See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/339110495

Sensory and Monosaccharide Analysis of Drip Brew Coffee Fractions versus

Brewing Time

Article  in  Journal of the Science of Food and Agriculture · February 2020

DOI: 10.1002/jsfa.10323

CITATIONS READS

10 2,661

5 authors, including:

Mackenzie Batali Carlito Lebrilla

University of California, Davis University of California, Davis
7 PUBLICATIONS   30 CITATIONS    618 PUBLICATIONS   26,973 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

The effects of diet on the functional biology of HDL particles View project

All content following this page was uploaded by Mackenzie Batali on 27 February 2020.

The user has requested enhancement of the downloaded file.

Research Article
Received: 29 June 2019 Revised: 3 December 2019 Accepted article published: 7 February 2020 Published online in Wiley Online Library:

(wileyonlinelibrary.com) DOI 10.1002/jsfa.10323

Sensory and monosaccharide analysis of drip

brew coffee fractions versus brewing time
Mackenzie E Batali,a Scott C Frost,a Carlito B Lebrilla,b
William D Ristenpartc* and Jean-Xavier Guinarda

Abstract
BACKGROUND: The composition of drip brew coffee versus brewing time has been chemically characterized in previous studies,
and it is known that the total dissolved solids (TDS) systematically decreases with each fraction during the brew. Little informa-
tion exists regarding the corresponding sensory attributes versus time, however, and it is unclear how TDS correlates with fla-
vor profile.
RESULTS: Standard drip brews were fractionated into distinct samples by switching in an empty carafe every 30 s during the
brew. Using a trained sensory descriptive panel, we found that most taste and flavor attributes decreased with brew time;
for example, the earlier fractions were systematically more bitter and more sour than later fractions. Surprisingly, however, sev-
eral flavor and taste attributes increased in time; for example, later fractions were systematically sweeter and more floral than
earlier fractions. Since later fractions had lower TDS, these results indicate that perceived sweetness in drip brew coffee is neg-
atively correlated with TDS. Mass spectrometry measurements of the monosaccharide content in the brews showed that none
of the fractions had perceptible concentrations of any monosaccharide.
CONCLUSION: The results of the sensory analysis and the monosaccharide analysis suggest that perceptible sweetness in coffee
is a consequence of masking effects and/or the presence of sweet-associated aromas and flavors. The results further suggest
that unique flavor profiles could be obtained from the same coffee grounds by judicious combinations of specific fractions.
© 2020 Society of Chemical Industry
Supporting information may be found in the online version of this article.

Keywords: Coffea arabica; descriptive analysis; mass spectrometry; monosaccharides

INTRODUCTION
Coffee flavor and taste are the result of a complex relationship of
volatile aroma compounds and nonvolatile carbohydrates, acids,
melanoidins, antioxidants, and lipids.1–3 These components are
influenced by factors ranging from the genetics of the plant all
the way to brewing,4 with over 100 important sensory attributes
identified in the official Coffee Taster's Flavor Wheel.5
The coffee industry has generally conducted drip brewing fol-
lowing the guidelines of the Coffee Brewing Control Chart and
associated Coffee Brewing Handbook, which relates total dissolved
solids (TDS) to percentage extraction (PE) as an index of the sen-
sory quality of coffee.6,7 This approach, which was originally
developed by Lockhart in the 1960s, has not been rigorously
updated as understanding of coffee composition and flavor has
expanded. A detailed picture into the sensory and chemical pro-
files of coffee based on extraction parameters will give brewers
a clearer understanding of how to extract exactly the flavors they
want. For example, anyone who has watched a drip brew in pro-
gress will agree that the extraction and resulting sensory and
chemical profiles change as a function of time; it is readily
observed that the first brewed coffee that comes out is stronger
than the coffee that comes out later. Quantitative measures of this
behavior and corresponding impact on the sensory qualities,
however, are lacking. Past studies have looked at chemical
extraction over time, showing that a brew time of 10–11 min will
extract 90% of the material in coffee,8 and that the majority of
antioxidants (which can contribute to sourness and bitterness)
are extracted in the first 75–150 s of drip brew.9 Other studies
have characterized how aroma compounds are extracted relating
to polarity and volatility, and have demonstrated that more polar
(and therefore more water-soluble) compounds are extracted
early.10 Furthermore, sweetness is a desirable sensory property
for coffee consumers,11,12 and coffee is known to have sucrose,
oligosaccharides, and polysaccharides in green, roasted, and
brewed coffee in varying quantities.13–15
Despite the prominence of TDS as a key metric for brewed cof-
fee, few published studies have looked directly at the relationship
* Correspondence to: WD Ristenpart, Department of Chemical Engineering,
University of California, Davis, One Shields Avenue, Davis, CA 95616, USA,
E-mail: wdristenpart@ucdavis.edu
a Department of Food Science and Technology, University of California, Davis,
Davis, CA, USA
b Department of Chemistry, University of California, Davis, CA, USA
c Department of Chemical Engineering, University of California, Davis, Davis,
CA, USA
1

