BK

TP HCM




























©
2009











SPISE2009 1

SPISE2009, the second symposium on sensory evaluation in ASEAN, was held on August 7–
9, 2009 at Ho Chi Minh-city University of Technology, Vietnam. Besides the former purpose
of encouraging the implementation and development of sensory evaluation in Vietnam and in
neighboring countries, this year SPISE aimed to capitalise on and to encourage further
collaborations among Asian food scientists. The focus of SPISE2009 was on “Food
Consumer Insights in Asia: Current Issues and Future.” We had the great pleasure to welcome
more than one hundred of scientists from (among other countries) Vietnam, China, Thailand,
Singapour, the USA, France, Australia, the Netherlands, and Denmark.
The present proceedings are organised into four topics which we used to organised this
meeting sessions and themes:
1. New methods and research tools in consumer research

2. Food choice and consumer behaviour studies

3. Applications in the food industry/Product development

4. Sensory-instrumental relations

The presenters who submitted papers came from a wide range of disciplines, partly because
sensory evaluation so often depends on interdisciplinary collaboration for successful
implementation. We also feel that these papers reflect the wide range of interests and methods
of the field and that they give an accurate image of the state of the art of sensory evaluation
practices and theoretical concerns in the Asian countries. As such, this document should be
of interest for anyone interested in consumer research in Asia.
Our special thanks are due to our partners which participated to the organisation of this
meeting: the HCMC University of Technology, AgroSup Dijon, CESG-Dijon, and
AgroCampus-Ouest. We would also like to thank our sponsors for their generous help:
VITAGORA, Fizz-Biosystème, LogicStream, Vinamilk, and Saoviet. We extend our special
thanks to those who have helped us so much and worked so hard to make this event possible:
Le Minh Tam, Phan Thuy Xuan Uyen, Nguyen Thi Hang, Vu Thi Thanh Phuong, Lam Minh
Thuy, Truong Thi Xuan, Vu Thi Phuong Mai.
SPISE2009 2

Table of Content 
Part 1: New methods and research tools in consumer research ..................................... 4
1. The ideal profile method: combining classical profiling with jar methodology .............. 5
P.H. Punter & T. Worch
2. Consumer preferences for visually presented meals ....................................................... 19
G. Gabrielsen, M.D. Aaslyng, & H.H Reisfelt
3. Observational research: a tool for collecting behavioral data and validating surveys .... 29
S.L. Godwin & E. Chambers IV

Part 2: Food choice and consumer behaviour studies ..................................................... 36

4. Hedonic response to the tastes of wine in Vietnam: Does the region of origin of
consumers matter?........................................................................................................... 37
V.B. Do, B. Patris, T. D. Ha, D.H. Nguyen, & D. Valentin
5. Coffee or margarita: impact of ambiences on beverage choices in a bar........................ 44
C. Dacremont, A. Sutan, F. Galia, J-F. Desmarchelier, & D. Valentin
6. The soupe du jour effect: language as a country-of-origin cue and its impact
on product perception...................................................................................................... 51
T.X.U. Phan & C-F Sheu
7. Attitudes and spending behaviours from a middle class point of view during
the current economic recession in Vietnam .................................................................... 62
T.M. Le, D.T. Do, & D.H. Nguyen
8. Effects of age and geographical origin in preference of consumers for yam
tubers and mix wheat yam flours cookies ....................................................................... 70
B. Ranaivosoa, D. Valentin, V.H. Jeannoda, & J.L. Razanamparany

Part 3: Application in the food industry / Product development.................................... 78

9. Gaining insight into marketing strategies and retailer perceptions of us beef in
Vietnam: a focus group approach.................................................................................... 79
T.T.N. Dinh, D.H. Nguyen, S.S. Harp, L.D. Thompson, J.C. Brooks, M.F. Miller,
J.C. Boyce, D.B. Reed, K.C. Le, S.F. Maxner, & J.L. Lusk

10. Formulation of flour based peanut snack using mixture design ..................................... 90
N. Pengboon, S. Punsuwan, & P. Siriwongwilaichat
11. Application of green tea extract to biscuit cream............................................................ 95
M.N. Dang
SPISE2009 3

Part 4: Sensory-instrumental relationship....................................................................... 104

12. Exploring the optimisation model of Vietnamese consumers for sterilised
milks ................................................................................................................................105
T.B. Nguyen, T.M. Le, & D.H. Nguyen

13. Three compounds with potent -glucosidase inhibitory activity purified from
sea cucumber Stichopus japonicus ................................................................................. 112
H.T. Nguyen & S.M. Kima
14. Volatile components and sensory characteristics and consumer liking of
commercial brand oyster sauces..................................................................................... 123
T.H.D Nguyen, X.C. Wang, & Y.Z. Zhu

15. Relationship between sensory descriptive and chemical property of jiaogulan

(Gynostemma pentaphyllum.) tea with lime juice.......................................................... 140
N. Utama-ang & S. Jaisam
SPISE2009 4

Part 1: New methods and research tools in consumer research 
SPISE2009 5

THE IDEAL PROFILE METHOD: COMBINING CLASSICAL PROFILING WITH

JAR METHODOLOGY
P.H. Punter* and T.Worch
OP&P Product Research, Utrecht, The Netherlands
*Corresponding author: [email protected], Tel: +31.30.2516772
Abstract
In order to make competitive and successful products, companies need to know what
consumers really want. They need detailed insights in the perceived product characteristics
and their contribution to consumer liking. Traditionally, sensory information is collected by
the R&D department and liking information by the market research department. These two
disciplines use very different methodologies and subjects to obtain this information. For the
sensory information, experts or trained panellists are used and for the acceptance data
consumers from the target population are used. The sensory professional in R&D is employed
by the food company, for the collection of consumer preferences companies will use their
own market research department or rely on outside market research agencies. In recent years,
there has been an increasing interest in obtaining both sensory and hedonic information from
the target consumers. Market researchers have been doing this for many years using the JAR
methodologies. But when consumers can answer JAR questions about sensory characteristics,
they should also be able to rate the perceived and ideal intensities directly. The Ideal Profile
(IP) method combines both JAR and classical profiling methodology: consumers are requested
to rate the perceived and ideal intensities for the relevant product characteristics and to give
acceptance ratings. The resulting data are used to compute the effect of the difference from
ideal for each individual attribute on overall liking and to guide R&D in product optimisation.
Keywords: profiling, consumer, product optimisation

1. INTRODUCTION
The success of a food product depends to a large extend on consumer liking: The higher the
liking, the more successful the product. Most methodologies applied in product optimisation
work from the implicit assumption of the existence of an “ideal” product. A key challenge in
product development and optimisation is to modify and redefine the existing product closer to
this ideal. A core assumption underlying this thinking is that consumer’s loss in liking ( ) of
a product is a weighted linear combination of the absolute attribute level deviations from the
ideal product. More formally this can be expressed as:
SPISE2009 6

where is the consumer’s loss in liking for product p, is the relative importance of
deviations on attribute j for consumer’s overall liking judgment, is the consumer
perception of product p on attribute j, and is the ideal level of attribute j (averaged over all
products of reference) that would generate maximum liking (Engel, Blackwell, & Miniard,
1995). When all attributes have the ideal level, there is no loss in liking and liking is
maximised: The larger the deviations from ideal, the higher the loss in liking and the lower
the liking for that product.

