1. Ghanaian Cocoa Bean Fermentation
Characterized by Spectroscopic and
Chromatographic Methods and Chemometrics
Patrick C. Aculey, Pia Snitkjaer, Margaret Owusu, Marc Bassompiere, Jemmy Takrama, Lars Nørgaard, Mikael A. Petersen, and
Dennis S. Nielsen
Abstract: Export of cocoa beans is of great economic importance in Ghana and several other tropical countries. Raw cocoa
has an astringent, unpleasant taste, and flavor, and has to be fermented, dried, and roasted to obtain the characteristic cocoa
flavor and taste. In an attempt to obtain a deeper understanding of the changes in the cocoa beans during fermentation
and investigate the possibility of future development of objective methods for assessing the degree of fermentation, a
novel combination of methods including cut test, colorimetry, fluorescence spectroscopy, NIR spectroscopy, and GC-MS
evaluated by chemometric methods was used to examine cocoa beans sampled at different durations of fermentation and
samples representing fully fermented and dried beans from all cocoa growing regions of Ghana. Using colorimetry it
was found that samples moved towards higher a∗ and b∗ values as fermentation progressed. Furthermore, the degree of
fermentation could, in general, be well described by the spectroscopic methods used. In addition, it was possible to link
analysis of volatile compounds with predictions of fermentation time. Fermented and dried cocoa beans from the Volta
and the Western regions clustered separately in the score plots based on colorimetric, fluorescence, NIR, and GC-MS
indicating regional differences in the composition of Ghanaian cocoa beans. The study demonstrates the potential of
colorimetry and spectroscopic methods as valuable tools for determining the fermentation degree of cocoa beans. Using
GC-MS it was possible to demonstrate the formation of several important aroma compounds such 2-phenylethyl acetate,
propionic acid, and acetoin and the breakdown of others like diacetyl during fermentation.
Keywords: cocoa, fluorescence, GC-MS, NIR, spectroscopy
Practical Application: The present study demonstrates the potential of using colorimetry and spectroscopic methods as
objective methods for determining cocoa bean quality along the processing chain. Development of objective methods for
S: Sensory & Food
determining cocoa bean quality will be of great importance for quality insurance within the fields of cocoa processing
and raw material control in chocolate producing companies.
Quality
Introduction cocoa flavor and taste (Thompson and others 2001). Following
Cocoa beans originate as seeds in the fruit pods of the tree harvest the cocoa pods are opened and the pulp surrounding the
Theobroma cacao. They are the principal raw material of chocolate beans spontaneously inoculated with a range of microorganisms
(Schwan and Wheals 2004). Approximately 2/3 of the world’s that metabolise pulp sugars producing mainly ethanol and lactic
cocoa is produced in West Africa. Ghana is the world’s 2nd largest acid. Some of the ethanol is further oxidised to acetic acid by
producer accounting for around 20% of the world production acetic acid bacteria through an exothermal process. The ethanol
(Anon 2007). Being the largest export commodity cocoa is of great and acetic acid penetrate the testa and in combination with the
economical importance for Ghana as a country and of even bigger heat produced (reaching around 50 ◦ C in the fermenting mass)
socio-economic importance in the cocoa growing communities kill the germ and break down cell walls in the bean. This initiates
and villages around the country. the processes leading to well fermented beans (Thompson and
Raw cocoa has an astringent, unpleasant taste, and flavor and others 2001; Schwan and Wheals 2004; Nielsen and others 2007).
has to be fermented, dried, and roasted to obtain the characteristic Following breakdown of the cell walls in the bean, numerous bio-
chemical processes take place leading to the breakdown of proteins
to peptides and amino acids (Hashim and others 1998; Lerceteau
and others 1999; Buyukpamukcu and others 2001); and break-
MS 20090994 Submitted 10/7/2009, Accepted 5/24/2010. Authors Aculey and down of anthocyanins to anthocyanidins and sugars (Pettipher
Takrama are with Cocoa Research Inst. of Ghana (CRIG), Tafo-Akim, Ghana. 1986; Wollgast and Anklam 2000). During fermentation and dry-
Authors Snitkjaer, Owusu, Bassompiere, Nørgaard, Petersen, and Nielsen are with ing the polyphenol content (including anthocyanidins) decrease
Dept. of Food Science, Faculty of Life Sciences, Centre for Advanced Food Science
(LMC), Univ. of Copenhagen, Denmark. Direct inquiries to author Nielsen (E-mail:
as some diffuse out of the beans while others are oxidised and
dn@life.ku.dk). polymerise to insoluble high-molecular-weight compounds (tan-
nins) during fermentation and drying (Pettipher 1986; Wollgast
C 2010 Institute of Food Technologists R
S300 Journal of Food Science r Vol. 75, Nr. 6, 2010 doi: 10.1111/j.1750-3841.2010.01710.x
Further reproduction without permission is prohibited

2. Cocoa bean fermentation characterization . . .
and Anklam 2000). This leads to changes in the internal color of covered with plantain leaves. The whole stack was left to ferment
the dried cocoa beans from the dark grey (slaty) color of the unfer- for 6 d.