J Sci Food Agric 2020 www.soci.org © 2020 Society of Chemical Industry

 www.soci.org
between TDS and sensory properties. Although the TDS of the
coffee exiting the brew basket clearly decreases over the course
of the brew, it is unclear how the sensory properties of the brew,
such as sweetness, sourness, and bitterness, and the correspond-
ing TDS vary with time.
The objective of this study was to measure the sensory and
physical properties of a drip brew from a commercial coffee
machine, where the carafe was changed every 30 s to collect ‘frac-
tions’ of the brewed coffee. The individual fractions were evalu-
ated by trained descriptive sensory panelists16 for intensities of
22 attributes. The TDS was measured for each fraction; likewise,
to investigate the chemical basis of sweetness, the total and free
monosaccharide compositions of each fraction were measured
by mass spectrometry. The experiments were designed to yield
insight into the correlations between the sensory and physical
attributes of brewed coffee.
MATERIALS AND METHODS
Coffee preparation
Excelso European Prep grade Coffea arabica from the Huila region
of Colombia, donated by Java City Coffee Roasters in Sacramento,
CA, was roasted to an Agtron score of 54. Whole beans were
stored in 450 g quantities, vacuum sealed in a −20 °C freezer until
the day of evaluation, when they were brought to room tempera-
ture. Thawed coffee beans were ground to setting 5 (‘percolator’)
on a Mahlkönig Guatemala grinder, yielding a particle size of
1200 ± 44 μm.17 Commercial drip brewers (Wilbur Curtis Co. G4
Gold Cup) were used to brew the samples. Each set of fractions
was brewed with 210 g of ground coffee and a 3.8 L dose of water
at 91 °C dispensed continuously for 3 min 20 s, followed by a
2 min ‘drip out’ where the brewed coffee continued to drain by
gravity. Stainless steel 1 L thermal carafes were used to collect
the respective fractions (1 through 8) in 30 s time increments,
with the timer starting upon observation of the first drip. The mass
of coffee collected in each fraction was measured to determine
the average flow rate for that fraction; note that earlier fractions
had lower masses because of water absorption into the initially
dry grounds.18 A second full brew under equivalent conditions
was also performed without fractionating to yield a ‘whole brew’
sample. One full set of nine samples (eight fractions and one full
brew) was brewed per day, with panelists returning three times
on different days to complete three tasting replicates. Serving
began 30 min after brewing.
Panel recruitment and training
Panelists were recruited from the UC Davis community, giving
preference to those who had previously served on descriptive
analysis panels and who were regular coffee drinkers. Twelve pan-
elists between 20 and 35 years old were selected and underwent
5 weeks of training twice per week.
Panelists used the Coffee Taster's Flavor Wheel for reference
and were given the opportunity for term generation across the
training set of coffees.5 The most commonly generated terms
were put on an unlabeled line scale, and panelists ranked four dif-
ferent coffee samples on each attribute. Further training was per-
formed using flavor/taste references spiked directly into the
evaluation sample coffee, as well as Folgers instant coffee, to com-
pare attribute intensity (Table 1). Panelists also underwent an
assessment session in the booths, with identical serving condi-
tions as the final evaluation, after which all judges received
2
wileyonlinelibrary.com/jsfa © 2020 Society of Chemical Industry
ME Batali et al.
feedback on their performance (i.e. ability to discriminate, repro-
ducibility, and concept alignment).
Descriptive analysis testing procedure
To compare samples, a generic descriptive analysis that com-
bined elements of the quantitative descriptive analysis and spec-
trum methods16 was performed using RedJade sensory science
software (RedJade Sensory Solutions LLC, Redwood City, CA,
USA). Sixteen flavors, one mouthfeel, and the five basic tastes
were retained for evaluation (Table 1). The individual fractions
were served in a blind randomized Williams Latin square block
design, randomized across panelists and replicates to minimize
the effects of sensory fatigue on the results, with codes generated
from RedJade to prevent sample recognition. Participants were
seated in isolated, temperature-controlled, and red-lit sensory
booths. Three to five participants were served per session, each
receiving 20–30 g of each coffee served between 60 and 65 °C,
in white ceramic mugs. Panelists recorded attributes via an iPad
logged in to RedJade, with a 15 cm unstructured line scale for
each specific sensory attribute from ‘low’ to ‘high’. Participants
were served one sample at a time and evaluated all nine samples
(eight fractions and one ‘whole brew’) in a 1 h session, with a
1–2 min break to rinse their mouths with water and unsalted sal-
tine crackers between samples. Three trial replicates were tasted
on three separate tasting days.
Physical measurements of coffee samples
Small aliquots of coffee from each fraction were cooled to room
temperature in sealed containers, and the TDS was measured with
a VST LAB coffee refractometer (Cambridge, MA, USA) using one
drop of the cooled coffee. Independent calibration tests of the
mass of dissolved solids remaining after drying indicated the
refractometer yielded excellent measurements of the TDS (see
Supporting Information Fig. S1). Measurements of the TDS were
taken for every fraction and whole brew during every replicate.
Monosaccharide analysis
Preparation for monosaccharide analysis
Samples of the brewed coffee were saved in 10 mL cryovials in a
−20 °C freezer prior to chemical analysis. Methods presented here
for workup were based on work by Xu et al. then refined for mono-
saccharide composition of polysaccharides by Amicucci et al.19,20
For total monosaccharide analysis, frozen brewed coffee samples
(fractions 1–9 from Table 2, along with the whole brew sample)
were vortexed and aliquoted in a 96-well plate for hydrolysis.
Dry and spent grounds were put into a 10 mg mL−1 solution with
water and blended in a Storm 24 Bullet Blender made by Next
Advance (Averill Park, NY, USA) with 1.4 mm stainless steel beads,
heated at 100 °C for 1 h, and then blended again and aliquoted
into a 96-well plate for hydrolysis. All of these samples were
performed in triplicate. Standard-curve solutions containing
14 monosaccharides (fructose, allose, mannose, glucose, galac-
tose, xylose, arabinose, ribose, fucose, rhamnose, glucuronic
acid, galacturonic acid, N-acetylglucosamine, and N-acetyl-galac-
tosamine, purchased from Sigma Aldrich, St. Louis, MO, USA or
Fisher Scientific, Waltham, MA, USA) were prepared at concentra-
tions ranging from 1 ng mL−1 to 100 μg mL−1. The food and poly-
saccharide standard solutions underwent acid hydrolysis with
trifluoroacetic acid (TFA). Upon completion of hydrolysis, samples
were quenched with cold water.
Aliquots of TFA-hydrolyzed coffee samples (for total monosaccha-
ride analysis), untreated coffee samples (for free monosaccharide
J Sci Food Agric 2020
Analysis of drip brew coffee fractions versus time www.soci.org

Table 1. Consensus attribute descriptions and preparation method for descriptive analysis standards

Attribute Description Aroma reference in training Flavor/taste reference in training

Bitter Basic taste of bitter 0.0085 g/0.017 g/0.029 g Prolab

caffeine pills crushed per 1 g Folgers
instant coffeea
Sour Basic taste of sour 0.013 g/0.025 g/0.043 g citric acid per
1 g Folgers instant coffee
Salty Basic taste of salty 0.5 tbsp table salt in sample coffee
Sweet Basic taste of sweet 0.5 tbsp baking sugar in sample coffee,
0.85 g/1 g/1.15 g baking sugar per
1 g Folgers instant coffee
Umami Basic taste of savory 1% monosodium glutamate solution in
water
Smoky/Burnt Liquid smoke, charcoal, ash 5 drops liquid smoke dried under a
vent in a mug
Tea/Floral Black tea/tannic, chamomile, 1 pack each Lipton black tea and
dried floral Celestial Seasonings chamomile
tea dry
Caramelized Brown sugar, maple syrup, 1 tbsp brown sugar drizzled with 0.5 tbsp brown sugar in sample coffee
molasses maple syrup
Cereal Dried cereal, grain, oat, 50/50 grape nuts and Chex
woody, musty
Fruity Apple, peach, pear (fruit and 2 tbsp unsweetened applesauce 1 tbsp pear syrup from fruit cups in
juice) sample coffee
Meaty/Brothy Beef broth 2 tbsp Campbell's beef broth
Rubber Rubber band Rubber band
Citrus Lemon juice, fresh lemon 2 tbsp of lemon juice 1 tbsp lemon juice in sample coffee
Nutty Almond, hazelnut, peanut Equal parts crushed almond, 5 drops almond extract in sample coffee
hazelnut, and peanut
Berry Blackberry, blueberry, Smucker's blackberry jam and 1 tbsp Smucker's blackberry jam in
cranberry frozen cranberries, microwaved sample coffee
30 s
Dried fruit Prune/prune juice Dried prunes with 1 tbsp prune 1 tbsp prune juice in sample coffee
juice, microwaved 30 s
Green/Vegetative Cooked spinach, olive oil, 2 tbsp canned spinach, drizzled with 1 tbsp olive oil in sample coffee
herbs olive oil
Alcoholic/Winey Sour, fermented, winey, 2 tbsp red wine
ethanol
Vinegar White vinegar, pungent, 2 tbsp white vinegar 1 tbsp white vinegar in sample coffee
acidic
Honey Honey, floral sweetness 1 tbsp honey 1 tbsp honey in sample coffee
Cocoa Cocoa powder, dark 1 tbsp Hershey's cocoa powder 1 tbsp Hershey's cocoa powder in
chocolate sample coffee
Astringent Mouth drying, puckering 0.07% solution of alum in water, or
0.026 g/0.043 g/0.06 g alum per 1 g
Folgers instant coffee
a
Coffee in training was brewed with 85 g water per 1 g Folgers instant coffee.