The classical sensory approach is to obtain a sensory profile of the different products from
trained assessors, to obtain liking ratings from target consumers and to find out which sensory
aspects drive liking through regression modelling (Lawless & Heymann, 1999; Stone & Sidel,
2004; Meilgaard et al., 2006; Greenhoff & MacFie, 1999) market research has a different
approach to this problem: Just-About-Right (JAR) scaling. Since market research has no
access to trained assessors, they have to extract the necessary information directly from
consumers (Epler et al., 1997; Popper et al, 2004; Popper & Kroll, 2005; ASTM E-18,
Moskowitz et al., 2004). In JAR scaling, consumers give a liking rating for a specific product
and rate it on a series of sensory attributes. In contrast to standard sensory profiling, they do
not simply state their perceived sensory intensity but make an assessment of the difference
between their perceived sensory intensity and their ideal intensity for that aspect. For each
sensory attribute, they indicate whether the intensity is “just about right”, “too weak”, or “too
strong”. From a psychological point of view, this is a complex task and the only output is the
result of the mental arithmetic. This task could be made more explicit by asking the
consumers to give both the perceived and ideal intensity instead of asking them the result of
the subtraction. This is actually the procedure in Ideal Profiling (IP). Consumers rate the
products on a set of relevant sensory attributes (both perceived and ideal intensity) and on
hedonic aspects.
Comparison of sensory profiles from consumers and experts show that the results do not
differ (Moskowitz, 1996; Husson et al., 2001; Worch et al., 2009). The final
recommendations for product improvement based on expert, consumer, or JAR profiles also
suggest a high degree of convergent validity between the various methods in terms of the
order of ideal points (van Trijp et al., 2007).

A number of alternative profiling methods have been developed in the past years which allow
the use of naïve consumers. Most are based on the assumption that consumers should not be
asked attribute questions and use ranking procedures with free choice profiling (Delarue &
Sieffermann, 2004) or use “Napping®” (Pagès, 2005).

Since the results from these different methodologies lead to similar conclusions, the choice
for one particular method can be made on practical grounds.
SPISE2009 7

From a business point of view, presenting product developers with quantitative sensory and
ideal profiles and information about the effect of deviations from ideal on liking is more
efficient than showing them a product map or the results of a ranking. The IP method provides
accurate sensory and ideal spider webs, drivers of liking and suggestions for optimisation, as
will be shown in this paper with an example for eight different fruit yoghourts.

2. MATERIALS AND METHOD

2.1 Samples
The following eight commercially available fruit yoghourts have been tasted:
Raspberry/Peach Orange
Fruit mix Mandarin
Cherry/Orange Maracuja
Mango Dades
2.2 Consumers
A total of 130 consumers participated in the test, 78 females and 52 males. They ranged in
age between 19 and 65 years (median age 42 years). All users fruit yoghurt, 49% is heavy
user (several times a week) and 51% light user (at least once every two weeks). They have
been recruited by telephone from the OP&P consumer database.

2.3 Questionnaire
The questionnaire consisted of six acceptance questions including overall liking (rated on the
9-point hedonic scale), buying intention (5-point scale) and 28 intensity questions (both
perceived and ideal intensity has been asked on 100 mm line scales with anchors at 10% and
90%.).

Acceptance (9-point Intensity and ideal (100 mm line scales)

category scales)
appearance gloss fruity taste Thick in mouth
odour colour intensity freshness fruit smooth in mouth
taste amount fruit sweetness watery/thick
mouth feel recogn. fruit sourness airy mouth feel
aftertaste thick appearance. bitterness firmness fruit
overall liking odour intensity astringent amount fruit mf
buying intention (5-p) fruity odour creamy taste aftertaste intensity
sweet odour mild taste aftertaste length
sour odour fresh taste
taste intensity off taste
The questionnaire has been administered by computer (EyeQuestion software, 2005).

2.4 Procedure
The consumers were invited to the sensory facility (a sixteen booth tasting room) on two
consecutive days for two 60-minute sessions. They were paid for their participation. On each
day, they tasted four of the eight variants. The variants were presented one after another at 12
SPISE2009 8

minute intervals, according to a balanced presentation design (MacFie et al., 1989). The
participants received 125 gram of each variant in a plastic cup with plastic spoons. Between
the presentations, they were advised to clean their palate using crackers and water.

2.5 Statistical analysis

The acceptance and intensity ratings for the different products were analysed by ANOVA
(repeated measurements ANOVA, each subject has tasted all products) and LSD post hoc
analysis. A significant product effect shows that the eight products are rated differently, the
post hoc test shows which individual product differs from which other products. To see if
there is an effect of gender, age class or cluster membership a factorial design with block-
treatment confounding (split-plot design) has been carried out. The consumer panel has also
been segmented on the basis of their liking ratings (Clustering around latent variables,
Vigneau et al., 2001).
For the multivariate analysis of the intensity ratings, a principal component analysis (PCA) on
the correlation matrix for product  panellist with varimax rotation has been carried out.
Liking is regressed on the extracted principal components (this method is also called Principal
Component Regression). All analyses have been carried out using Senstools.NET (beta
version, OP&P product research, 2009).

3. RESULTS
3.1 Acceptance aspects
Comparison of the overall liking ratings show that the eight variants differ significantly from
each other (figure 1). Fruit mix and Cherry/Orange are liked most; Dades and Mango are
liked least. There is no effect of user type (heavy or light users) or gender on overall liking
but there is a significant effect of age class (consumers between 34-49 give higher ratings
than the younger and older consumers, but there is no interaction with yoghourt).
Furthermore, the consumers can be divided into two clusters of 85 and 45 consumers
respectively with different liking patterns (figure 2).
9

8
7 ,1 6 ,9
7 6 ,7 6 ,7 6 ,7
6 ,4
5 ,9
6 5 ,4
5
a a 
4 b  b 
c  c 
3 d  e  b  e   
d  d  d  d 
f  f  d  f   
2 h h h h
h g  h g 
1
/ ) ) ) ) ) f)( ) )
y b d e g h
rr (a ( / (c ( ( ( (
e
b h ix ry
r e o
g
e
g in
r ja s
e
p c m e g n n a u d
s a ti h n a d c
a e u C a
r ra n a
r
a
D
R P r M O a a
F O M M

Figure 1: Overall liking for the different yoghurts and post hoc significances (LSD). The letters in a
bar indicate from which other product this specific product differs significantly
SPISE2009 9

overall liking Cl1 overall liking Cl2

9
8
7
6
5
4
3
2
1
Raspb/

Dades
Orange

Orange
Mango
Fruits

Maracuja
Mandarin
Cherry/
Peach

Figure 2: Overall liking for Cluster 1 (n = 85) and 2 (n = 45).