mented bean through the deep-purple color of under fermented At 24 h intervals approximately 300 g beans were sampled with
beans to the brown color of the fully fermented bean (Pettipher a sterile plastic bag. After sampling the samples were divided into
1986; Wollgast and Anklam 2000). 2 portions (12 samples), with one portion being sun-dried for 6
The change in cocoa bean color during the fermentation is to 10 d on raised raffia-mats and one portion being dried in a
exploited in a simple method for determining the degree of fer- fan-assisted air extraction oven at 50 ◦ C for 3 d. After sun-drying
mentation called the cut test. In short, a number of beans are all samples were finally dried in a fan-assisted air extraction oven
cut lengthwise and the internal color assessed. Brown beans are at 50 ◦ C for 48 h.
considered well fermented, violet beans partly fermented and grey Furthermore, 26 samples representing fully fermented and dried
(slaty) beans unfermented. It is generally believed that there is an (“ready-to-sell”) cocoa beans from farms representing all co-
inverse relationship between the flavor developed and the purple coa growing regions (Western, Brong-Ahafo, Ashianti, Central,
color retained in the beans. The method is widely used due to its Volta and Eastern region) of Ghana and 4 samples represent-
simplicity but is not without problems as color determination us- ing fermentations carried out at CRIG were included in the
ing the cut test is considered subjective and difficult to standardise study. The beans were bought directly from the farmers and sub-
(Lopez 1984; Wood and Lass 1985). jected to same analyses as the beans sampled during the controlled
Spectroscopic techniques such as NearInfraRed (NIR) spec- fermentations.
troscopy and fluorescence spectroscopy offer valuable alterna- The samples were airlifted to Denmark and stored at – 20 ◦ C
tives to traditional methods for rapid determination of selected until further analysis.
chemical compounds as well as spectroscopic fingerprinting in a
wide range of food and food stuff matrices (Williams and Nor- Cut test
ris 2001; Christensen and others 2006). Previously, NIR spec- Cut test was carried out basically as described by Wood and Lass
troscopy has been used to determine quality parameters such (1985) using a Magra Cutter (Model 12, Teserba, Switzerland).
as fat, protein, and carbohydrates in cocoa and cocoa products
(Kaffka and others 1982; Permanyer and Perez 1989). More re- Sample preparation for analysis
cently, Whitacre and others (2003) successfully applied NIR to Total of 100 beans from each sample were ground in liquid
predict the content of proanthocyanidins in cocoa and Davies nitrogen using a coffee grinder (Braun). The grounded powder
and others (1991) attempted to predict the sensory quality of was stored at – 20 ◦ C until further analysis. To investigate the
the finished chocolate with NIR measurements of ground raw difference between whole and de-shelled beans, samples of sun-
cocoa beans with promising results. However, previously only dried beans from a tray fermentation (6 samples) carried out at
little attention has been given to the possibilities of investigat- CRIG were divided into 2 sub-samples. One sub-sample was
ing changes during fermentation and none of the mentioned carefully de-shelled before grinding, and 1 sub-sample was ground
methods have found commercial use in the cocoa/chocolate as whole beans yielding a total of 12 samples.
industry.
S: Sensory & Food
In an attempt to obtain a deeper understanding of the changes Color measurement (lab system)
in the cocoa beans during fermentation and investigate the possi- The surface color of each sample was measured using a tris-
Quality
bility of future development of objective methods for assessing the timulus color analyzer for measuring reflective colors of surfaces
degree of fermentation, a novel combination of methods includ- with a Chroma Meter CR-300 (Minolta Camera Co. Ltd, Osaka,
ing NIR spectroscopy, fluorescence spectroscopy, colorimetry, and Japan). The instrument was standardised using a white calibration
GC-MS evaluated by chemometric methods was used to examine plate (Calibration Plate CR-A43). Total of 20 g of ground co-
cocoa beans sampled at different times of fermentation. Further- coa beans was transferred into a white plastic dish and the surface
more, to investigate whether local differences in fermentation and color determined. The colorimeter was set to perform 3 repeated
drying practices among farmers influence the quality of the fi- scans for each measurement and the measurements were done at
nal product, samples representing fully fermented and dried beans 4 different places on each sample giving 4 replicates per sample
from all cocoa growing regions of Ghana were included in the (Møller and others 2000).
study.
Fluorescence spectroscopy—BioView
Materials and Methods Auto-fluorescence was measured on a BioView spectrofluo-
rometer from Delta Lights & Optics (Lyngby, Denmark), with a
Cocoa fermentations pulsating xenon lamp exciting at 20 nm broad bands from 260 to
Cocoa pods were harvested by traditional methods during Oc- 560 nm using rotating filters. The emission landscapes consisting
tober (average ambient temperature during the day 28 to 30 ◦ C) of 20 nm broad bands from 300 to 600 nm were provided also by
in Mixed Hybrid Cocoa plantations in Ghana, near New Tafo the use of rotating filters and recollected in the BioView program
(Eastern Region). The pods were harvested over 3 to 7 d and (Svenstrup and others 2005). The samples were measured using a
opened on the following day with a cutlass at the Cocoa Research fiber optic probe with direct contact to the sample analyzed. One
Insti. of Ghana (CRIG) where the fermentation were carried out scan was performed per measurement. Each sample was measured
as well. in duplicate.
