_2 5.5) | RC_5 5.5 and RC_6 5 | Arabica higher Tested, only some pairs significant |
| Šeremet 2022 | Filter (1 g to 15 mL, 92 °C) | mean on a 1–10 scale (20 panelists) | RC_3 4.5 and RC_4 5.5 (decaf RC_1 4.5 and RC_2 5.5) | RC_5 5.5 and RC_6 6.5 | Robusta higher Tested, only some pairs significant |
| Seninde 2018 | Mean over the 8 cold-brew conditions (main effect) | least-squares mean on a 0–15 scale (5 panelists × 3 replicates) | AW 9.36 and AD 9.24 | RW 10.28 and RD 10.08 | Robusta higher Tested, significant |
| Seninde 2018 | The 8 combinations of 2 roast levels × 2 doses × 2 brewing methods | 0–15 scale | Values per combination are in Figure 2-3 (a scatter plot) | Same | Depends on the combination No test of the difference |
| Portela 2020 | Cold brew, coarse grind, 5 °C | maximum rate at which bitterness was the dominant sensation (0–1, 24 people) | 0.29 | 0.31 | About the same No test of the difference |
| Portela 2020 | Cold brew, fine grind, 15 °C | maximum rate at which bitterness was the dominant sensation (0–1, 24 people) | 0.26 | 0.36 | Robusta higher No test of the difference |
Conditions and results, study by study
For each study, the conditions for beans, roasting, brewing and rating, and the results. Each row is one study × one measured result, so different brewing methods or compounds in the same study are separate rows. Under each row we give where it is in the paper and whether it can be used to compare the direction of bitterness between the species, with the reason. Nothing is left blank: "not given" (not stated in the paper), "not read" (not checked by us) and "not applicable" are kept apart.
Vezzulli 2022 sensory only
Metabolomics Combined with Sensory Analysis Reveals the Impact of Different Extraction Methods on Coffee Beverages from Coffea arabica and Coffea canephora var. Robusta. Foods 11(6):807 — doi:10.3390/foods11060807
- How deeply checked
- Checked in full text — Read in the full text on PMC and its Table 2
- Samples
- One commercial blend of 100% arabica (Brazil, natural) and one of 100% robusta (India, natural); different countries of origin; one sample each
- Roasting
- Roasted by the same roaster with the same roasting process (no roast-degree value reported)
- Brewing
- 4 methods: espresso (coffee to beverage 1:2 by weight, about 16 g, 92 °C), filter (50 g/L), moka (76 g/L) and Neapolitan pot (72 g/L)
- Who rated, on what scale
- 6 trained panelists; scale from 0 (absent) to 9 (very intense); the table gives the panel median
| Measured | Condition | Arabica | Robusta | Unit | Direction / test |
|---|---|---|---|---|---|
| Bitterness intensity (sensory) | Espresso (1:2) | 4.0 | 4.7 | median on a 0–9 scale (6 panelists) | Robusta higher No test of the difference |
| No test of the species difference is reported; bitterness is not among the descriptors the ANOVA in the text found to differ significantly Where: Table 2, Section 3.3 For the direction of bitterness between arabica and robusta: Limited — Two samples from the same roasting process rated by the same panel; but one commercial blend per species from different countries, a 6-person median, and no test of the species difference | |||||
| Bitterness intensity (sensory) | Filter (home drip machine; 50 g/L) | 1.2 | 3.5 | median on a 0–9 scale (6 panelists) | Robusta higher No test of the difference |
| As above Where: Table 2, Section 3.3 For the direction of bitterness between arabica and robusta: Limited — As above | |||||
| Bitterness intensity (sensory) | Moka (76 g/L) | 3.4 | 3.3 | median on a 0–9 scale (6 panelists) | About the same No test of the difference |