analysis), and monosaccharide standard curves were added directly
to polypropylene 96-well plates containing ammonium hydroxide
(NH4OH) and 1-phenyl-3-methyl-pyrazolone (PMP) in methanol
and incubated at 70 °C. Samples were then dried to completion
by vacuum centrifugation and reconstituted in water. Two
chloroform–water extractions were performed to remove excess
PMP. The aqueous layer was collected and stored at −20 °C until
the time of analysis.
Analysis was performed using the following procedures pro-
vided by Amicucci et al.19 Samples were analyzed on an Agilent
6495 triple quadrupole mass spectrometer coupled to an Agilent
1290 Infinity II ultrahigh-performance liquid chromatograph. Sepa-
ration was performed on an Agilent Zorbax C18 column
(2.1 mm × 100 mm i.d., 1.8 μm particle size, Santa Clara, CA, USA).
Chromatographic solvents consisted of 25 mmol L−1 ammonium
acetate adjusted to pH 8.2 using NH4OH in 5% acetonitrile in water
(v/v; solvent A) and 95% acetonitrile in water (v/v; solvent B). The
mass spectrometer was operated in positive ion mode, and
dynamic multiple reaction monitoring mode was used for data
acquisition. Separation of PMP-labeled monosaccharides was
3

J Sci Food Agric 2020 © 2020 Society of Chemical Industry wileyonlinelibrary.com/jsfa

 www.soci.org ME Batali et al.

Table 2. Mean attribute intensities for each factor, with Fischer's LSD groups indicated by letters

Attribute Fraction 1 Fraction 2 Fraction 3 Fraction 4 Fraction 5 Fraction 6 Fraction 7 Fraction 8 Whole brew LSD

Smoky/Burnt 47.58a 40.94ab 37.97b 27.36c 21.89cde 19.94de 17.03ef 10.92f 24.83cd 6.91
Tea/Floral 9.36e 7.83e 13.5e 20.92d 29.69bc 34.39b 41.83a 41.97a 23.42cd 7.34
Caramelized 21.03b 21.75ab 27.81a 17.61b 19.69b 17.22b 18.92b 20.19b 19.61b 6.28
Cocoa 24.47a 24.75a 22.44ab 15.25c 13.92cd 16.56bc 8.42d 11.47cd 16.47bc 6.14
Cereal 11.42d 13.5bcd 13.39cd 17.94abc 17.86abc 17.5abcd 19.83ab 22.03a 12.14cd 6.42
Fruity 9.81d 12.22cd 14.47bcd 15.17bcd 18.22abc 18.92ab 23.39a 12.08cd 20.19ab 6.58
Meaty/Brothy 22.11a 17.89ab 13.92bc 9.78cd 7.92d 6.81d 5.53d 6.17d 9.22cd 5.15
Rubber 24.42a 17.31b 15.36bc 10.03de 6.08e 8.39de 5.75e 7.08e 11.81cd 4.52
Citrus 19.31a 20.78a 19.58a 12.97bc 15.31ab 15.44ab 8.61c 9.75bc 20.19ab 5.71
Nutty 15.89ab 18.61ab 21.56a 19.83a 12.89b 16.44ab 17.67ab 16.92ab 17.75ab 6.16
Berry 14.11abc 13.56abc 19.03a 17.72a 15.53ab 8.36c 9.25bc 15.22ab 18a 6.78
Dried/Fruit 27.69a 23.61ab 26.67a 19.42bc 14.14cd 9.78d 13.61cd 16.08cd 23.42ab 6.67
Green/Vegetative 19.36a 18.14ab 18.31ab 16.97ab 13.72ab 16.03ab 13.03b 18.17ab 16.14ab 5.82
Alcoholic/Winey 43.47a 36.28b 25.36c 13.06d 10.64de 6.78e 5.14e 6.56e 15.25d 6.27
Vinegar 26.33a 22.19a 10.94b 8.25bcd 7.83bcd 3.69d 4.5d 5.58cd 10.25bc 5.14
Honey 6.97b 9.89b 8.89b 11.31b 18.31a 19.89a 18.64a 22.25a 9.08b 4.88
Astringent 42.67a 41.19ab 35.08bc 28.19d 29.28cd 28.39cd 20.61e 24.89de 31.03cd 6.88
Bitter 56.56a 47.89b 41.03c 31.53de 26.19ef 20.81fg 19.75g 21.06fg 35.53d 5.41
Sour 55.89a 52.03a 38.69b 27.64c 23.53cd 18.89de 15.92ef 11.47f 37.44b 5.38
Sweet 9.92e 13.64de 17.53cd 19.31c 24.86b 27.39b 34.11a 29.97ab 18cd 5.38
Salty 12.14a 10.97ab 6.42bc 6.39c 6.22c 5.25c 4.36c 5.92c 7.53bc 3.50
Umami 20.69a 18.06a 10.42bc 8.81bcd 7cd 6.92cd 5.33d 4.67d 11.5b 4.31

performed using the parameters developed by Xu et al., with an
optimized 10 min binary gradient of 12–15% B at 0–7 min, 99% B
at 7.1–8.5 min, and 12% B at 8.6–10 min.20 All measurements used
a flow rate of 0.5 mL min−1 and a column temperature of 35 °C.
Data analysis was performed on the Agilent Mass Hunter B.08 quan-
titative analysis software and peaks were automatically integrated.
To account for the addition of water during hydrolysis, the concen-
tration of each monosaccharide was multiplied by a correction fac-
tor derived from the ratio of the monosaccharide residue molar
mass to the monosaccharide molar mass.
of judge effect.21 One attribute (Cereal) was dropped. All factors
determined to be significant by ANOVA were then analyzed using
Fisher's least significant difference (LSD) through the Agricolae
package. Total and free monosaccharide concentrations were
analyzed by a two-way ANOVA, with Fisher's LSD run through
the Agricolae package to determine significant differences
among samples. Principal component analysis22 biplots were pro-
duced in R using the FactoMineR package, the factoextra package,
and the ggplot2 package to select for significant attributes and
visually represent the outcomes. Pearson's correlations were cal-