The two clusters differ mainly in their appreciation for the Dades yoghourt. Cluster 1 clearly
dislikes Dades but Cluster 2 likes them. The clusters differ in usage frequency: Cluster 1 has
more light users and Cluster 2 has more heavy users.
Besides overall liking, participants rated several other aspects on liking. Table 1 shows the
averaged acceptance ratings (9-point scales) and buying intention (5-point scale).
Table 1: Average acceptance ratings and buying intention for the eight yoghourts
Rasp/ Fruit Cherry/ Mango Orange Mandarin Maracuja Dades
Peach mix Orange
appearance 6,7 6,5 7,0 6,7 6,6 7,1 6,5 5,6
odour 6,6 6,6 6,7 6,4 6,8 7,1 6,4 5,9
taste 7,1 7,2 7,0 6,1 7,0 6,6 6,9 5,2
mouth feel 6,8 7,1 6,9 6,4 6,9 6,5 6,7 5,5
aftertaste 6,6 6,9 6,7 6,0 6,6 6,2 6,6 5,3
overall liking 6,7 7,1 6,9 5,9 6,7 6,4 6,7 5,4
buying 3,7 3,9 3,7 3,0 3,7 3,3 3,6 2,6

The eight yoghourts differ significantly from each other for every aspect.

3.2 Sensory and ideal profiles

The eight products differ significantly from each other at each aspect except sour odour. Table
2 presents the ANOVA results (F-ratio’s and p-values).
Table 2: Results of the ANOVA for the intensity attributes (model main effects for product and subject
without interaction)
F-Ratios p-values products F-Ratios p-values
products subjects subjects
gloss 7.57 <.01 10.88 <.01
colour int. 67.20 <.01 4.94 <.01
amount fruit 22.99 <.01 5.03 <.01
recogn. fruit 25.56 <.01 5.75 <.01
thick app. 3.99 <.01 8.39 <.01
odour int. 10.93 <.01 7.38 <.01
SPISE2009 10

fruity odour 18.66 <.01 6.05 <.01

sweet odour 3.76 <.01 9.59 <.01
sour odour 1.02 0.42 10.40 <.01
taste int. 14.45 <.01 7.62 <.01
fruity taste 19.09 <.01 5.01 <.01
freshness fruit 6.22 <.01 6.00 <.01
sweetness 35.85 <.01 5.44 <.01
sourness 12.03 <.01 6.51 <.01
bitterness 11.05 <.01 8.64 <.01
astringent 10.46 <.01 7.19 <.01
creamy taste 9.31 <.01 9.21 <.01
mild taste 3.93 <.01 6.88 <.01
fresh taste 18.82 <.01 6.90 <.01
off taste 5.32 <.01 8.70 <.01
thick in mouth 2.24 0.03 8.85 <.01
smooth in mouth 9.65 <.01 8.48 <.01
watery/thick 2.14 0.04 6.40 <.01
airy mouth feel 5.57 <.01 9.08 <.01
firmness fruit 24.71 <.01 4.02 <.01
amount fruit mf 11.77 <.01 6.35 <.01
int. aftert. 7.16 <.01 7.22 <.01
length aftert. 4.61 <.01 7.14 <.01

Raspberry/Peach Fruit mix Cherry/Orange Mango

Orange Mandarin Maracuja Dades

gloss
length aftert. 80 colour int.
int. aftert. amount fruit
70
amount fruit
recogn. fruit
mf 60

firmness fruit 50 thick app.

airy mouth 40
odour int.
feel
30

watery/thick 20 fruity odour

10
smooth in
0 sweet odour
mouth

thick in
sour odour
mouth

off taste taste int.

fresh taste fruity taste

freshness
mild taste
fruit
creamy taste sweetness
astringent sourness
bitterness

Figure 3: Sensory profiles of the eight variants (n=130)

SPISE2009 11

ID Raspb/Peach ID Fruit mix ID Cherry/Orange ID Mango

ID Orange ID Mandarin ID Maracuja ID Dades

gloss
length aftert. 80 colour int.
int. aftert. amount fruit
70
amount fruit
recogn. fruit
mf 60

firmness fruit 50 thick app.

airy mouth 40
odour int.
feel
30

watery/thick 20 fruity odour

10
smooth in
0 sweet odour
mouth

thick in
sour odour
mouth

off taste taste int.

fresh taste fruity taste

freshness
mild taste
fruit
creamy taste sweetness
astringent sourness
bitterness

Figure 4: Ideal profiles of the eight variants (n=130)

Figure 5: Sensory and average ideal profiles of Fruit mix (liking 7.1) and Dates (liking 5.4) n = 130.
SPISE2009 12

Comparison of the sensory and ideal profiles of the most liked and least liked variant shows
that they clearly differ in their deviation from ideal (figure 5). They both lack recognisable
fruit, Fruit mix lacks more fruit than Dades, both lack fruity odour and taste, odour intensity
and the fruit lacks freshness (more so for Dades). Fruit mix is ideal on the basic tastes, Dades
clearly deviates from ideal on these aspects. Dades has too much off taste and it lacks
mildness and freshness and the mouth feel lacks smoothness and airiness.

4. DRIVERS OF LIKING

First, a PCA is carried out to extract the underlying sensory dimensions (PCA on the product 
consumer  attributes matrix with varimax rotation). The 28 sensory attributes can be
summarised into the following eight underlying factors or dimensions (67% variance
accounted for).
Table 2: Extracted factors
F1 Fruity/fresh F3 Thickness F6 Smooth/mild
fruity taste watery/thick smooth in mouth
taste int. thick in mouth airy mouth feel
freshness fruit thick app. mild taste
firmness fruit creamy taste fresh taste
sweetness F4 Odour aspects gloss
F2 Fruit aspects fruity odour F7 Aftertaste
amount fruit odour int. length aftert.
recogn. fruit sweet odour int. aftert.
amount fruit mf F5 Bitter/off taste F8 Sourness
colour int. off taste sourness
bitterness sour odour
astringent

Since the analysis has been carried out on the product  consumer matrix we can also estimate
the amount of variance which is explained by differences between the products and which can
be attributed to differences between consumers. Table 3 shows the ratio between the product
and consumer variance for each dimension.