| As above Where: Table 2 For the direction of bitterness between arabica and robusta: Limited — As above | |||||
| Bitterness intensity (sensory) | Neapolitan pot (72 g/L) | 3.1 | 2.9 | median on a 0–9 scale (6 panelists) | About the same No test of the difference |
| As above Where: Table 2 For the direction of bitterness between arabica and robusta: Limited — As above | |||||
Šeremet 2022 sensory + composition
Antioxidant and Sensory Assessment of Innovative Coffee Blends of Reduced Caffeine Content. Molecules 27(2):448 — doi:10.3390/molecules27020448
- How deeply checked
- Checked in full text — Read in the full text on PMC, Tables 1, 2 and 6, and Supplementary Tables S1–S3 (the statistics for bitterness)
- Samples
- 6 single-origin beans: 4 arabica (RC_1 Mexico decaf, RC_2 Colombia decaf, RC_3 Brazil, RC_4 Costa Rica) and 2 robusta (RC_5 Guatemala, RC_6 Brazil)
- Roasting
- Conditions set by trial and error for each bean (temperature and time differ by bean); all medium roast
- Brewing
- Espresso (1 g coffee to 20 mL water), Turkish (1 g to 10 mL, boiled) and filter (1 g to 15 mL at 92 °C)
- Who rated, on what scale
- 20 staff with experience in coffee sensory evaluation; scale from 1 (detectable) to 10 (very intense) (the paper calls it a 10-point hedonic scale, but its end points are intensities); one-way ANOVA with Tukey's test
| Measured | Condition | Arabica | Robusta | Unit | Direction / test |
|---|---|---|---|---|---|
| Bitterness intensity (sensory) | Espresso (1 g to 20 mL) | RC_3 8.5 and RC_4 8.5 (decaf RC_1 7.5 and RC_2 9) | RC_5 9 and RC_6 9.5 | mean on a 1–10 scale (20 panelists) | Robusta higher Tested, not significant |
| None of the pairs between the 2 non-decaf arabicas and the 2 robustas differ significantly; decaf RC_1 is significantly lower than RC_2, RC_5 and RC_6. The statement in the abstract and Section 3.4 that decaf arabica was more bitter does not match this table Where: Supplementary Table S1 For the direction of bitterness between arabica and robusta: Limited — Same panel, same scale, with a test; but roasting was set per bean, samples were 4 arabica (2 decaf) and 2 robusta, and origins differ | |||||
| Bitterness intensity (sensory) | Turkish (1 g to 10 mL, boiled) | RC_3 6.5 and RC_4 6 (decaf RC_1 6.5 and RC_2 5.5) | RC_5 5.5 and RC_6 5 | mean on a 1–10 scale (20 panelists) | Arabica higher Tested, only some pairs significant |
| RC_6 (robusta from Brazil) is significantly lower than RC_3 (arabica from Brazil) and RC_1; other pairs do not differ significantly Where: Supplementary Table S2 For the direction of bitterness between arabica and robusta: Limited — As above | |||||
| Bitterness intensity (sensory) | Filter (1 g to 15 mL, 92 °C) | RC_3 4.5 and RC_4 5.5 (decaf RC_1 4.5 and RC_2 5.5) | RC_5 5.5 and RC_6 6.5 | mean on a 1–10 scale (20 panelists) | Robusta higher Tested, only some pairs significant |
| RC_6 (robusta from Brazil) is significantly higher than RC_3 (arabica from Brazil) and RC_1; other pairs do not differ significantly Where: Supplementary Table S3 For the direction of bitterness between arabica and robusta: Limited — As above | |||||
| Astringency (sensory) | All 3: espresso, Turkish and filter | Espresso 3–4.5, Turkish 2.5–3.5, filter 3 (4 beans) | Espresso 7, Turkish 6–6.5, filter 5.5 (2 beans) | mean on a 1–10 scale (20 panelists) | Robusta higher Tested, significant |