Free monosaccharide concentration×Monosaccharide molar mass g mol−1 −18 g mol−1
Bound monosaccharide concentration =

Monosaccharide molar mass g mol−1

Free monosaccharides were analyzed without hydrolysis; so, for
those, the correction factor was not used.
DATA ANALYSIS
Descriptive analysis results were exported from RedJade, convert-
ing positions on the 15-point line scale into scores on a scale from
0-100 for each attribute. R version 3.5.3 ‘Great Truth’ (R Core Team,
2019) was used to perform a three-factor analysis of variance
(ANOVA) with two-way interactions across panelists, samples,
and replicates to determine which attributes were significant, cal-
culated at ⊍ = 0.05. F values were adjusted from the original
ANOVA by performing a pseudomixed ANOVA to calculate new
F ratios and determine which attributes were significant in spite
4
culated in R using the ‘cor.test’ function both between attribute
and TDS with significance determined at ⊍ = 0.05. The attribute
correlation table was plotted with the Corrplot package with sig-
nificance determined at ⊍ = 0.05.
RESULTS
Physical measures of coffee
The measured TDS and dispensed coffee mass for each individual
fraction, as well as the full brew, are shown in Fig. 1. The TDS of the
first fraction was highest, at about 3%, and then decreased sys-
tematically to about 0.3% by fraction 8 (Fig. 1(A)). The standard
deviation for the TDS in the early fractions was larger, approxi-
mately ±0.5%, compared with the more consistent later fractions

wileyonlinelibrary.com/jsfa © 2020 Society of Chemical Industry J Sci Food Agric 2020

Analysis of drip brew coffee fractions versus time
Figure 1. Boxplots of (A) total dissolved solids (TDS) percentage by fraction and (B) coffee mass by fraction (n = 9). Fraction 1: 0–30 s; fraction 2: 31–60 s;
fraction 3: 61–90 s; fraction 4: 91–120 s; fraction 5: 121–150 s; fraction 6: 151–180 s; fraction 7: 181–210 s; fraction 8: 211 –240 s; fraction 9 (only for phys-
ical and chemical evaluations, 241–300 s); and whole brew (WB): 0–300 s.
with a standard deviation of approximately ±0.03%. The TDS
increased slightly in the final ‘drip out’ phase collected in
fraction 9, the last minute of the brew. The whole brew (unfractio-
nated) sample had a TDS of 0.84 ± 0.05%.
The corresponding mass flow rates steadily increased from
fraction 1, with an average of 141 g per 30 s, to a peak at
fraction 6, with an average of 543 g per 30 s. The slight increase
in TDS observed after fraction 8 corresponded with the end of
water dispensation after 3 min, as the 2:20 drip out had a gradual
decrease in flow rate down to 100 g per 70 s in the last fraction of
the brew.
Descriptive analysis
The ANOVA showed significant sensory differences among coffee
fractions for 19 of the 22 attributes (bold F values in Supporting Infor-
mation Table S1). Eighteen of those attributes also showed a signifi-
cant judge × sample interaction (bold F values in Supporting
Information Table S1), and therefore a pseudomixed ANOVA was
performed to determine which attributes were significant in spite
of the judge effect. This process yielded 17 attributes significantly dif-
ferent across all samples. These attributes were Sweet, Sour, Bitter,
Salty, and Umami taste, Astringent mouthfeel, Tea/Floral, Honey, Fru-
ity, Citrus, Smoky/Burnt, Rubber, Alcoholic/Winey, Meaty/Brothy, Vin-
egar, Cocoa, and Dried Fruit flavors. Of these attributes, 13 were
highest in fraction 1, and then systematically decreased as the brew
progressed. In contrast, four attributes (Sweet, Tea/Floral, Honey, and
Fruity) were lowest in fraction 1 and increased significantly as the
brew progressed. The mean intensities, LSD values, and groupings
for all attributes and all samples are shown in Table 2.
A representative subset of attributes versus fraction number is
shown in Fig. 2. We highlight these attributes here because they
had large changes in intensity during the brew; data for all mea-
sured attributes is available in Table 2, with additional statistical
information in the supplementary material. Four attributes that
decreased in intensity over the course of the brew were Bitter
and Sour taste, Astringent mouthfeel, and Smoky/Burnt flavor
(Fig. 2(A)–(D) respectively). These four sensory attributes
decreased significantly and systematically versus fraction number.
Figure 2(E)–(H) respectively show the four attributes that
J Sci Food Agric 2020 © 2020 Society of Chemical Industry
www.soci.org
increased in intensity over the course of the brew: Sweet taste,
Tea/Floral, Honey, and Fruity flavor. These four attributes
increased and then plateaued in intensity at the end of the brew
except in the case of Fruity flavor, which exhibited a slight
decrease between fraction 7 and fraction 8.
A principal component analysis of all statistically significant
attributes gives an overall view of how the samples trend across
the attributes. The first two components, plotted in Fig. 3, account
for 95.8% of the variance in the products, with the coffees separat-
ing primarily across the first principal component, which
accounted for 93.1% of the variance. In dimension 1 we see Frac-
tion 1 on the positive side associated with all the attributes with all
attributes seen to decrease as the brew progresses. The fractions
move sequentially toward the negative side, to Fraction 8 charac-
terized most by the attributes shown to increase as the brew pro-
gresses, particularly Sweet, Tea/Floral, and Honey. Importantly,
dimension 2 only accounts for 2.7% of the variance, indicating
that the linear spread in order of fractions on Dimension 1 is the
only relevant information to describe these products.
Since the different fractions have highly different TDS values,
with the TDS decreasing steadily from the first to the last fractions
(see Fig. 1), a natural question is how the flavor intensities corre-
late with TDS. To our knowledge, this relationship has not been
explored in existing coffee literature. Figure 4 shows the attribute
intensity correlations to TDS for three representative sample attri-
butes. Bitter, positively correlated in Fig. 4(A), decreases as TDS
decreases. Sweet, negatively correlated in Fig. 4(B), increases as
TDS decreases. Berry, not correlated in Fig. 4(C), is present at rela-
tively consistent levels in all samples. Figure 4(D) shows the corre-
lation coefficients of all samples. As shown in Fig. 4(D), in order of
decreasing correlation coefficient, Vinegar, Meaty/Brothy, Alco-
holic/Winey, Umami, Salty, Rubber, Bitter, Sour, Astringent,
Smoky/Burnt, Cocoa, Dried Fruit, and Citrus attributes were all
positively correlated to TDS. In order of increasing correlation
coefficient, Sweet, Tea/Floral, Honey and Fruity attributes were
all negatively correlated to TDS. In total, 13 attributes were posi-
tively correlated to TDS, four attributes were negatively corre-
lated, and the remaining five attributes showed no statistically
significant relationship with TDS.
5
wileyonlinelibrary.com/jsfa
 www.soci.org ME Batali et al.