Table 3: Variance ratio by dimension

Variance between Variance between Ratio
products consumers
F1 Fruity/fresh 24.6 2.8 8.2
F2 Fruit aspects 17.6 3.0 5.87
F3 Thickness 5.7 4.3 1.32
F4 Odour aspects 11.9 3.8 3.16
F5 Bitter/off taste 7.4 4.7 1.55
F6 Smooth/mild 4.5 4.3 1.05
F7 Aftertaste 3.4 4.0 0.85
F8 Sourness 6.9 4.3 1.59

For most dimensions, the variance explained is mainly due to product differences and not to
subject differences.
SPISE2009 13

Regression of liking on the factor scores (with backward deletion) shows that liking can be
explained by seven of the eight factors (multiple correlation coefficient 0,63). Figure 7 shows
the standardised beta weights.
standardized beta weights

F1 Fruity/fresh

F6 Smooth/mild

F2 Fruit aspects

F3 Thickness

F4 Odour aspects

F7 Aftertaste

F5 Bitter/off taste

-0,50 -0,40 -0,30 -0,20 -0,10 0,00 0,10 0,20 0,30 0,40 0,50

Figure 6: Standardised regression weights (standardised beta) for the significant factors (blue bars
contribute positively to liking and red bars negatively).

attribute ß weight
fresh taste
freshness fruit
fruity taste
mild taste
smooth in mouth
firmness fruit
creamy taste
taste int.
sweetness
airy mouth feel
amount fruit mf
fruity odour
thick in mouth
amount fruit
odour int.
recogn. fruit
gloss
thick app.
watery/thick
sweet odour
int. aftert.
length aftert.
sour odour
colour int.
sourness
off taste
bitterness
astringent

‐0,50 ‐0,40 ‐0,30 ‐0,20 ‐0,10 0,00 0,10 0,20 0,30 0,40 0,50

Figure 7: Attribute beta weights (standardised).

SPISE2009 14

To obtain the “attribute beta weights,” the factor loadings for each attribute within a factor
have been multiplied with the standardised beta weight and the resulting values have been
summed for each attribute. Figure 7 shows these values.

The values indicate the relative contribution to overall liking for that specific attribute. Here,
it appears that the “fresh” attributes have the largest positive contribution to overall liking. On
the other side, astringent is equally important, but contributes strongly to disliking (strong
negative contribution). In the final step, the deviations from ideal are weighted by these
attribute betas and the impact on liking when the specific attribute would be ideal is
computed.

5. OPTIMISATION
For each attribute, the relative effect on overall liking is computed provided that this attribute
would be rated ideal (the assumption is that liking will be rated 9 on the 9-point scale when
the product is ideal).
The final result is presented as a fishbone plot in which the potential increase in overall liking
and the deviations from ideal are shown. Only attributes which show a relative improvement
of 2% or more are shown in the plot, the attributes are ordered according to the improvement
(attributes on the right have the highest impact).

Figure 8 shows the results for Fruit mix. The yellow bars show the potential increase in liking
if that attribute would be ideal (in percentages, left axis) and the red diamonds show the
absolute deviations from ideal (right axis).

Fruit mix deviation f rom ideal (+ needs more, - needs less)

10% 20

15

5% 10

0% 0
f ruity odour
f ruity taste

f reshness f ruit
recogn. f ruit
creamy taste

# f ruit
# f ruit mf

-5

-5% -10

-15

-10% -20

Figure 8: Optimisation or fishbone plot for Fruit mix.

Since Fruit mix already has a high liking rating (7.1) there is little room for improvement.
Increasing the freshness and making the fruit more recognisable and increasing amount of
fruit will increase liking (Note that the increase in liking is not directly related to the amount
of change: for recognisable fruit, the deviation is larger than for freshness Fruit, but the
possible gain in liking for freshness is higher than that for recognisable Fruit, not every
deviation is equally important).
SPISE2009 15

For the Dades variant more improvements are needed. As can be seen on the right side of the
graph, increasing fresh and fruity taste and decreasing astringency have a large effect on
liking. Increasing creamy and smooth mouth feel and decreasing bitterness and off taste also
will improve liking.

Dades deviation f rom ideal (+ needs more, - needs less)

10% 20

15

5% 10

0% 0

of f taste

f ruity odour
mild taste
f irmness f ruit

recogn. f ruit

f ruity taste
airy mouth f eel

creamy taste

f resh taste
smooth in mouth
odour int.

f reshness f ruit
bitterness

astringent
amount f ruit mf
sweetness

-5

-5% -10

-15

-10% -20

Figure 9: Optimisation or fishbone plot for Dades.

6. DISCUSSION AND CONCLUSIONS

Compared to the classical procedure (expert panel for the sensory profile and consumers for
liking) the IP method is faster and more efficient because all information is obtained in one
test. The final results are the fishbone plots which show for each attribute how much it
deviates from ideal and the relative effect on liking when it would be made ideal. In contrast
to expert panels, consumers do not use a specific language and cannot tell R&D which
ingredient is missing. The final translation has to be made by the expert. For the translation of
the results, a knowledgeable researcher is indispensible. Identical conclusions with respect to
all methods of product improvement were reached by van Trijp et al. (2007): “…….the
results therefore show that statistical criteria only are an inadequate basis to decide on
product adjustment or not and that this criterion needs to be complemented by managerial
judgment as to whether the deviation from ideal is large enough to justify product
adjustment..” and with respect to JAR scaling by Popper (2004): “….Nevertheless, JAR scales
in the hands of knowledgeable researchers and along with the appropriate analysis can do a
just-about-right job ….”.
Compared to the JAR procedure, flash profiling and napping, the IP method provides more
detailed information. It shows not only how the different products are perceived (the sensory
profiles), but also the ideal profiles and it shows estimates of the effect on overall liking when
an attribute would become ideal.

REFERENCES
ASTM-E18. WK371 New guide for use, benefits and risks associated with just about right scales in
sensory evaluation testing with consumers (E-18.04.26 wk 371). Committee E18 on Sensory
Evaluation.
SPISE2009 16