| In every brewing method, each of the 2 robustas is significantly more astringent than each of the 4 arabicas Where: Supplementary Tables S1–S3, Section 3.4 For the direction of bitterness between arabica and robusta: Not usable — This is astringency, not bitterness; it is listed because the two sensations are easy to confuse | |||||
| Caffeine | Caffeine in the brew (espresso / Turkish / filter) | RC_3 3.91 / 1.10 / 0.64 and RC_4 3.63 / 0.90 / 0.53 | RC_5 4.94 / 1.83 / 0.86 and RC_6 8.65 / 1.71 / 1.17 | mg/mL | Robusta higher No test of the difference |
| The tests in Table 6 compare brewing methods within one bean; the species difference is not tested Where: Table 6 For the direction of bitterness between arabica and robusta: Not usable — A compound amount, not perceived bitterness | |||||
| Caffeine | Caffeine in roasted beans (dry basis) | Shown in Figure 2 (values not read) | RC_6 22.38 is the highest of the 6 | mg/g (dry basis) | Robusta higher No test of the difference |
| No test of the species difference is reported Where: Abstract, Figure 2, Section 3.2 For the direction of bitterness between arabica and robusta: Not usable — As above | |||||
Caporaso 2018 composition only
Non-destructive analysis of sucrose, caffeine and trigonelline on single green coffee beans by hyperspectral imaging. Food Research International 106:193-203 — doi:10.1016/j.foodres.2017.12.031
- How deeply checked
- Checked in full text — Read in the full text on PMC
- Samples
- 277 green beans from 27 lots (13 countries, both arabica and robusta) analysed one by one (the count per species was not found in the text)
- Roasting
- Not applicable (green beans)
- Brewing
- Not applicable (each bean ground and extracted in boiling water for HPLC-MS)
- Who rated, on what scale
- Not applicable (no sensory evaluation)
| Measured | Condition | Arabica | Robusta | Unit | Direction / test |
|---|---|---|---|---|---|
| Trigonelline | Single green beans (as-is basis) | 8.6 ± 1.8 | 7.3 ± 1.7 | mg/g (mean ± SD) | Arabica higher Tested, significant |
| The species differ at p < 0.001, though the distributions overlap in the middle Where: Section 3.1 (text) For the direction of bitterness between arabica and robusta: Not usable — As above (a study without sensory evaluation) | |||||
| Caffeine | Single green beans (as-is basis) | 15.7 ± 2.3 | 19.9 ± 4.9 | mg/g (mean ± SD) | Robusta higher Tested, significant |
| The species differ at p < 0.001, though the distributions overlap in the middle Where: Section 3.1 (text) For the direction of bitterness between arabica and robusta: Not usable — As above (a study without sensory evaluation) | |||||
Lang 2015 composition only
Mozambioside Is an Arabica-Specific Bitter-Tasting Furokaurane Glucoside in Coffee Beans. Journal of Agricultural and Food Chemistry 63(48):10492-10499 — doi:10.1021/acs.jafc.5b04847
- How deeply checked
- Abstract only — Abstract only (full text is with the publisher and was not reached)
- Samples
- Arabica and robusta green and roasted beans (the number of samples is not given in the abstract)
- Roasting
- Not given in the abstract
- Brewing
- Not applicable (compound quantification)
- Who rated, on what scale
- A bitter recognition threshold was measured, but the abstract does not give the panel size
| Measured | Condition | Arabica | Robusta | Unit | Direction / test |
|---|---|---|---|---|---|
| Mozambioside | Green beans | 396–1188 | below 5 | nmol/g | Arabica higher No test of the difference |
| The abstract reports no test (robusta has only traces) Where: Abstract For the direction of bitterness between arabica and robusta: Not usable — As above (abstract only) | |||||