Figure 2. Attribute intensity by fraction for eight of the 22 attributes, with error bars and letter codes indicating significant difference based on LSD
values (Supporting Information Table S1): (A) Bitter taste; (B) Sour taste; (C) Astringent mouthfeel; (D) Smoky/Burnt flavor; (E) Sweet taste; (F) Tea/Floral
flavor; (G) Honey flavor; (H) Fruity flavor.

Figure 3. Principal component biplot showing evaluated samples (fractions 1–8 and the ‘whole brew’) and significant descriptive attributes as deter-
mined by the pseudomixed ANOVA.

Monosaccharide analysis
After an increase in Sweet taste was measured by the descriptive
panel, the chemical analysis on total monosaccharide and free
monosaccharide content was carried out to investigate the
hypothesis that sugar would be responsible for the increase in
sweetness.
ANOVA of all samples showed significant differences between
total monosaccharide composition of samples for all monosaccha-
rides, aside from allose and N-acetyl-galactosamine, and significant
6
differences between samples for all free monosaccharides (bold
values shown in F table of Supporting Information Table S2). Mean
intensities, LSD values, and groups for all measured monosaccha-
rides and all samples are shown in Table 3.
Figure 5 shows the change in total and free monosaccharides
over the course of the brew, omitting those that were measured
in less than 1.5% abundance for all samples, color coded by chem-
ical component showing relative composition. The graphs indi-
cate the highest concentration of monosaccharides always

wileyonlinelibrary.com/jsfa © 2020 Society of Chemical Industry J Sci Food Agric 2020

Analysis of drip brew coffee fractions versus time
Figure 4. Correlations for (A) Sweet/TDS, (B) Bitter/TDS, and (C) Berry/TDS. (D) R values for all sensory attributes to TDS showing which attributes are pos-
itively and negatively correlated with percentage TDS.
occurred in the early fractions. Specifically, the total concentra-
tions were 2.6 mg mL−1 for total monosaccharide and
0.016 mg mL−1 for free monosaccharide in fraction 1. Both the
total and free monosaccharides then exhibited an overall but
non-monotonic decrease, with the lowest concentration of mono-
saccharides in fraction 6 (0.23 mg mL−1 for total monosaccha-
rides and 0.0018 mg mL−1 for free monosaccharides). A
subsequent statistically significant but relatively slight uptick in
overall monosaccharide concentration occurred in fractions 7, 8,
and 9 near the end of the brew. A statistically significant but slight
increase also occurred between fractions 3 and 4, followed by a
decreased from fractions 4 to 5. Table 3 shows the concentrations
of all measured total and free monosaccharides, with LSD group-
ings by letter and standard deviation based on three analytical
replicates.
DISCUSSION
The most surprising result presented here is that sweetness and
several other sensory attributes increased in later fractions; that
is, these attributes were negatively correlated with TDS. A simple
explanation would be that the later fractions actually had higher
sugar content. However, our chemical measurements of mono-
saccharide content do not support this hypothesis, with the
monosaccharide concentrations following the same general pat-
tern as TDS, decreasing as the coffee dilutes in the later fractions.
Additionally, when comparing monosaccharide concentrations
with recognition and detection thresholds found in literature, glu-
cose and fructose have an average sensory detection threshold in
the range of 3.1 mg mL−1 and 1.7 mg mL−1 respectively, and an
J Sci Food Agric 2020 © 2020 Society of Chemical Industry
www.soci.org
average recognition threshold of 6.4 mg mL−1 and 3.5 mg mL−1
respectively.23 The aforementioned literature does report an
incredibly wide range of detection and recognition thresholds
based on the method used and the population detected, but even
assuming trained sensory panelists are better than average for
sweet detection, the minimum reported detection threshold
(0.18 mg mL−1 for glucose and 0.108 mg mL−1 for fructose) found
for the most sensitive tasters is still significantly higher (by about
one order of magnitude) than the highest free monosaccharide
concentration measured here at 0.015 mg mL−1 (see Fig. 5(A), first
column). Additionally, monosaccharides aside from fructose and
glucose (making up the majority of both the total and free mono-
saccharides) are even less sweet.24 Nor do all polysaccharides
measured in the total monosaccharide composition contribute
to perceptible sweetness. There may be some degree of enhance-
ment of sweet taste compared with literature detection and rec-
ognition thresholds based on temperature, since some studies
have found that higher temperature increases perceived sweet-
ness intensity.25 Even so, the combination of decreasing concen-
trations in later fractions and their overall low magnitudes
indicate that the physiological perception of sugar is not the pri-
mary contribution to the perceived sweetness in any of these cof-
fee samples.
If the monosaccharides are not responsible for the perceived
sweetness, the natural question is what does drive the perceived
sweetness? There are several potential hypotheses that can be
put forth to explain the trends observed in the sensory analysis.
It is possible that there is some degree of masking of attributes
from the sour‑ and bitter-tasting compounds that overpower
some of the other more delicate flavors, which then appear to
7
wileyonlinelibrary.com/jsfa
 8

Table 3. Mean and standard deviation of all measured monosaccharide concentrations with Fisher's LSD value and letter groups indicating significant differences (n = 3). All quantities reported in

wileyonlinelibrary.com/jsfa
mg L−1

Monosaccharide Fraction 1 Fraction 2 Fraction 3 Fraction 4 Fraction 5 Fraction 6 Fraction 7 Fraction 8 Fraction 9 Whole brew LSD