Delarue, J.& Sieffermann, J-M. (2004). Sensory mapping using Flash profile. Comparison with a
conventional descriptive method for the evaluation of the flavour of fruit dairy products; Food
Quality and Preference, 15, 383-392.
Engel, J.E., Blackwell, R.D., & Miniard, P.W. (1995). Consumer Behavior. The Dryden Press.
Harcourt Brace College Publishers, 8 Edition.
Epler, S., Chambers, E., & Kemp, K.E. (1997). Hedonic Scales are Better Predictor than Just About
Right Scales of Optimal Sweetness in Lemonade. Journal Of Sensory Studies, 13, 191-197.
Greenhoff, K. & MacFie, H.J.H. (1999). Preference mapping in practice. In Measurement of Food
Preferences. Editors: H.J.H. MacFie and D.M.H. Thomson.
Husson, F., Le Dien, S. & Pagès, J. (2001). Which value can be granted to sensory profiles given by
consumers? Methodology and results. Food Quality and Preference, 12, 291-296
Lawless, H.T. & Heymann, H. (1999). Sensory evaluation of food: Principles and practices. New
York: Kluwer Academic/Plenum Publishers.
MacFie, H.J., Bratchell, N., Greenhoff, K. & Vallis, L.V. (1989). Designs to balance the effect of
order of presentation and first-order carry-over effects in hall tests; Journal of Sensory Studies, 4,
129-148.
Meilgaard, M.C., Civille, G.V., & Carr, B.T. (2007). Sensory Evaluation Techniques. Fourth Edition.
CRC Press. Boca Raton, FL.
Moskowitz, H.R. (1996); Expert versus consumers: a comparison. Journal of Sensory Studies, 11, 19-
37.
Moskowitz, H. R. (2004). Just about right (JAR) directionality and the wandering sensory unit. In data
analysis workshop: getting most out of just-about-right data. Food Quality and Preference, 15,
891–899.
Pagès, J. (2005). Collection and analysis of perceived product inter-distances using multiple factor
analysis: Application to the study of 10 white wines from the Loire Valley; Food Quality and
Preference, 16, 642-649
Popper, R. (2004). Workshop summary getting the most out of JAR data, Food Quality and
Preference, 891-999.
Popper, R. & Kroll, D. R. (2005). Just-about-right scales in consumer research. Chemo Sense, 7, 1–6.
Popper, R., Rosenstock, W., Schraidt, M., & Kroll, B. J. (2004). The effect of attribute questions on
overall liking ratings. Food Quality and Preference, 15, 853–858.
Stone, H. & Sidel, J.K. (2004). Sensory evaluation practices; Academic Press.
Trijp van, H.C.M., Punter, P.H., Mickartz, F., & Kruithof, L. (2007). The quest for the ideal product:
Comparing different methods and approaches; Food Quality and Preference, 18, 729-740.
Vigneau, E., Qannari E. M., Punter P. H., & Knoops S. (2001). Segmentation of a panel of consumers
using clustering of variables around latent directions of preference. Food Quality and Preference,
12, 359-363.
Worch, T., Lê, S., & Punter P.H. (2009). How reliable are the consumers? Comparison of sensory
profiles from consumers and experts; Food Quality and Preference, In Press.
SPISE2009 17

APPENDIX
The average intensity and ideal rating are shown in table 4 and 5
Table 4: Average intensity ratings for the different products:
Raspberr Fruits Cherry/ Mango Orange Mandarin Maracuja Dades
y/Peach Orange
gloss 61 64 61 63 64 62 63 58
colour int. 42 35 62 44 36 40 38 49
amount fruit 54 48 50 44 43 56 40 54
recogn. fruit 43 44 47 40 42 57 34 39
thick app. 58 57 57 60 59 60 59 62
odour int. 48 51 52 46 52 56 46 45
fruity odour 45 47 51 42 51 56 43 40
sweet odour 45 46 44 40 44 45 42 42
sour odour 40 40 42 41 40 42 41 39
taste int. 56 63 61 53 57 54 62 55
fruity taste 53 59 61 47 54 51 61 48
freshness fruit 54 58 56 49 53 54 56 49
sweetness 51 59 54 43 49 38 54 43
sourness 40 38 38 43 42 51 38 41
bitterness 26 23 27 30 27 30 27 35
astringent 30 27 30 37 30 34 30 39
creamy taste 52 54 49 50 54 47 55 47
mild taste 51 52 50 47 52 48 53 48
fresh taste 57 61 60 52 62 59 60 48
off taste 36 39 40 40 35 38 38 45
thick in mouth 52 54 52 53 54 55 55 56
smooth in mouth 55 60 55 53 57 54 55 48
watery/thick 53 53 52 55 54 52 55 56
airy mouth feel 50 52 52 47 51 50 51 45
firmness fruit 57 65 51 58 46 43 56 52
amount fruit mf 50 53 49 42 42 49 41 50
int. aftert. 50 56 54 48 51 52 57 53
length aftert. 48 55 52 48 50 51 55 51

Table 5: Average ideal ratings for the different products.

Raspber Fruits Cherry/O Mango Orange Mandari Maracuj Dades
ry/Peach range n a
ID gloss 61 61 61 61 62 62 62 60
ID colour 48 41 59 45 43 44 42 47
ID amount fruit app. 64 64 63 61 61 64 60 62
ID recogn fruit 60 61 61 59 60 64 57 58
ID thick app 61 60 61 60 61 61 60 60
ID odour int 59 58 59 56 60 61 56 57
ID fruity odour 60 61 61 59 61 63 58 57
ID sweet odour 49 50 49 49 49 49 49 50
ID sour odour 38 39 40 37 39 39 37 37
ID taste 62 64 64 59 62 61 63 58
ID fruity taste 64 65 65 62 63 63 64 60
ID freshness fruit 65 67 66 64 64 65 65 61
ID sweet taste 51 54 53 50 52 50 53 50
ID sour taste 38 39 39 38 39 42 39 37
ID bitterness 25 24 25 26 25 26 25 27
SPISE2009 18

ID astringent 26 25 26 28 26 28 25 29
ID creamy taste 62 62 60 62 62 61 62 61
ID mild taste 56 56 56 53 57 55 56 55
ID fresh taste 63 66 64 62 66 65 65 61
ID off taste 32 34 36 33 32 32 33 34
ID thick mf 59 59 58 58 58 59 59 58
ID smooth mf 59 61 59 59 61 60 59 59
ID watery/thick 58 58 57 58 59 59 58 58
ID airy mf 54 54 54 53 55 53 54 53
ID firmness fruit 63 67 61 62 59 57 61 58
ID amount fruit mf 63 65 64 62 62 64 62 60
ID int aftert. 53 55 54 51 53 53 54 52
ID aftert 51 53 52 50 52 51 52 51
SPISE2009 19

CONSUMER PREFERENCES FOR VISUALLY PRESENTED MEALS

G. Gabrielsen*a, M.D.Aaslyng b, and H.H Reisfelt c
(a) Copenhagen Business School, Denmark,
(b) Danish Meat Research Institute, Denmark,
(c) Department of Food Science University of Copenhagen
*Corresponding author: [email protected], Tel: +45 38153512

Abstract
The aim of the present study was to develop methods to model preferences at the individual
or even contextuallevel. Such model may increase the explanatory power of the data and
thereby be a stronger tool for understanding the consumer variation, the contextual variation
and the cultural variation. A better understanding of these variations may replace the concept
of “best product” by a portfolio of good products and give input to new product development.
As the study object we applied ready meals, the appearance of such products being very
important for our choices of convenient food products. A conjoint layout using 32 pictures as
a 25 design was constructed: Collation (modern/traditional), Meat cutting (medallions or
pieces), Vegetables (wok or root fruits), Herbs (with or without), and Sauce (with or without).
A Total of 880 consumers were interviewed in three different towns in Denmark and at two
different stores. The method of scoring the pictures makes it possible to analyse the results as
an analysis of variance (ANOVA). The statistical analyses showed that is it possible to estimate
the variability of preferences according to the five different aspects in the conjoint layout and
furthermore to estimate trade-offs between different aspects. Finally, the analyses indicated
the amount and type of products in a product portfolio.