Following opening of the pods, the beans were placed in
wooden trays (122 × 91 × 10 cm with a raffia mat at the bottom Near infrared (NIR) spectroscopic analysis
of each tray) with approximately 100 kg beans placed in each tray. Near infrared spectra were acquired in reflectance mode, on
Total of 8 trays were stacked on top of each other and the top tray a spectrophotometer from Foss NIRSystems Inc. (Model 6500,
Vol. 75, Nr. 6, 2010 r Journal of Food Science S301

3. Cocoa bean fermentation characterization . . .
Silver Springs, Md., U.S.A.) with a tungsten halogen lamp and an Another sample set represented fully fermented and dried (com-
internal ceramic lamp (Nielsen and others 2008). The instrument mercial quality) cocoa beans obtained from farmers representing
was calibrated and standardized with a pectin standard before the all cocoa growing regions in Ghana.
measurements. The samples were measured in a special glass ring
cap capable of holding approximately 5 g of ground powder. The
cup was carefully compacted with a back plate rubber seal and Cocoa bean changes during fermentation investigated by
spun in a Spinning Module (NR6506). The spectrum recorded cut test, colorimetry, fluorescence, and NIR spectroscopy
was the average of 16 scans per sample. Each sample was measured Principal component analysis score plots describing the progress
in duplicate. The results were recorded as log(1/R), where R is of fermentation using the traditional cut test and instrumen-
the reflectance and a total of 850 variables were obtained in the tal methods based on colorimetry, fluorescence and NIR spec-
range 800 nm to 2498 nm (with 2 nm intervals). troscopy are shown in Figure 1A to 1D.
As seen from Figure 1A, the drying method (sun compared
Dynamic headspace analysis–GC-MS with oven drying) heavily influences the cut test scores during
Volatile aroma components from 5 g of ground, dried beans the early stages of fermentation. From approximately 72 h of fer-
were detected and identified basically following the procedure de- mentation and onwards the drying method has only negligible
scribed by Juric and others (2003). Sampling was done on Tenax- influence on the cut test scores. During the fermentation a clear
TA traps at 30 ± 1◦ C. The trapped volatiles were desorbed using progress is observed, with the unfermented and shortly fermented
an automatic thermal desorption unit (ATD 400, Perkin Elmer, samples being judged as “slaty” and “purple” followed by a move
Norwalk, Conn., U.S.A.) and automatically transferred to a gas towards “purple-brown” and “brown” by the end of fermentation
chromatograph-mass spectrometer (GC-MS, G1800A GCD Sys- (Figure 1A) which is in agreement with previously published re-
tem, Hewlett-Packard, Palo Alto, Calif., U.S.A.). Separation of sults (Lopez 1986). It is generally recognised that the drying rate
volatiles was carried out on a DB-Wax capillary column 30 m and temperature influence the development of flavor and aroma
long × 0.25 mm internal diameter, 0.25 μm film thickness. The precursors during drying and that drying at a too high temperature
mass spectrometric detector operated in the electron ionisation may impair the development of these precursors (Thompson and
mode at 70 eV. Mass-to-charge ratios between 15 and 300 were others 2001). In the present study, it was found that the oven-dried
scanned. Volatile compounds were identified by matching their samples appeared browner than the corresponding sun-dried beans
mass spectra with those of a commercial database (Wiley275.L, and consequently more beans were judged as “brown” by the cut
HP product nr G1035A). The software program, MSDChemsta- test, probably due to more pronounced nonenzymatic browning
tion (Version E.01.00.237, Agilent Technologies, Palo Alto, Calif., by Maillard-type reactions (Kyi and others 2005).
U.S.A.), was used for data analysis. In an attempt to determine the color of the fermented co-
coa beans in a more objective manner than the visual assessment
Data analysis of the cut test, colorimetry was used to determine development
Cut test, colorimetric, fluorescence, NIR, and GC-MS data in the L ∗ a∗ b∗ -values of the cocoa beans during fermentation.
were all subjected to Principal Component Analysis (PCA) for In agreement with the cut test results, it was found that the
S: Sensory & Food
evaluating systematic variations. In short PCA provides an ap- drying method also influences the colorimetry measurements as
proximation of a data matrix, X, expressed as the product of 2 seen from Figure 1B. As fermentation time increase the sam-
Quality
sets of vectors, T (scores) and P (loadings), that capture the latent ples move towards higher a∗ - and b∗ -values as seen in the bi-plot
factors of X and are referred to as principal components (Wold (Figure 1B) corresponding well with the samples becoming in-
and others 1987). The predictive performance of the GC-MS data creasingly brown with time. Ilangantileke and others (1991) also
were evaluated using Partial Least Squares (PLS) regression (Haa- used color-measurements to estimate the degree of cocoa bean
land and Thomas 1988). All models were validated using full cross fermentation. Despite the experimental differences, comparable
validation (Wold 1978). The NIR spectra were Standard Nor- results were obtained indicating the potential of colorimetry to
mal Variate (SNV) transformed and mean-centered (Barnes and become an objective and simple replacement for the cut test and
others 1989) and the GC-MS data were log10 -transformed before most probably grouping of samples in the categories slaty, purple,
further analysis. All multivariate data analyses were performed us- purple-brown, and brown is obtainable from a linear combination
ing MATLAB version 14 (MathWorks, Natick, Mass., U.S.A.) of the L ∗ a∗ b∗ -values.