Farah 2005 composition only
Effect of Roasting on the Formation of Chlorogenic Acid Lactones in Coffee. Journal of Agricultural and Food Chemistry 53(5):1505-1513 — doi:10.1021/jf048701t
- How deeply checked
- Abstract only — Abstract only (full text is with the publisher and was not reached)
- Samples
- Two arabica cultivars (Bourbon and Longberry) and one robusta
- Roasting
- A series of roast degrees; peaks at light-medium roast (about 14% weight loss). Whether roasting was matched between species is not given in the abstract
- Brewing
- Not applicable (compound quantification)
- Who rated, on what scale
- Not applicable (no sensory evaluation)
| Measured | Condition | Arabica | Robusta | Unit | Direction / test |
|---|---|---|---|---|---|
| Chlorogenic acid lactones | At the peak (light-medium roast) | 3-CQL 230 ± 9 and 4-CQL 116 ± 3 | 3-CQL 254 ± 4 and 4-CQL 139 ± 2 | mg/100 g (dry basis) | Robusta higher No test of the difference |
| The abstract reports no test of the species difference Where: Abstract For the direction of bitterness between arabica and robusta: Not usable — As above (abstract only) | |||||
Seninde 2018 sensory only
Determining the impact of roasting degree, coffee to water ratio and brewing method on the sensory characteristics of cold brew Ugandan coffee. Master's thesis, Kansas State University — source
- How deeply checked
- Checked in the full text of the underlying thesis — Read in the full text of the master's thesis (Kansas State University 2018) that the paper (Food Research International 2020) is based on, including Appendix Table C-1 and Figure 2-3; the paper itself was not reached on the publisher's site
- Samples
- 4 Ugandan green coffees: 2 arabica (washed AW, dry-processed AD) and 2 robusta (washed RW, dry-processed RD); each from a single origin (4 different regions and altitudes), no decaf
- Roasting
- All 4 in the same roaster at 2 levels, medium and dark (set by bean colour L*)
- Brewing
- 2 cold-brew methods (16-hour steeping, 3-hour slow drip) × 8 g and 12 g of coffee per 100 mL; 32 samples in all; served cold (4–7 °C)
- Who rated, on what scale
- 5 highly trained panelists; 0–15 scale (end-point meanings not given); three replicates; mixed model with pairwise comparisons
| Measured | Condition | Arabica | Robusta | Unit | Direction / test |
|---|---|---|---|---|---|
| Bitterness intensity (sensory) | Mean over the 8 cold-brew conditions (main effect) | AW 9.36 and AD 9.24 | RW 10.28 and RD 10.08 | least-squares mean on a 0–15 scale (5 panelists × 3 replicates) | Robusta higher Tested, significant |
| Each of the 2 robustas is significantly more bitter than each of the 2 arabicas Where: Appendix C, Table C-1 (p. 56) For the direction of bitterness between arabica and robusta: Limited — Same roaster, same panel, tested, averaged over 8 conditions; but 2 samples per species, all Ugandan, cold brew only | |||||
| Bitterness intensity (sensory) | The 8 combinations of 2 roast levels × 2 doses × 2 brewing methods | Values per combination are in Figure 2-3 (a scatter plot) | Same | 0–15 scale | Depends on the combination No test of the difference |
| The four-way interaction is significant; tests for individual combinations are not reported. The one example the text states: at medium roast and 12 g, robusta was more bitter than arabica by slow drip but about the same by steeping. Read from the figure, in 5 of the 8 combinations both robustas exceeded both arabicas, and in 3 they were level or reversed (approximate values read from the plot) Where: Text pp. 29–32, Figure 2-3 For the direction of bitterness between arabica and robusta: Limited — No tests per combination are reported, and values are approximate readings from a plot | |||||