Keywords: conjoint layout, analysis of variance, multilevel models, product portfolio

1. INTRODUCTION
The overall aim of the study was to investigate consumer preference for the appearance of a
meal in a situation of buying convenience food. These products are sold at a rather low and
stagnant rate in Denmark compared to, for example, the rest of Europe where they are sold at
a high and increasing rate. The meals can be targeted at various consumer groups.
When we buy a ready meal, the appearance is very important for our choice. However, the
knowledge of how the plates should be arranged is scarce and mainly empirical. The preferred
appearance may depend on geographical location, gender, age, and many other factors. This
knowledge can be used in product development and in the design of the meal according to
consumer groups: The busy high income family who do not have time to cook themselves, but
still want some well tasting food; the singles who do not want to cook for themselves; the
family with children who do not have time to cook in everyday life etc.
SPISE2009 20

However, often it is in practice impossible to connect preferences to segments and one has to
consider the distribution of preferences in the targeted market to decide the characteristic of
products, price level of products and marketing strategies.

Visually based preferences

Visual surveys are an inexpensive and quick method for assessing preferences and the
potential of these surveys have expanded with recent computer and visually based techniques
in fields such as consumer preference research, convenience meal development, package
labelling, and advertising.
Visual surveys are increasingly being applied to methods such as discrete choice analysis
(Louviere, Hensher, Swait, & Adamowicz, 2000; Train, 2003), which is used to predict
preferences for goods or services according to choices made between distinct alternatives with
varying attributes.
A conjoint design can be used to determine the value placed on attributes by having
respondents choose between strategically included alternatives characterised by a limited
number of different attributes, each at levels, which are evenly distributed between the
alternatives (i.e., orthogonal design). In conjoint design, it is assumed that choice is governed
by the maximisation of preferences for the attributes (Moskowitz & Silcher, 2006). The use of
conjoint design in a discrete choice analysis makes it possible to determine the value placed
on attributes through decomposition of the choices and trade-offs made between alternatives.
The validity and discriminatory power of using visual surveys instead of real products has
previously been established for conjoint design in the sense that there is no difference
between preferences of food products determined by choices made between pictures versus
tangible food products (Jaeger, Hedderley, & MacFie, 2001). However, it is important that
alternatives investigated are realistic. Consequently, meal attributes should be well known,
well-liked, properly organised, and appear palatable to respondents. Despite this, most visual
surveys investigate relatively few distinct attributes. A visual survey based on multiple
combinations of attributes might vastly reduce effort and costs associated with investigating
various attributes. Furthermore, visually-based choices can be rapidly made and used to infer
preferences and thereby help improve food product development and institutional food
services. In the present study, the meals seemed to closely resemble each other, but were in
fact strategically organised with varying attributes as based on a conjoint design
Since visual working memory has a limited capacity and the perceptual workload of our task
was high, it was deemed very unlikely that the results would be based on either short-term
working memory or long-term visual memory between sessions. Rather, choices were more
likely to be based on subjects’ previous meal experience (Humphreys & Bruce, 1991; Repovs
& Baddeley, 2006). Hence, we investigated the reliability of the method by conducting the
experiment twice.
SPISE2009 21

2. MATERIALS AND METHODS

The investigation was carried out by means of a computerised questionnaire in shopping
centres in three middle-sized Danish towns: Amager, a suburban part of the Danish capital,
Copenhagen; Naestved, a town situated on the isle of Zealand about 75 km from Copenhagen;
and Kolding, situated on the peninsula of Jutland about 225 km from the Danish capital.
The sample comprised 880 respondents, 542 women and 338 men, who completed the
questionnaire. Respondents were recruited in shopping centres on location, adjacent to
supermarkets of interest for the investigator Bilka a relatively large and cheap supermarket
(three stores), and “Irma,” a small supermarket with a reputation for high-quality food (two
stores).

The program was written in microsoft access and functioned interactively, forcing the
respondents to pick their choice to any question posed by pressing the key marker on one of
the displayed response options. Consequently, the respondents were allowed progression onto
the next question that appeared on the interface once the former question had been answered
and registered in the computer database.
The program was divided into a pictorial part, which was followed by the regular
questionnaire that included a total of 26 questions that focused mainly on geographic,
demographic, sociological and lifestyle variables. The 32 photos were divided into four series,
and each series of eight photos were displayed on the screen. The series each consisted of
eight unique pictures of randomised food dishes. For each screen display, respondents were
instructed to pick their most preferred choice, second most preferred choice and least
preferred choice, respectively, from eight pictorial variations of a meal presented in
randomised series (blocks). Having pressed the key marker, the most preferred image was
removed from the display, and similarly, for the second most preferred choice. When the
respondent had selected the least preferred image of the displayed series, the sequence was
terminated, and the screen display shifted with emergence of a new series of another eight
photos until all of the series had been displayed.
Figure 1 displays the first of four series that respondents were presented within the program.
The screen display thus consisted of a 3  3 array in which the questions were posed in the
centre square. In order to obtain an indication of respondents’ degree of overall liking of the
visually presented stimuli, they were asked to rate two images that were diametrically
opposed to each other with respect to variations of the component. We chose the first photo
randomly from a pile of images and subsequently paired it with its opposite. In this way the
majority of meal components used in the experiment was subjected to a hedonic evaluation.
We used a hedonic 5-point scale, with values ranking from “like very much the look of the
meal” to “don’t like the look of the meal at all”.
SPISE2009 22

Figure 1: One of four series consisting of eight photos.

2.1 Meal Components

The five meal components and their corresponding levels (in brackets) were defined in the
following way:
Dish: modern or traditional level; this component denotes the mode of presentation.
Vegetable mix: root mix or wok mix.
Meat: slices or pieces of tenderloin pork; this component represents the carving of the
meat.
Sauce: dishes with or without sauce.
Herbs: dishes with or without parsley.
When selecting the stimuli we were cautious to meet popular Danish choices of meat,
vegetable mixes, and sauce, this insured that the stimuli reflected Danish cultural aspects.
Moreover, the energy contents and constitution of macronutrients of the meals largely were in
accordance with guidelines from Nordic Nutrition Recommendations, NORDEN 2004.
Three base ingredients were used for all photos: tenderloin pork, potatoes (either white and
cooked, or fried, brown boats) and vegetables.
Dish was presented in a modern and traditional collation, respectively. The former was
defined as a dish with fried potato boats, and glasses were used if the component sauce was
present. For the component variant meat in pieces, the meat pieces were arranged on spears
SPISE2009 23

and whole leaves of parsley were used. The traditional dish was arranged with cooked white
potatoes and the sauce was arranged in a small puddle, and the parsley had been chopped and
poured over the potatoes. Two different vegetable mixes were used; root mix and wok mix.