and LatentiX Version 1.00 (Latent5, Copenhagen, Denmark). All The samples are clearly separated on the basis of fermentation
data were transferred to MATLAB and LatentiX using in-house time using fluorescence spectroscopy as seen from the score plot
routines written in MATLAB. in Figure 1C (PC2 compared with PC3) indicating the potential
use of this method for determining the degree of fermentation.
Results and Discussion Interestingly, whole beans grouped away from the de-shelled oven
Two different sample sets were investigated in the present study. and sun-dried beans indicating that the shell-content influenced
One set represented samples taken during controlled tray fermen- the fluorescence spectra more than the drying method. PCA of
tations with 24 h intervals during 6 d of fermentation. The method the NIR-spectra (Figure 1D) showed the same trend as the fluo-
used for drying the fermented beans is known to influence the rescence spectra (Figure 1C) with a clear separation on the basis
quality of the final product (Thompson and others 2001). To of fermentation time as well as shell content (whole beans com-
investigate the effect of drying method, the beans were either pared with de-shelled beans). Recently, we have shown that NIR-
sun-dried or oven-dried following fermentation. A portion of the spectra of cocoa beans sampled during fermentation are strongly
sun-dried beans was manually de-shelled following drying before correlated with microbiological changes (investigated using De-
further analysis as it is known that the chemical composition of naturing Gradient Gel Electrophoresis) in the pulp surrounding
the shell is different from the cotyledon (Lopez and Dimick 1995). the beans (Nielsen and others 2008) underlining the potential of
S302 Journal of Food Science r Vol. 75, Nr. 6, 2010

4. Cocoa bean fermentation characterization . . .
spectroscopic techniques as a valuable tool for investigating and methylbutanal to be among the most odour-active compounds
categorising cocoa beans. identified in an extract from cocoa powder. During roasting a sig-
nificant increase will occur due to Strecker degradation (M¨ nch
u
Formation of volatile compounds during and Schierberle 1998; Schieberle 2005), and this is probably the
fermentation investigated using GC-MS most important source of 2/3-methylbutanal in cocoa. The co-
Fermentation and drying are essential steps in the flavor forma- coa beans in this study were analyzed unroasted, and the 2/3-
tion of cocoa beans and the flavor potential of the beans is an im- methylbutanal found is most probably of microbiological origin. It
portant quality parameter (Schwan and Wheals 2004). Figure 2A is well documented that lactic acid bacteria can convert isoleucine
shows a clear change in the content and composition of volatile to 2-methylbutanal and leucine to 3-methylbutanal (Singh and
components during fermentation. Samples fermented 0 to 24 h others 2003; Smit and others 2004). Yeasts have, however, been
are placed to the left in the plot having negative values on demonstrated to reduce 2/3-methylbutanal to the corresponding
PC1 while samples fermented for 72 to 120 h are placed to alcohols rather efficiently (Perp´ te and Collin 2000), which would
e
the right having positive values on PC1. From Figure 2B it explain the decline demonstrated during fermentation here. Mi-
is seen that cocoa beans fermented for 24 h or less have rela- crobial activity is therefore not expected to be the main source of
tively high concentrations of 2-methylpropanal, 2,3-butanedione 2/3-methylbutanal in cocoa.
(diacetyl)/2-pentanone, 2-pentanol, methyl acetate, 2-heptanone, Cocoa beans fermented for 72 h or more have higher lev-
2-pentyl propanoate, 1-pentanol, 2/3-methylbutanal, tetrahydro- els of propanoic acid, linalool oxide, 3-hydroxy-2-butanone
2-methyl furan, 2-methyl-1-propanol, and ethyl acetate. Of these (acetoin), 2-methylpropanoic acid, 1-hydroxy-2-propanone, 3-
compounds, Frauendorfer and Schierberle (2006) reported 2/3- methylbutanoic acid, acetic acid, 2-phenylethyl acetate, 2,3,5,
S: Sensory & Food
Quality
Figure 1–(A) Bi-plot (combining scores and loadings) of PC1 compared with PC2 from a PCA on oven-dried (OD, solid line) and sun-dried (SD, dashed)
fermentation samples and cut test variables. (B) Bi-plot of PC1 compared with PC2 from a PCA on oven-dried (OD, solid line) and sun-dried (SD, dashed)
fermentation samples and color (L∗ a∗ b∗ ) variables. (C) Score plot of PC2 compared with PC3 from a PCA on oven-dried (OD, solid line), sun-dried (SD,
dashed), and sun-dried whole beans (SDW, dotted) fermentation samples and fluorescence data. (D) Score plot of PC1 compared with PC2 from a PCA
on oven-dried (OD, solid line), sun-dried (SD, dashed), and sun-dried whole beans (SDW, dotted) fermentation samples and near infrared data (800 to
2498 nm, SNV-transformed and mean-centered). Time is given in hours. P: purple, PB: purple brown, B: brown, S: slaty.