Portela 2020 sensory + composition
Impacto da espécie de café e condições de preparo na composição e características sensoriais de bebidas extraídas a frio (cold brew). Dissertação de mestrado, Universidade Estadual de Londrina — source
- How deeply checked
- Checked in the full text of the underlying thesis — Read in chapter 3 of the master's thesis (State University of Londrina 2020) that the paper (Food Research International 2022) is based on, including Tables 3.1 and 3.2; the paper itself is subscription-only and was not read
- Samples
- One arabica (Brazil, Mundo Novo) and one robusta (Brazil, Rondônia); single-origin, no decaf
- Roasting
- Both medium roast (L* 37.4 and 36.9); roasting conditions not given
- Brewing
- Cold brew (10 g coffee per 100 mL water, 15 hours), coarse grind at 5 °C and fine grind at 15 °C
- Who rated, on what scale
- 24 consumers with sensory experience; temporal dominance of sensations (TDS), choosing the most dominant sensation over time rather than rating intensity; maximum dominance rate (0–1)
| Measured | Condition | Arabica | Robusta | Unit | Direction / test |
|---|---|---|---|---|---|
| Bitterness dominance (sensory) | Cold brew, coarse grind, 5 °C | 0.29 | 0.31 | maximum rate at which bitterness was the dominant sensation (0–1, 24 people) | About the same No test of the difference |
| Table 3.2 gives no test of differences between brews; difference curves for each species pair exist (supplementary Figures 3.S1 and 3.S2) but carry no significance marks; bitterness stayed dominant longer for the arabica Where: Table 3.2 For the direction of bitterness between arabica and robusta: Limited — Measures whether bitterness was the most dominant sensation, not its intensity; one sample per species and no test of the difference | |||||
| Bitterness dominance (sensory) | Cold brew, fine grind, 15 °C | 0.26 | 0.36 | maximum rate at which bitterness was the dominant sensation (0–1, 24 people) | Robusta higher No test of the difference |
| Table 3.2 gives no test of differences between brews; difference curves for each species pair exist (supplementary Figures 3.S1 and 3.S2) but carry no significance marks Where: Table 3.2 For the direction of bitterness between arabica and robusta: Limited — As above | |||||
| Caffeine | Cold brew (coarse 5 °C / fine 15 °C) | 121 / 123 | 167 / 175 | mg/100 mL | Robusta higher Tested, significant |
| Significant differences among the four brews (Tukey) Where: Table 3.1 For the direction of bitterness between arabica and robusta: Not usable — A compound amount, not perceived bitterness | |||||
| Chlorogenic acids | Total chlorogenic acids in the cold brew (coarse 5 °C / fine 15 °C) | 301 / 313 | 396 / 411 | mg/100 mL | Robusta higher Tested, significant |
| Significant differences among the four brews (Tukey) Where: Table 3.1 For the direction of bitterness between arabica and robusta: Not usable — As above | |||||
| Trigonelline | Cold brew (coarse 5 °C / fine 15 °C) | 108 / 107 | 58 / 58 | mg/100 mL | Arabica higher Tested, significant |
| Significant difference between species (Tukey) Where: Table 3.1 For the direction of bitterness between arabica and robusta: Not usable — As above | |||||
What was established within an experiment
Only results where the same panel compared changed conditions within one study.