Regarding meat, the pork tenderloins were cut in 1-cm-thick pieces and weighed 50 g on
average. The smaller pieces of meat were approximately 1  1 cm in size, weighing 10 g on
average. The slices were fried for 4 min. and the pieces for 3 min, respectively. Averages of
10 g of virgin olive oil were used for each pan. The meal components, sauce (whisky sauce)
and herbs (parsley), varied in the series by being either present or absent. The dishes were
arranged placing the vegetables on the plate approximately between 6:00 and 10:00 a.m., the
potatoes approximately between p.m., and the meat covering the rest of the plate.

2.2 Experimental design

The applied design was a 25 conjoint layout combining a balanced block design and an
incomplete ranking test. In this factorial design, 25 = 32 photos were allocated into four
blocks (fractionals, series) of eight, so that each level of each of the five meal components
appeared equally often on the screen, and at various positions. The same layout was repeated
for all respondents. Thus, the main and two-way effects were balanced with respect to the
block, thereby enabling estimation of the main effects and two-way interaction effects of the
meal components, irrespectively of the block on the individual level as well as on the
aggregate level. The variable position described the position of the items in the 3  3 array
used for all blocks.

2.3 Data Analyses

The study presented a conjoint layout and was analysed by analysis of variance (ANOVA). The
position of the photos on the screen influenced respondents’ choices and the effect has been
corrected for as “design” variables. Likewise, respondents were shown only eight photos at a
time, which resulted in a series effect that has also been corrected for. Age and gender were
regarded most important variables, and effects related to these variables have been corrected
for in the analyses of data as categorical variables (factors) in the analysis.
The five treatment factors (meal components) may be represented as dummies XC ~ the factor
dish, 0 = traditional and 1 = modem; XV ~ vegetable mix, 0 = wok mix of vegetables and 1 =
root mix of vegetables; XM ~ meat. 0 = tenderloin meat in pieces and 1 = tenderloin slices; XS
~ sauce, 0 = without sauce and 1 = with sauce; XH ~ herb, 0 = without herbs and 1 = with
herbs.
The 32 photos consisted of the 32 combinations of the five factors and may be indexed by i, j,
k, l, m. We defined a factor Block (B: b = 1,2,3,4) allocating the 32 photos to the four blocks b
= b(i, j, k, l, m) indicating to which of the four blocks a photo belonged.
Dijklm was a dummy indicating the position of the photo (on the screen): 0 = lower right
comer, 1 = upper left comer, including the diagonal. For convenience, position has been
SPISE2009 24

reduced to a two-level “design” factor (level 1 ~ position 1, 2, 3, 4 and 6; and level 2 ~

position 5, 7 and 8) because more than 95% of the variation between the eight positions was
explained by these two levels. As such, the core of the analysis of variance modelwith only
main actionslooked like this:
E(Yijklms) = μ + γbB + δDijklm + βCXiC + βVXjV + βMXkM + βSXlS + βHXmH
in which E(Yijklms) denoted the expected response of subject s to photo i, j, k, l, m corrected for
the effect of photos displayed and the effect of the block. The 32 photos were the explanatory
variables/factors. To fit a model to all the data, each of the 32 photos must have a response.
Thus, the method used had to take into account the photos not chosen, representative of
“missing values” in this context.

The values for the hedonic scores within a block were: 2, 1, 0 and –1, where 2 represented the
score of the most preferred choice, 1 represented the score of the second most preferred
choice and –1 represented the score of the least preferred choice. Finally, 0 represented the
score of each of the five images of a block with no mutual ranking. To calculate the main
effect of a factor, the difference between the means of the 16 scores on each of the two factor
levels was used as the sufficient statistic. As the blocks are orthogonal to the (treatment)
factors, this calculation could be performed across blocks.

Because of the central limit theorem, therefore, the sufficient statistics for each subject are
approximately normal. Calculating means across subjects would make the approximation to
normality even better. Furthermore, the sum of scores for a specific factor level is the fraction
of times this level is present in the most preferred photo with weight 2 plus the fraction of
times this level is present in the second most preferred photo with weight 1 plus the fraction
of times this level is present in the least preferred photo with weight 1. This means that
although the scores are (incomplete) ranks, the sufficient statistics are means of weighted
ranks (or smoothed ranks). By using the ranks as scores one theoretical difficulty arose: for
each subject and each block the rating or scores of the eight photos comprising the block are
given by the numbers 2, 1, 0, 0, 0, 0, 0, –1. Thus, the mean of the scores in each block is 0.25
and thereby the mean of the 32 scores given by a subject is also 0.25 (and the grand mean of
the 32 scores given by the 768 subjects is also 0.25).

Furthermore, for each subject and each block the SSQ is (2 – 0.25)2 + (1 – 0.25)2 + 5  (0 –
0.25)2 + (–1 – 0.25)2 = 5.5. For each subject, therefore, we have the total SSQ = 4  5.5 = 22.
This means that the grand mean is fixed (and no degrees of freedom are used to estimate it).
Furthermore, the total SSQ is fixed being number of subjects  22 = 768  22 = 16,896. Thus,
the data can be considered as standardised, having for each respondent the same mean (equal
to 0.25) and the same SSQ (equal to 22). This implied that from a theoretical point of view the
usual F-test in the ANOVA is not valid. However, the numerator is still chi-square distributed
and the usual F-test should have been replaced by a chi-square test. As the number of degrees
SPISE2009 25

of freedom in the denominator in the present case is large, the chi-square test and the F-test
are approximately equal and therefore. We use the F-test as usual.