Vol. 75, Nr. 6, 2010 r Journal of Food Science S303

5. Cocoa bean fermentation characterization . . .
transformed). (B) Loading plot of PC1 compared with PC4 from a PCA on oven-dried, sun-dried, and sun-dried whole beans samples and GC-MS data (log10 transformed). (C) Zoom of the marked area in Fig.
Figure 2–(A) Score plot of PC1 compared with PC4 from a PCA on oven-dried (OD, solid line), sun-dried (SD, dashed), and sun-dried whole beans (SDW, dotted) fermentation samples and GC-MS data (log10
6-tetramethylpyrazine, 2-pentyl acetate, benzaldehyde, trimethyl
pyrazine, benzene ethanol, 3-methylbutyl acetate, and linalool
(Figure 2B and 2C). The increased levels of organic acids are a re-
sult of the metabolisation of sugars from the pulp surrounding the
cocoa beans (Bonvehi 2005). Propanoic acid, 2-methylpropanoic
acid, 3-methylbutanoic acid, and acetic acid are all reported to
be important odor-active compounds in cocoa (Bonvehi 2005;
Frauendorfer and Schierberle 2006; Krings and others 2006).
During fermentation diacetyl decreases and acetoin increases.
This indicates activity of lactic acid bacteria and relatively anaer-
obic conditions (reduction of diacetyl to acetoin; Kandler 1983).
Furthermore, some acetate esters are formed during fermenta-
tion (2-phenylethyl acetate, 2-pentyl acetate, and 3-methylbutyl
acetate) while others decrease (methyl acetate and ethyl acetate)
(Figure 2B and 2C). Of these 2-phenylethyl acetate is reported to
be important for the aroma of cocoa (Bonvehi 2005; Frauendorfer
and Schierberle 2006; Krings and others 2006). The ester pro-
duction is most likely a result of yeast metabolism. 2-Phenylethyl
acetate and 3-methylbutyl acetate are also important esters in beer
and their production is reported to depend on the growth condi-
tions of the yeast, among others, low levels of oxygen (Verstrepenm
and others 2003).
The main part of the pyrazines in cocoa are formed during roast-
ing. In fact, the level of tri- and tetramethylpyrazine has been sug-
gested as an indicator of the degree of roasting (Ramli and others
2006). The increasing levels of tri- and tetramethylpyrazine during
fermentation (Figure 2C) are, however, due to enzymatic activity.
Gill and others (1984) demonstrated that tetramethylpyrazine is a
metabolic product of Bacillus subtilis, and the formation indicates
B. subtilis activity during the fermentation. In general Bacillus spp.
including B. subtilis reach high numbers during the later stages
of Ghanian cocoa fermentations (Carr and Davies 1980; Nielsen
and others 2007). Ramli and others (2006) also found tetram-
ethylpyrazine in unroasted beans, but not trimethylpyrazine.
S: Sensory & Food
Linalool (Figure 2B and C) is reported to be an odour-active
compound in cocoa (Bonvehi 2005; Frauendorfer and Schierberle
Quality
2006; Krings and others 2006) and to contribute to the flowery
and tea-like flavor of some cocoa varieties (Hansen and others
1998). It has been suggested that glycosidases release glycoside-
bound terpenes like linalool during fermentation, explaining the
increase observed (Hansen and others 1998).
Figure 3A and 3B support the close relationship between fer-
mentation and the changes described previously. A PLS1 model
based on all volatiles predicts the fermentation time reasonable
well (r = 0.86, RMSECV = 21.3, Figure 3B). In fact, acetic acid
concentration alone can reasonably predict fermentation time as
seen from Figure 3A (r = 0.86, RMSECV = 21.3) and could pos-
sibly be exploited as an objective and fairly simple measurement
of the degree of fermentation. Overall, it is seen that the changes
in levels of volatiles demonstrated here correspond well with the
present knowledge on fermentation processes. Although roasting
has an immense impact on the odor of cocoa, for example by
formation of pyrazines (Jinap and others 1998), some compounds
formed during fermentation such as organic acids, linalool, and
2B. Time is given in hours.
esters persist and strongly influence taste and flavor of the final
product, chocolate.