- In a study where 5 trained panelists rated 4 Ugandan coffees (2 arabica, 2 robusta) roasted in the same roaster and cold-brewed 8 ways, averaged over conditions both robustas (10.28 and 10.08) were significantly more bitter than both arabicas (9.36 and 9.24) on a 0–15 scale. [rows: Seninde 2018 average]
- In the same study, at medium roast and 12 g, robusta was more bitter by slow drip but about the same by steeping. Reading the figure, in 3 of the 8 combinations of roast, dose and method, robusta did not come out ahead of arabica. Tests for individual combinations are not reported. [rows: Seninde 2018 combinations]
- In a study where 6 panelists rated 1 commercial blend of each species (Brazilian arabica, Indian robusta) from the same roasting process, brewed hot in 4 ways, the robusta's median bitterness was higher as espresso (4.7 vs 4.0) and filter (3.5 vs 1.2), and about the same as moka (3.3 vs 3.4) and Neapolitan pot (2.9 vs 3.1). No test of the species difference is reported. [rows: Vezzulli 2022 espresso; Vezzulli 2022 filter; Vezzulli 2022 moka; Vezzulli 2022 Neapolitan]
- In a study where 20 panelists compared a Brazilian arabica (RC_3) and a Brazilian robusta (RC_6), the arabica was significantly more bitter as Turkish coffee (6.5 vs 5) and the robusta was significantly more bitter as filter coffee (4.5 vs 6.5). With the same pair of beans and the same panel, changing the brewing method swapped which one was more bitter. [rows: Šeremet 2022 Turkish; Šeremet 2022 filter]
- As espresso in the same study, the bitterness of the 2 non-decaf arabicas (8.5 and 8.5) and the 2 robustas (9 and 9.5) did not differ significantly. The decaffeinated Mexican arabica (7.5) was significantly less bitter than RC_2, RC_5 and RC_6. [rows: Šeremet 2022 espresso]
- Astringency was significantly stronger in each of the 2 robustas than in each of the 4 arabicas, in all 3 brewing methods of the same study. [rows: Šeremet 2022 astringency]
- In another cold-brew study (1 sample each, 24 people), the maximum rate at which bitterness was the dominant sensation was about the same with coarse grind at 5 °C (0.29 vs 0.31) and higher for robusta with fine grind at 15 °C (0.26 vs 0.36). The table gives no test of the difference. [rows: Portela 2020 coarse; Portela 2020 fine]
- For compounds, a bean-by-bean analysis of green beans found more caffeine in robusta (19.9 vs 15.7 mg/g) and more of the bitter compound trigonelline in arabica (8.6 vs 7.3 mg/g). In cold brew too, robusta had more caffeine (167–175 vs 121–123 mg/100 mL) and arabica more trigonelline (107–108 vs 58 mg/100 mL). [rows: Caporaso 2018 caffeine; Caporaso 2018 trigonelline; Portela 2020 caffeine; Portela 2020 trigonelline]
Candidate explanations for the differences
Conditions that differ between studies, or between brewing methods, and could explain differences in bitterness. Several conditions differ at once, so which one mattered is unknown. These are candidates, not causes.
- Brewing. In all 4 included sensory studies, changing the brewing method (and the dose or roast) changed the size or direction of the bitterness difference. But each method differs at once in coffee-to-water ratio, water temperature, time and grind, so which of these mattered is unknown. [rows: Vezzulli 2022 espresso; Šeremet 2022 Turkish; Seninde 2018 combinations; Portela 2020 fine]
- Differences in compounds. Caffeine and chlorogenic acids (and the lactones roasting makes from them), which are higher in robusta, are candidates for bitterness, but arabica is higher in other bitter compounds (trigonelline, mozambioside). As Turkish coffee, robusta RC_6, whose brew had more caffeine (1.71 mg/mL) than the 2 arabicas (1.10 and 0.90 mg/mL), was rated less bitter. Compound amounts alone do not decide the direction of bitterness. [rows: Šeremet 2022 caffeine in brew; Šeremet 2022 Turkish; Portela 2020 chlorogenic acids; Caporaso 2018 trigonelline; Lang 2015 mozambioside; Farah 2005 lactones]
- How roasting was matched. In the 2 studies that roasted both species in the same roaster or process, robusta was more bitter on average or in some brewing methods; in the study that roasted each bean separately, non-decaf beans did not differ significantly as espresso. Beans, panels and brewing also differ between the studies, so the effect of roast matching remains a candidate. [rows: Vezzulli 2022 espresso; Seninde 2018 average; Šeremet 2022 espresso]