Total monosaccharides
Fructose 150 ± 6.4ab 70 ± 28bc 70 ± 49bc 190 ± 120a 54 ± 32bc 39 ± 12c 150 ± 89ab 59 ± 31bc 120 ± 77abc 46 ± 2.6c 100
Mannose 700 ± 87a 340 ± 12b 150 ± 31c 150 ± 20c 75 ± 19de 34 ± 5.3e 29 ± 06.8e 46 ± 16de 99 ± 28cd 76 ± 6.3de 57
Glucuronic acid 6.6 ± 1.5a 4.1 ± 0.47b 2.3 ± 0.1c 1.6 ± 0.37cd 1.3 ± 0.8cd 0.58 ± 0.15d 2.2 ± 1.6c 0.78 ± 0.39d 1.6 ± 0.56cd 1.4 ± 0.18cd 1.3
Ribose 8.9 ± 0.83a 6.3 ± 0.52b 3.3 ± 0.09 c 2.9 ± 0.28cd 1.7 ± 0.32efg 1.2 ± 0.18g 1.9 ± 0.6defg 1.7 ± 0.73fg 2.8 ± 1.2cde 2.3 ± 0.16cdef 1.1
Allose 0.026 ± 0.24ab 0.75 ± 0.92a 0.38 ± 0.18ab 0.25 ± 0.21ab 0.2 ± 0.18ab 0.35 ± 0.3ab 0.25 ± 0.36ab 0.2 ± 0.17ab 0.37 ± 0.49ab 0b 0.6
Rhannose 170 ± 24a 96 ± 6b 34 ± 5.9c 21 ± 0.32cd 11 ± 1.6de 6.5 ± 0.9e 5.2 ± 0.95e 5.2 ± 0.97e 13 ± 1.5de 22 ± 1.6cd 14
Galacturonic acid 5.0 ± 0.52a 2.3 ± 0.23b 1.0 ± 0.024cd 1.4 ± 0.75c 0.64 ± 0.43de 0.33 ± 0.086e 0.68 ± 0.23de 0.45 ± 0.27de 0.87 ± 0.46cde 0.68 ± 0.17de 0.67
Glucose 170 ± 22a 100 ± 32bc 70 ± 24abc 150 ± 88bc 70 ± 63bc 27 ± 4.8c 97 ± 42abc 46 ± 22c 140 ± 67ab 40 ± 7c 0.56
N-Acetylglucosamine 1.6 ± 0.32a 1.4 ± 0.44ab 1.1 ± 0.25abcd 1.0 ± 0.26bcd 0.90 ± 0.31bcd 0.74 ± 0.41cd 0.58 ± 0.12d 0.74 ± 0.066cd 1.2 ± 0.52abc 0.91 ± 0.15bcd 0.56
N-Acetylgalactosamine 0.16 ± 0.071ab 0.12 ± 0.053ab 0.055 ± 0.071b 0.17 ± 0.061ab 0.25 ± 0.15a 0.18 ± 0.033ab 0.23 ± 0.068a 0.15 ± 0.088ab 0.19 ± 0.12a 0.18 ± 0.074ab 0.14
Galactose 700 ± 87a 390 ± 12b 170 ± 32de 240 ± 19cd 95 ± 19e 56 ± 3.3e 160 ± 62de 350 ± 210bc 140 ± 26de 120 ± 8.2de 130
Xylose 17 ± 3.5a 11 ± 1.4ab 4.1 ± 1.6bcd 110 ± 12abc 4.0 ± 1.2cd 1.8 ± 0.14d 3.0 ± 1.4d 3.4 ± 02.6d 4.2 ± 2.2bcd 3.4 ± 0.48d 67.9
Arabinose 710 ± 110a 480 ± 41b 190 ± 24c 150 ± 21cd 84 ± 6.9de 59 ± 7.8e 62 ± 9.2e 53 ± 13e 140 ± 32cd 140 ± 11cd 71.5
Fucose 1.9
www.soci.org

5.7 ± 1.1a 4.2 ± 0.81ab 2.3 ± 0.43bcd 3.0 ± 1.9bcd 1.9 ± 0.28cd 1.5 ± 0.31d 2.6 ± 0.93bcd 1.9 ± 0.5cd 3.7 ± 2bc 2.1 ± 0.14cd
Free monosaccharides
Fructose 3.8 ± 0.28a 3.3 ± 1.2ab 0.93 ± 0.37def 1.8 ± 0.61cde 0.93 ± 0.45def 0.36 ± 0.14f 0.91 ± 0.035ef 0.99 ± 0.09def 1.9 ± 0.93cd 2.3 ± 0.029bc 1.0
Mannose 0.43 ± 0.015a 0.34 ± 0.063b 0.099 ± 0.0048e 0.18 ± 0.012d 0.071 ± 0.0014ef 0.049 ± 0.0027f 0.1 ± 0.014e 0.09 ± 0.017e 0.16 ± 0.0063d 2.6 ± 0.0032c 0.04
Glucuronic acid 0.78 ± 0.046a 0.63 ± 0.12b 0.18 ± 0.013e 0.33 ± 0.014d 0.13 ± 0.015ef 0.087 ± 0.003f 0.2 ± 0.029e 0.2 ± 0.022e 0.6 ± 0.0028d 0.46 ± 0.032c 0.079