Fermented and dried cocoa beans representing all cocoa
growing regions in Ghana investigated by cut test,
colorimetry, fluorescence, and NIR spectroscopy
It is well known that different cocoa subspecies and varieties pro-
duce cocoa beans with different aroma, flavor, and color potential
S304 Journal of Food Science r Vol. 75, Nr. 6, 2010

6. Cocoa bean fermentation characterization . . .
(Wood and Lass 1985). Additionally, the influence of geographic though distinct groupings are not evident. Cocoa beans from
origin at the country level on aroma and flavor of cocoa beans has the Western farms do, however, differ from most of the other
been investigated to some extent (Hernandez and Rutledge 1994; farms by having higher levels of benzaldehyde, 2-heptanone,
Jinap and others 1995; Caligiani and others 2007). However, the 2/3-methylbutanal, 2-nonanone, and 2,6-dimethyl-4-heptanol
vast majority of cocoa of commercial quality is traded as bulk ca- and lower levels of acetic acid, propanoic acid, 1-hydroxy-2-
cao (for example, Ghana cocoa, grade I) where each lot represent propanone, and 2-phenylethyl acetate (Figure 5B). This probably
cocoa from numerous farmers and only little is known about the reflect differences in fermentation practices as the production of
influence of geographic origin at the farm to farm level on co- acetic acid during fermentation is dependent on oxygen avail-
coa beans. In the present study, samples of fermented and dried ability (Camu and others 2008) and esters such as 2-phenylethyl
“ready-to-sell” cocoa beans representing all major cocoa grow- acetate are most likely a result of yeast metabolism (Verstrepenm
ing regions of Ghana have been investigated using the cut test, and others 2003).
colorimetry, fluorescence, and NIR spectroscopy and GC-MS to Many of these compounds are—as mentioned previously—
elucidate differences from farm to farm and between the different expected to be odor-active in the roasted cocoa and the re-
Ghanaian cocoa growing regions. gional differences are therefore important for the quality. The
As seen from Figure 4A, no major differences are seen between data could indicate that cocoa beans from Western farms are
the cut test values of the different Ghanaian cocoa growing regions less fermented with lower levels of organic acids and higher
as the majority of the samples cluster together between the Brown levels of 2/3-methylbutanal as mentioned previously. Corre-
and the Purple-Brown loading indicating that the cocoa is well- sponding to Figure 4B to 4D, the Volta farms are also here
fermented (Wood and Lass 1985). Four farms (010, 015, 020, and generally placed opposite to the Western farms in the PCA
229) including both farms representing the Ashanti Region cluster plot.
closer to the Purple and Slaty loadings indicating that the cocoa is
poorly fermented (Figure 4A). Conclusions
The colorimetric analysis in general confirmed the cut test anal-Degree of fermentation could, in general, be well described by
ysis. However, of the 4 farm samples that clustered closer to the
all the methods used. It was demonstrated that colorimetry has
Slaty and Purple loadings (Figure 4A), only Farm 015 clustered the potential to become an objective and simple replacement for
away from the majority of samples in the PCA bi-plot based on the
the cut test, but samples were also clearly separated on the basis
colorimetric results (Figure 4B). The fluorescence and NIR spec- of fermentation time using fluorescence spectroscopy and NIR-
troscopic results confirmed this pattern with Farm 015 clusteringspectrometry. In addition, it was possible to link analysis of volatile
away from the main cluster in Figure 4C as well as Figure 4D. compounds and the formation of acetic acid with predictions of
Interestingly, the samples representing farms in the Volta and
fermentation time.
the Western regions in general (with the exception of Farm 019) When cocoa beans from different regions in Ghana were com-
clustered separately in the score plots based on colorimetric, fluo-
pared no major differences were seen between the cut test values.
rescence and NIR (Figure 4B to D) indicating regional differences
On the contrary, the samples representing farms in the Volta and
in the composition of Ghanaian cocoa beans. the Western regions, in general, clustered separately in the score
S: Sensory & Food
plots based on colorimetric, fluorescence, NIR, and GC-MS in-
Volatile compounds in cocoa beans representing all
Quality
dicating regional differences in the composition of Ghanaian co-
Ghanaian cocoa growing regions investigated using GC-MS coa beans. Many of the detected compounds are expected to be
The score plot in Figure 5A indicates that geographical dif- odor-active in the roasted cocoa and the regional differences are
ferences in the composition of Ghanaian cocoa beans do exist therefore important for the quality.
Figure 3–(A) Actual compared with Predicted plot (one PLS component) for time in hours for oven-dried (OD), sun-dried (SD), sun-dried whole beans
(SDW) based on acetic acid as determined by GC-MS. (B) Actual compared with Predicted plot (one PLS component) for time in hours for oven-dried
(OD), sun-dried (SD), sun-dried whole beans (SDW) based on all GC-MS data (log10 transformed).
Vol. 75, Nr. 6, 2010 r Journal of Food Science S305

7. S: Sensory & Food Cocoa bean fermentation characterization . . .
Quality
Figure 4–(A) Bi-plot of PC1 compared with PC2 from a PCA on farm samples and cut test variables. (B) Bi-plot of PC2 compared with PC3 from a PCA
on farm samples and color data. (C) Score plot of PC2 compared with PC2 from a PCA on farm samples and autoscaled fluorescence data. (D) Score plot
of PC1 compared with PC3 from a PCA on farm samples and near infrared data (800 to 2498 nm, SNV and mean-centered).
Figure 5–(A) Score plot of PC1 compared with PC2 from a PCA on farm samples and GC-MS data (log10 transformed). (B) Loading plot of PC1 compared
with PC2 from a PCA on farm samples and GC-MS data (log10 transformed).