- How it was measured. 3 studies scored bitterness intensity; 1 recorded whether bitterness was the most dominant sensation. Bitterness can be strong and still go uncounted if another sensation dominates. [rows: Seninde 2018 average; Portela 2020 coarse]
What is not known — what this comparison alone cannot separate
- The effect of species itself. Per species the sensory studies used 1–4 arabica and 1–2 robusta samples (1–2 of each if decaf is excluded), so species overlaps with origin, processing and roasting. Whether a bitterness difference comes from the species or from that particular bean cannot be separated. [rows: Vezzulli 2022 espresso; Šeremet 2022 espresso; Seninde 2018 average; Portela 2020 fine]
- Why the studies' results differ. Panels (5–24 people with different training), scales (0–9, 1–10, 0–15, dominance rates), roast matching, origins and brewing conditions all differ at once. [rows: Vezzulli 2022 espresso; Šeremet 2022 espresso; Seninde 2018 average; Portela 2020 fine]
- Hot or cold brewing. The 2 hot-brew studies and the 2 cold-brew studies also differ in beans and panels. Whether cold brewing makes robusta come out more bitter cannot be separated by this comparison. [rows: Vezzulli 2022 espresso; Šeremet 2022 espresso; Seninde 2018 average; Portela 2020 fine]
- Which condition within the brewing method swapped the direction (coffee-to-water ratio, temperature, time, boiling, grind). [rows: Šeremet 2022 Turkish; Šeremet 2022 filter; Portela 2020 fine]
- The link between bitterness and compounds. The composition-only studies have no sensory evaluation, and in the sensory studies the link between measured compounds (such as caffeine) and bitterness rests on too few samples to be checked by this comparison. [rows: Caporaso 2018 caffeine; Šeremet 2022 caffeine in brew; Portela 2020 caffeine]
How studies and rows were counted
The number of evidence rows is not the number of studies. Table rows, ledger evidence rows, papers and research groups are counted separately.
- Sources in the table: 7, of which 4 had people rate bitterness; the others measured something else (compounds, smell) or are used only for context.
- Rows in the table (study × measured result): 21. Rows on perceived bitterness: 11.
- Evidence rows in the lab's ledger these rows rest on: 6. One paper can supply several rows. Rows without a ledger evidence row come from other tables of the same papers, or from sources newly read for this question.
- Independent research groups among the sensory studies: 4 (reports from the same lab count as 1).
What we looked at
The sources examined for this question, why each was included or excluded, and which full texts could not be reached. We searched PubMed for papers whose title or abstract mentions arabica, robusta (or canephora) and bitterness, and added the sources already in the lab's ledger.
| Source | Decision | Why |
|---|---|---|
| Metabolomics Combined with Sensory Analysis Reveals the Impact of Different Extraction Methods on Coffee Beverages from Coffea arabica and Coffea canephora var. Robusta (2022) doi | Included in the table | Two species from the same roasting process rated by people in four brewing methods; read in full (CC BY) |
| Antioxidant and Sensory Assessment of Innovative Coffee Blends of Reduced Caffeine Content (2022) doi | Included in the table | Six single-origin beans rated by people in three brewing methods, with bitterness statistics in the supplement; read in full (CC BY) |
| Non-destructive analysis of sucrose, caffeine and trigonelline on single green coffee beans by hyperspectral imaging (2018) doi | Included in the table | Measured bitterness-related compounds by species; no sensory evaluation, so not used for the direction of bitterness |
| Mozambioside Is an Arabica-Specific Bitter-Tasting Furokaurane Glucoside in Coffee Beans (2015) doi | Included in the table | An amount of a bitter compound almost limited to arabica; abstract only (full text with the publisher) |
| Effect of Roasting on the Formation of Chlorogenic Acid Lactones in Coffee (2005) doi | Included in the table | Amounts of bitter lactones formed in roasting by species; abstract only (full text with the publisher) |
| Determining the impact of roasting degree, coffee to water ratio and brewing method on the sensory characteristics of cold brew Ugandan coffee (2020) doi | Included in the table | The paper itself was not reached on the publisher site; read through the full text of the underlying master's thesis (2018). It compares bitterness intensity under controlled conditions with statistical tests |