© 2020 Society of Chemical Industry

Ribose 0.43 ± 0.01a 0.33 ± 0.043b 0.094 ± 0.0029e 0.16 ± 0.0064d 0.065 ± 0.0021fg 0.046 ± 0.0036g 0.094 ± 0.0084ef 0.09 ± 0.012e 0.14 ± 0.015d 0.25 ± 0.00093c 0.028
Allose 0.057 ± 0.0056a 0.041 ± 0.004b 0.015 ± 0.0023e 0.022 ± 0.0025d 0.012 ± 0.0018ef 0.01 ± 0.0015f 0.015 ± 0.0034e 0.013 ± 0.0022ef 0.02 ± 0.0031d 0.035 ± 0.0043c 0.0045
Rhannose 0.61 ± 0.029a 0.44 ± 0.1b 0.11 ± 0.012de 0.13 ± 0.0078d 0.044 ± 0.0025f 0.032 ± 0.0013f 0.059 ± 0.0067ef 0.051 ± 0.0083f 0.087 ± 0.0076def 2.4 ± 0.0045c 0.057
Galacturonic acid 0.19 ± 0.0043a 0.13 ± 0.023b 0.027 ± 0.0014ef 0.042 ± 0.0025d 0.016 ± 0.0031fg 0.0098 ± 0.0012g 0.02 ± 0.000044fg 0.018 ± 0.0017fg 0.036 ± 0.0027de 0.095 ± 0.0068c 0.014
Glucose 0.87 ± 0.012a 0.93 ± 0.49a 0.32 ± 0.14b 0.59 ± 0.21ab 0.39 ± 0.28b 0.2 ± 0.023b 0.4 ± 0.12b 0.29 ± 0.047b 0.61 ± 0.39ab 0.57 ± 0.039ab 0.43
N-Acetylglucosamine 0.03 ± 0.00014a 0.03 ± 0.0096a 0.012 ± 0.0015de 0.02 ± 0.0016b 0.0087 ± 0.00075de 0.0068 ± 0.00099cd 0.014 ± 0.0023cd 0.014 ± 0.0025cd 0.019 ± 0.0018bc 0.024 ± 0.0025ab 0.0062
N-Acetylgalactosamine 0.0047 ± 0.0029ab 0.0025 ± 0.0012cd 0.0014 ± 0.00049d 0.002 ± 0.00048cd 0.0015 ± 0.0013d 0.0017 ± 0.00047d 0.003 ± 0.0011bcd 0.0035 ± 0.00045abc 0.0028 ± 0.00064cd 0.0051 ± 0.00094a 0.0018
Galactose 4.5 ± 0.21a 3.6 ± 0.66b 1.1 ± 0.047ef 2.0 ± 0.14d 0.75 ± 0.03fg 0.55 ± 0.011g 1.2 ± 0.19ef 1.2 ± 0.28e 1.9 ± 0.14de 2.7 ± 0.11c 0.45
Xylose 0.25 ± 0.027a 0.18 ± 0.041b 0.05 ± 0.0022def 0.074 ± 0.018d 0.035 ± 0.0034d 0.021 ± 0.0033f 0.04 ± 0.0035ef 0.04 ± 0.0033ef 0.054 ± 0.012d 0.12 ± 0.0057c 0.031
Arabinose 4.0 ± 0.16a 3.1 ± 0.51b 0.91 ± 0.064ef 1.6 ± 0.11d 0.65 ± 0.015fg 0.42 ± 0.023g 1.0 ± 0.11e 0.97 ± 0.17ef 1.6 ± 0.18d 2.4 ± 0.08c 0.35
Fucose 0.26 ± 0.004a 0.18 ± 0.022b 0.044 ± 0.0043ef 0.063 ± 0.0039d 0.026 ± 0.0018gh 0.016 ± 0.0024h 0.028 ± 0.0016gh 0.031 ± 0.0027fg 0.048 ± 0.0025e 0.11 ± 0.0047c 0.014