S306 Journal of Food Science r Vol. 75, Nr. 6, 2010

8. Cocoa bean fermentation characterization . . .
Acknowledgment Krings U, Zelena K, Wu S, Berger RG. 2006. Thin-layer high-vacuum distillation to isolate
volatile flavour compounds of cocoa powder. Eur Food Res Technol 223:675–81.
The research performed was partly financed through the EU Kyi TM, Daud WR Mohammed AB, Samsudin MW, Kadhum AH, Talib MZM. 2005. The
W,
INCO project: “Developing biochemical and molecular markers kinetics of polyphenol degradation during the drying of Malaysian cocoa beans. Int J Food
Sci Technol 40:323–31.
for determining quality assurance in the primary processing of Lerceteau E, Rogers J, P´ tiard V, Crouzillat D. 1999. Evolution of cacao bean proteins during
e
cocoa in West Africa—COCOQual” (ICA4-CT-2002–10040). fermentation: a study by two-dimensional electrophoresis. J Sci Food Agric 79:619–25.
Lopez AS. 1984. Limitac˜ o da “prova de corte” no controle de qualidade do cacau comercial.
a
This article is published with the kind permission of the Director Revista Theobroma 14:199–207.
of CRIG. The authors are grateful for the technical assistance Lopez AS. 1986. Chemical changes occuring during the processing of cacao. In: Dimick PS,
of the staff of the Physiology/Biochemistry Div. of the Cocoa editor. Cacoa biotechnology symposium. Pennsylvania State Univ., Pa. p 19–53.
Lopez AS, Dimick PS. 1995. Cocoa fermentation. In: Reed G, Nagodawithana TW, editors.
Research Inst. of Ghana. Enzymes, biomass, food and feed. Weinheim, Germany: VCH. p 561–77.
Møller JKS, Jensen JS, Olsen MB, Skibsted LH, Bertelsen G. 2000. Effect of residual oxygen
on colour stability during chill storage of sliced, pasteurised ham packaged in modified
References atmosphere. Meat Sci 54:399–405.
Anon. 2007. ICCO annual report 2006/2007. International Cocoa Organization. Available M¨ nch P, Schierberle P. 1998. Quantitative studies on the formation of key odorants in thermally
u
from: http://www.icco.org. Accessed May 3, 2010. treated yeast extracts using stable isotope dilution assays. J Agric Food Chem 46:4695–701.
Barnes RJ, Dhanoa MS, Lister SJ. 1989. Standard normal variate transformation and de-trending Nielsen DS, Teniola OD, Ban-Koffi L, Owusu M, Andersson T, Holzapfel WH. 2007. The
of near-infrared diffuse reflectance spectra. Appl Spectrosc 43:772–7. microbiology of Ghanaian cocoa fermentations analysed using culture dependent and culture
Bonvehi JS. 2005. Investigation of aromatic compounds in roasted cocoa powder. Eur Food Res independent methods. Int J Food Microbiol 114:168–86.
Technol 221:19–29. Nielsen DS, Nielsen PS, Van Den Berg F. 2008. Investigating the fermentation of cocoa by
Buyukpamukcu E, Goodall DM, Hansen CE, Keely BJ, Kochhar S, Wille H. 2001. Characteri- correlating denaturing gradient gel electrophoresis profiles and near infrared spectra. Int J
zation of peptides formed during fermentations of cocoa beans. J Sci Food Agric 49:5822–7. Food Microbiol 125:133–40.
Caligiani A, Cirlini M, Palla G, Ravaglia R, Arlorio M. 2007. GC-MS detection of chiral Permanyer JJ, Perez ML. 1989. Compositional analysis of powdered cocoa products by near-
markers in cocoa beans of different quality and geographic origin. Chirality 19:329–34. infrared reflectance spectroscopy. J Food Sci 54:768–9.
a ´
Camu N, Gonz´ les A, De Winter T, Van Schoor A, De Bruyne K, Vandamme P, Takrama JS, Perp´ te P, Collin S. 2000. How to improve the enzymatic worty flavour reduction in a cold
e
Addo SK, De Vuyst L. 2008. Influence of turning and environmental contamination on the contact fermentation. Food Chem 70:457–62.
dynamics of populations of lactic acid and acetic acid bacteria involved in spontanous cocoa Pettipher GL. 1986. An improved method for the extraction and quantitation of anthocyanins in
heap fermentation in Ghana. Appl Environ Microbiol 74:86–98. cocoa beans and its use as an index of the degree of fermentation. J Sci Food Agric 37:289–96.
Carr JG, Davies PA. 1980. Cocoa fermentation in Ghana and Malaysia (part 2): Further micro- Ramli N, Hassan O, Said M, Samsudin W, Idris NA. 2006. Influence of roasting conditions on
biological methods and results. Long Ashton Research Station, Bristol. volatile flavor of roasted Malaysian cocoa beans. J Food Proc Pres 30(3):280–98.