| Effects of brewing conditions and coffee species on the physicochemical characteristics, preference and dynamics of sensory attributes perception in cold brews (2022) doi | Included in the table | The paper is subscription-only; read through chapter 3 of the underlying master's thesis (2020). It measures dominance of bitterness, not intensity |
| Sensory characterization and acceptance of coffee brews of C. arabica and C. canephora blended with steamed defective coffee (2019) doi | Full text not reachable | The publisher page returned an error, so the full text was not reached. The abstract says the 100% canephora brew was characterised by terms including bitter taste, but no bitterness values or tests were checked. The study is about blending in defective beans |
| Impact of roasting time on the sensory profile of arabica and robusta coffee (2013) doi | Full text not reachable | The repository holding the author version did not respond, so the full text was not reached. The abstract lists bitter flavour among attributes separated by roasting time but does not say which species was more bitter |
| Multivariate methods for characterization and classification of espresso coffees from different botanical varieties and types of roast by foam, taste, and mouthfeel (2001) doi | Excluded | Found by the reviewer with another search. It compares arabica with robusta blends (one of them torrefacto, roasted with added sugar), so it is not a species-to-species comparison; the abstract gives no bitterness values |
| Impact of consumption temperature on sensory properties of hot brewed coffee (2019) doi | Full text not reachable | Found by the reviewer with another search. A trained panel described 2 arabicas and 1 robusta brewed hot at different serving temperatures; the subject is temperature and the abstract gives no bitterness values. Subscription-only; full text not reached (a candidate to read next) |
| Coffee Sensory and Cupping Handbook (2021) | Read as context (not an experiment) | An industry handbook describing cultures that treat high-robusta coffee as bitter; not an experiment |
| Uncommon Grounds (1999) | Read as context (not an experiment) | A general coffee history describing how Europeans used dark roasts to hide robustas considered bitter; not an experiment |
| Unlocking the Quality Potential of Liberoid Coffee: Advances in Composition Processing and Microbial Fermentation (2026) PubMed | Excluded | A review (no new experiment), about liberica-type coffee |
| Influence of pre-and post-harvest factors on the organoleptic and physicochemical quality of coffee: a short review (2022) PubMed | Excluded | A review (no new experiment) |
| Metabolomics-Based Approach for Coffee Beverage Improvement in the Context of Processing, Brewing Methods, and Quality Attributes (2022) PubMed | Excluded | A review (no new experiment) |
| Effects of active compounds and their metabolites associated with coffee consumption on neurodegenerative disease (2025) PubMed | Excluded | A review on health; no bitterness comparison |
| Improvement of Robusta coffee aroma with l-leucine powder (2023) PubMed | Excluded | Measured volatile compounds labelled as bitter aromas with instruments; no human rating of bitterness |
| Roasting Kinetics and Chemical Composition Changes of Robusta Coffee Beans During Hot Air and Superheated Steam Roasting (2019) PubMed | Excluded | A roasting study of robusta only; bitterness appears only as a general statement in the introduction |
| Genetic mapping of a caffeoyl-coenzyme A 3-O-methyltransferase gene in coffee trees (2003) PubMed | Excluded | A genetics study; no bitterness comparison |
Read in the encyclopedia
- Arabica / Robusta — the two species in this question
- Caffeine — a bitter compound higher in robusta
- Trigonelline — an alkaloid higher in arabica green beans
- Mozambioside — a bitter compound found almost only in arabica green beans
- Chlorogenic acid — turns into bitter lactones during roasting
- Coffee bitterness — the topic hub covering roasting, caffeine and brewing
Source ledger
Each study's source can be reached from its section above. All of the lab's sources are listed in thesource ledger.
How to cite this page
Aoki, M. (NIGAMI LAB) (2026). Arabica or robusta: which is more bitter?. https://nigamilab.com/en/topics/arabica-vs-robusta/
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