J Sci Food Agric 2020

ME Batali et al.
Analysis of drip brew coffee fractions versus time
Figure 5. (A) Total and (B) free monosaccharide concentrations, with monosaccharides in less than 1.5% abundance in all samples omitted. With omitted
measurements, figure shows at least 98% of all total or free monosaccharides in each sample. All mean concentrations can be seen in Table 3.
increase once those masking compounds are flushed out of the
grounds. Caffeine26 and bitter, astringent chlorogenic acids27
are highly soluble and would be in high concentrations in early
fractions. Organic acids contributing to sour taste would be solu-
ble early extractors as well.28 The masking effect might explain the
perceived increase in flavor attributes; however, this observation
is at odds with the observation that sweet-active molecules,
particularly in the later fractions, were below the reported recog-
nition thresholds for sweetness even when sourness and bitter-
ness receded.
Potentially related to the masking effects, another likely factor is
the role of aroma compounds in the differences seen between cof-
fee taste attributes. Prior work has shown that the perception of
sweetness increases in the presence of sweet-associated aromas,
even if no sweet taste stimulation is being perceived.29 Previous
chemical studies have been conducted on the rates of extraction
of various aroma compounds from coffee, with Sánchez López
et al.30 performing online proton transfer reaction time-of-flight
mass spectrometry showing different groupings of when aroma
compounds extracted during the brewing process. The aforemen-
tioned results showed some flavors like guaiacol (sweet, floral,
vanilla), 4-allylguaiacol (spicy carnation floral), ethyl acetoacetate
(fruity), and 2-oxopropylacetate (fruity), extracting in the later
group.30 Further chemical studies have shown that, among other
things, floral/honey flavored ⊎-damascenone is a less-polar, later-
extracting compound.10 Thus, there is some chemical literature
that corroborates the sensory results of increase in Honey, Tea/-
Floral, and Fruity flavors, and these may be sufficiently sweet-
associated aromas. Coupled nosespace analysis with temporal
dominance of sensation in coffee has indicated that an increase
in sweetness by added sugars also increased the dominance of
certain aromas, such as roasted, caramel, and nutty.31 This behav-
ior suggests that the inverse effect might also be operative, where
an increase in those aromas may result in increased perceived
sweetness, as has been reported in non-coffee products.
The ambiguity left by the monosaccharide analysis indicates the
need for further chemical analysis; there is not a complete enough
J Sci Food Agric 2020 © 2020 Society of Chemical Industry
www.soci.org
picture of the extraction profile of nonvolatile and volatile com-
pounds to make claims as to what exactly is causing the increase
in Sweet, Honey, Tea/Floral, and Fruity attributes at the end of the
brew in spite of the overall decrease in TDS. The change in extrac-
tion with flow rate is also an interesting result that opens many
paths for further exploration as to how flow rate impacts what
specific components are extracted from coffee based on differing
amounts of contact time, from physical, chemical, and sensory
perspectives.
Regardless of the precise chemical mechanisms for the
increased sweetness in later fractions, there are immediate practi-
cal implications for the coffee industry. Our results clearly show
pronounced and systematic sensory differences among fractions,
suggesting that very different sensory profiles could be generated
from the same coffee grounds by judicious combination of differ-
ent fractions. For example, the same batch of coffee grounds
could be used to make a batch of bitter and smoky coffee for con-
sumers who prefer to use cream and sugar, and also to make a
batch of more sweet and floral coffee for consumers who prefer
that style.
CONCLUSION
Our results provide the first quantitative indication that some sen-
sory attributes in brewed coffee (Sweet taste, Tea/Floral, Fruity,
and Honey flavors) are negatively correlated with TDS. Chemical
analyses demonstrated that the perceived sweetness is not due
to any monosaccharides extracted during the brew, since both
the total and free monosaccharide concentrations were well
below the detection threshold of any sugars. In other words, our
data indicate the perceptible sweet taste is not related to the
physiological perception of coffee carbohydrates. This finding
helps motivate multiple new hypotheses for the driving factors
of perceived sweetness in black coffee, and further analysis
should investigate the relationship between TDS and perceived
sweetness, or aroma intensity and perceived sweetness. Nonethe-
less, understanding the timing of attribute extraction allows
9
wileyonlinelibrary.com/jsfa
 www.soci.org
coffee brewers to customize their process in order to extract the
exact flavors they want, offering the coffee industry more explicit
guidelines for brewing coffee with the desired profile.
FUNDING
Funding was provided by the Specialty Coffee Association with
underwriting by Breville Corporation.
ACKNOWLEDGEMENTS
We would like to thank the students in Dr Carlito Lebrilla's labora-
tory for assistance and training in chemical analysis (Eshani Nan-
dita, Juan Castillo, and Maurice Wong), as well as students with
the UC Davis Coffee Center who supported this project (Diego
Gabriel Docto, Jake Dykeman, Melissa Richards, Jiexin Liang, Ka
Chun Chan, Lik Xian Lim, and Reece Guyon). Credit goes to Tyler
Simons for R code for principal component analysis and general
statistical analysis help. Thanks go to Wilbur Curtis Co. for donat-
ing the brewers used in this project and Java City Coffee Roasters
for donating the coffee used in the sensory analysis. Finally, we
thank the Specialty Coffee Association and Breville Corp. for fund-
ing support of this research.
SUPPORTING INFORMATION
Supporting information may be found in the online version of this
article.
REFERENCES
1 Flament I and Bessière-Thomas Y, Coffee Flavor Chemistry. John Wiley &
Sons Inc., New York, p. 410 (2002).
2 Hofmann T, Czerny M, Calligaris S and Schieberle P, Model studies on
the influence of coffee melanoidins on flavor volatiles of coffee bev-
erages. J Agric Food Chem 49:2382–2386 (2001).
3 Sunarharum WB, Williams DJ and Smyth HE, Complexity of coffee fla-
vor: a compositional and sensory perspective. Food Res Int 62:
315–325 (2014).
4 Bhumiratana N, Adhikari K and Chambers E, Evolution of sensory
aroma attributes from coffee beans to brewed coffee. LWT – Food
Sci Technol 44:2185–2192 (2011).
5 Spencer M, Sage E, Velez M and Guinard J-X, Using single free sorting
and multivariate exploratory methods to design a new coffee
taster's flavor wheel. J Food Sci 81:S2997–S3005 (2016).
6 Lingle TR, The Coffee Brewing Handbook: A Systematic Guide to Coffee
Preparation [Internet]. Specialty Coffee Association of America,
Long Beach, CA (1996).
7 Lockhart EE and Pan American Coffee Bureau, National Coffee Associ-
ation of U.S.A, The Soluble Solids in Beverage Coffee as an Index to Cup
Quality. Coffee Brewing Institute, New York (1957).
8 Lee TA, Kempthorne R and Hardy JK, Compositional changes in
brewed coffee as a function of brewing time. J Food Sci 57:
1417–1419 (1992).
10
wileyonlinelibrary.com/jsfa © 2020 Society of Chemical Industry
View publication stats
ME Batali et al.
9 Ludwig IA, Sanchez L, Caemmerer B, Kroh LW, De Peña MP and Cid C,
Extraction of coffee antioxidants: impact of brewing time and
method. Food Res Int 48:57–64 (2012).
10 Mestdagh F, Davidek T, Chaumonteuil M, Folmer B and Blank I, The
kinetics of coffee aroma extraction. Food Res Int 63:271–274 (2014).
11 Cines BM and Rozin P, Some aspects of the liking for hot coffee and
coffee flavor. Appetite 3:23–34 (1982).
12 Geel L, Kinnear M and de Kock HL, Relating consumer preferences to sen-
sory attributes of instant coffee. Food Qual Pref 16:237–244 (2005).
13 Rogers WJ, Michaux S, Bastin M and Bucheli P, Changes to the content
of sugars, sugar alcohols, myo-inositol, carboxylic acids and inor-
ganic anions in developing grains from different varieties of Robusta
(Coffea canephora) and Arabica (C. arabica) coffees. Plant Sci 149:
115–123 (1999).
14 Redgwell R and Fischer M, Coffee carbohydrates. Braz J Plant Physiol
18:165–174 (2006).
15 Tian T, Freeman S, Corey M, German JB and Barile D, Chemical charac-
terization of potentially prebiotic oligosaccharides in brewed coffee
and spent coffee grounds. J Agric Food Chem 65:2784–2792 (2017).
16 Lawless HT and Heymann H, Sensory Evaluation of Food: Principles and
Practices. Elementary Food Science, 2nd edn. Springer, New York (2010).
17 Frost SC, Ristenpart WD and Guinard J, Effect of basket geometry on
the sensory quality and consumer acceptance of drip brewed coffee.
J Food Sci 3:2297–2312 (2019).
18 Ristenpart WD and Kuhl T, The Design of Coffee: An Engineering
Approach. Ristenpart/Kuhl Publishing, Davis, CA (2017).
19 Amicucci MJ, Galermo AG, Nandita E, Vo T-TT, Liu Y, Lee M et al., A
rapid-throughput adaptable method for determining the monosac-
charide composition of polysaccharides. Int J Mass Spectrom 438:
22–28 (2019).
20 Xu G, Amicucci MJ, Cheng Z, Galermo AG and Lebrilla CB, Revisiting
monosaccharide analysis – quantitation of a comprehensive set of
monosaccharides using dynamic multiple reaction monitoring. Ana-
lyst 143:200–207 (2018).
21 Gay C, Invitation to comment. Food Qual Pref 9:166 (1998).
22 Abdi H and Williams LJ, Principal component analysis. Wiley Interdiscip
Rev Comput Stat 2:433–459 (2010).
23 Low JYQ, McBride RL, Lacy KE and Keast RSJ, Psychophysical evaluation
of sweetness functions across multiple sweeteners. Chem Senses 42:
111–120 (2017).
24 Moskowitz HR, Ratio scales of sugar sweetness. Perceptions & Psycho-
physics 7:315–320 (1970).
25 Paulus K and Reisch AM, The influence of temperature on the thresh-
old values of primary tastes. Chem Senses 5:11–21 (1980).
26 Clarke RJ and Macrae R eds, Coffee: Volume 1: Chemistry. Springer Neth-
erlands, Dordrecht (1985).
27 Buffo RA and Cardelli-Freire C, Coffee flavour: an overview. Flavour Fra-
grance J 19:99–104 (2004).
28 Rodrigues CI, Marta L, Maia R, Miranda M, Ribeirinho M and Máguas C,
Application of solid-phase extraction to brewed coffee caffeine and
organic acid determination by UV/HPLC. J. Food Compos Anal 20:
440–448 (2007).
29 Spence C, Multisensory flavor perception. Cell 161:24–35 (2015).
30 Sánchez López JA, Wellinger M, Gloess AN, Zimmermann R and
Yeretzian C, Extraction kinetics of coffee aroma compounds using
a semi-automatic machine: on-line analysis by PTR-ToF-MS. Int J
Mass Spectrom 401:22–30 (2016).
31 Charles M, Romano A, Yener S, Barnabà M, Navarini L, Märk TD et al.,
Understanding flavour perception of espresso coffee by the combi-
nation of a dynamic sensory method and in-vivo nosespace analysis.
Food Res Int 69:9–20 (2015).
J Sci Food Agric 2020