Christensen J, Nørgaard L, Bro R, Engelsen SB. 2006. Multivariate autofluorescence of intact Schieberle P. 2005. The carbon module labelling (CAMOLA) technique: a useful tool for
food systems. Chem Rev 106:1979–94. identifying transistent intermediates in the formation of Maillard-type target molecules. Ann
Davies AMC, Franklin JG, Grant A, Griffiths NM, Shepherd R, Fenwick GR. 1991. Prediction New York Acad Sci 1043:236–48.
of chocolate quality from near-infrared spectroscopic measurements of the raw cocoa beans. Schwan RF, Wheals AE. 2004. The microbiology of cocoa fermentation and its role in chocolate
Vibra Spectrosc 2:161–72. quality. Crit Rev Food Sci Nutr 44:205–22.
Frauendorfer F, Schierberle P. 2006. Identification of the key aroma compounds in cocoa powder Singh TK, Drake MA, Cadwallader KR. 2003. Flavor of cheddar cheese: a chemical and sensory
based on molecular sensory correlations. J Agric Food Chem 54:5521–9. perspective. Compre Rev Food Sci Food Safety 2:139–62.
Gill MS, McLeod AJ, Moreau M. 1984. Volatile components of cocoa with particular reference Smit BA, Engels WJM, Wouters JTM, Smit G. 2004. Diversity of L-leucine catabolism in various
to glucosinolate products. Phytochem 23:1937–42. microorganisms involved in dairy fermentations, and identification of the rate-controlling step
Haaland DM, Thomas EV. 1988. Partial least squares-method for spectral analyses. 1. Relation in the formation of the potent flavour component 3-methylbutanal. Appl Microbio Biotechnol
to other quantitative calibration methods and the extraction of quantitative information. Ana 64:396–402.
Chem 60:1193–202. Svenstrup G, Heimdal H, Nørgaard L. 2005. Rapid instrumental methods and chemometrics
Hansen CE, del Olmo M, Burri C. 1998. Enzyme activities in cocoa beans during fermentation. for the determination of pre-crystallization in chocolate. Int J Food Sci Technol 40:953–
J Sci Food Agric 77:273–81. 62.
Hashim P, Selamat J, Muhammad SKS, Ali A. 1998. Changes in free amino acid, peptide-N, Thompson SS, Miller KB, Lopez AS. 2001. Cocoa and coffee. In: Doyle MJ, Beuchat LR,
sugar and pyrazine concentration during cocoa fermentation. J Sci Food Agric 78:535–42. Montville TJ, editors. Food microbiology—fundamentals and frontiers. Washington, D.C.:
Hernandez CV, Rutledge DN. 1994. Multivariate statistical-analysis of gas chromatograms to ASM Press. p 721–33.
S: Sensory & Food
differentiate cocoa masses by geographical origin and roasting conditions. Analyst 119:1171–6. Verstrepenm KJ, Derdelinckx G, Dufour J-P, Winderickx J, Thevelein JM, Pretorius IS, Delvaux
Ilangantileke SG, Wahyudi T, Bailon MAG. 1991. Assesment methodology to predict quality of FR. 2003. Review—flavor active esters: adding fruitiness to beer. J Biosci Bioeng 96:110–18.
Whitacre E, Oliver J, Van Den Broek R, van Engelen P, Kremers B, Van Der Horst B, Stewart
Quality
cocoa beans for export. J Food Qual 14:481–96.
Jinap S, Dimick PS, Hollender R. 1995. Flavor evaluation of chocolate formulated from cocoa M, Jansen-Beuvink A. 2003. Predictive analysis of cocoa procyanidins using near-infrared
beans from different countries. Food Control 6:105–10. spectroscopy techniques. J Food Sci 68:2618–22.
Jinap S, Rosli WIW, Russly AR, Nordin LM. 1998. Effect of roasting time and temperature on Williams P, Norris K. 2001. Near-infrared technology in the agricultural and food industries.
volatile component profiles during nib roasting of cocoa beans (Theobroma cacao). J Sci Food The American Association of cereal chemists, Minn.
Agric 77:441–8. Wold S. 1978. Cross-validatory estimation of number of components in factor and principal
Juric M, Bertelsen G, Mortensen G, Petersen MA. 2003. Light-induced colour and aroma component models. Technometrics 20:397–405.
changes in sliced, modified atmosphere packaged semi-hard cheeses. Int Dairy J 13:239– Wold S, Esbensen K, Geladi P. 1987. Principal component analysis. Chemomet Intel Lab Sys
49. 2:37–52.
Kaffka KJ, Norris KH, Kulcsar F, Draskovits I. 1982. Attempts to determine fat, protein and Wollgast J, Anklam E. 2000. Review on polyphenols in Theobroma cacoa: changes in composition
carbohydrate content in cocoa powder by the nir technique. Acta Alimentaria 11:271– during the manufacture of chocolate and methodology for identification and quantification.
88. Food Res Int 33:423–47.
Kandler O. 1983. Carbohydrate metabolism in lactic acid bacteria. Antonie van Leeuwenhoek Wood GAR, Lass RA. 1985. Cocoa. London: Longman.
49:209–24.
Vol. 75, Nr. 6, 2010 r Journal of Food Science S307