3. RESULTS AND DISCUSSION

Data can be analysed at different levels. Table 2 shows the result of the analyses of variance
at the aggregated level, assuming, that respondents are randomly drawn from a homogeneous
population. The most important meal component is Dish and in general the modern collation
is preferred. This means that the “mean respondent” prefers modern collation just a little bit
over traditional collation, however, the variation of preferences is very high and therefore
very few respondents are close to “the mean consumer.” The meal component Dish account
for only 6 % of the total variation of the preferences. If the other four meal components are
included in the analyses the explanatory power increases to R2 = 7.2 %. Including all order
interactions makes the explanatory power increases to 8.3 %, Table 2.
The remaining variation, 91.7 %, is what is sometimes called the biological variation –
variation of consumer preferences – and with a remaining biological variation as big 91.7 % it
has hardly any meaning to talk about a mean consumer.
Table 2: Main effects – Aggregated level.
Source of Variation Most preferred Sum of Squares p
Main effects
Collation Modern collation 997.1 < 0.0001
Vegetables Wok-like vegetables 21.7 < 0.0001
Meat Meat in Pieces 11.3 < 0.0001
Sauce With sauce 27.0 < 0.0001
Herbs With herbs 133.5 < 0.0001
R2 = 7.2 %
Main effects + all order interactions
R2 = 8.3 %

To reduce the biological variation we can try to segment the consumers according to, for
example, gender, age, etc. The idea of segmenting is that the consumers are clustered into
“groups” (i.e., segments) and that within segments the consumers are considered as a random
sample from a homogeneous population and the variation is due to differences between
segments. The effect of segmenting after different factors is shown in Table 3. It is seen that
segmenting the consumers increase the explanatory power of the model to at most 10 % of the
total variation.
Table 3: Main effects – Segmented.
Model R2
Main effects at aggregate level 7.2 %
Main effects segmented by gender 7.5 %
Main effects segmented by gender and age 8.7 %
Main effects segmented by gender and age and geography 8.8 %
Main effects segmented by gender and age and geography and … ~ 9.5 %
SPISE2009 26

At the individual level we can estimate the main effects of the meal component of each
individual. As the number of respondents is high we can illustrate the distribution of each of
the main effects - illustrating the unexplained individual variation. The explanatory power of
main effects estimated at the individual level is R2 = 45.8 %.

Effects of Dish Effects of Meat

0
Traditionel Modern Pieces Medalion

Effects of Herbs Effects of Vegetables

No herbs Herbs Wok-like Root fruits

Effects of Sauce

No sauce Sauce
Figure 4: Main effects of meal components.

Figure 4 shows the distribution of the estimated main effects for the five meal components
Dish, Meat, Herbs, Vegetables and Sauce. The variability of all of the meal components is
large: it does not seem possible to produce a ready meal which satisfies all consumer
preference concerning the five meal components. The meal component with the most
interesting distribution of preferences is Dish, where many consumers prefer Modern
collation; however, there is a long tail of preference towards Traditional. The preference of
meat is concentrated in the middle of the distribution meaning that most people have no
preferences concerning the meat being in pieces or medallion. Also the distribution of
preference of Herbs is concentrated a little towards Herbs, and there is no outspoken
preference for any Herbs.
SPISE2009 27

Many consumers have no outspoken preferences concerning Vegetables, however, there are
consumers who prefers Wok-like and consumers who prefers Root fruits. The distribution of
the preference of Sauce is heavy tailed in both ends, meaning that there are many consumers
who prefer sauce and there are many consumers who prefer no sauce. Therefore, it is not
possible to develop a ready meal satisfying all consumers and an approach in that direction
will result in a product satisfying the “mean consumer:” “what fits everybody fits none”. One
way of approaching this problem is to segment the consumers. However, as seen from Table 3
the segmenting factors do not reduce the variations very much, which means that we have a
variation which is not explainable by the usual factors (e.g., lifestyle). A solution, therefore, is
to develop a portfolio of products each product satisfying some consumers and hope that
consumers can find the products themselves.

One way the consumers segment themselves is by the store where they choose to shop. The
distributions of preferences of Dish segmented into costumers in Bilka and IRMAS,
respectively, is shown in Figure 5, where it is seen that consumers in IRMA concerning
preferences on Dish is rather concentrated on Modern being homogeneous whereas the
costumers in Bilka is more heterogeneous.

Effects of Dish, Bilka Effects of Dish, IRMA

Traditionel Modern Traditionel Modern

Figure 5: Effect of Dish segmented after store.

4. CONCLUSION
From a methodological point of view in market research it is important to distinguish between
measurements error, market variation, and biological variation. The purpose of a market
research is to explore variations of “preferences,” and, if possible, to determine sources of
variation of preferences. The sources of variation may be related to, for example, lifestyle,
culture and family but may also be contextual: “we do not eat the same every day.” The
market research should involve both product variation and biological variation and make it
possible to measure, model and compare both types of variations. One way to achieve this is
to use a conjoint layout; it can include many objects attributes (meal components) many
consumer attributes (gender, lifestyle etc.) and by use of a reasonable design it is possible to
separate different kinds of variation. Furthermore, data can be analysed at different levels
SPISE2009 28

(individual level, family-level, neighbourhood level, country level) such that the variations
can be attributed to the right level. Often it is appropriate to determine preferences by asking
respondent to choose between objects (e.g., discrete choice, best / worth). In these cases it is
often reasonable, if possible, to replace respondent’s choices by respondent’s scores of
objects. Furthermore, it is important that the research can be performed cheap and fast. Often
it is possible to use the internet, and to determine the preference for visually presented
objects/meals or one can use Hall-test.

REFERENCES
Humphreys, G.W. & Bruce, V. (1991). Visual Cognition. Lawrence Eribaum Associates. Hove, UK:
Psychology Press.
Jaeger, S.R., Hedderiey, D., & MacFie, H.J.H. (2001). Methodological issues in conjoint analysis: a
case study. European Journal of Marketing, 35, 1217-1237.
Louviere, J.J., Hensher, D.A., Swait, J.D., & Adamowicz, W. (2000). Stated choice analysis: Analysis
and applications. Cambridge University Press, Cambridge, UK.
Moskowitz, H.R. & Silcher, M. (2006). The applications of conjoint analysis and their possible uses in
Sensometrics. Food Quality and Preference, 17, 145-165.
NORDEN. (2004). Nordic Nutrition Recommendations. Nord 2004:13. Nordic Council of Ministers,
Copenhagen.
Reisfelt, H.H., Gabrielsen, G., Aaslyng, M.D., Bjerre, M.S., & Mdler, P. (2009). Consumer
preferences for visually presented meals. Journal of Sensory Studies, 24,182-203.
Train, K. (2003). Discrete choice methods with simulation. Cambridge: Cambridge University Press.
SPISE2009 29

OBSERVATIONAL RESEARCH: A TOOL FOR COLLECTING BEHAVIORAL

DATA AND VALIDATING SURVEYS
S. L. Godwin* and E. Chambers IV
School of Agriculture and Consumer Sciences, Tennessee State University, 3500 John A.
Merritt Blvd., Nashville, TN 37209, United States of America; Sensory Analysis Center,
Kansas State University, Department of Human Nutrition, Justin Hall, Manhattan, KS,
66506. USA.
*Corresponding author: [email protected]; phone (615)963-5619
Abstract
Observational research is an underused technique that involves the direct observation of
people in a natural setting or in surroundings that closely resemble a natural situation. This
differentiates it from survey research which may be completed in almost any location.
Observational techniques can be an effective method for collecting behavioural data and for
validating information we obtain from individual surveys. It measures behaviour directly,
rather than reports of behaviour or intentions. Thus, observational research is a direct
approach to collecting data, as contrasted to a survey where people are asked what they do or
would do in different situations, and we assume that they reported actual behaviours. In our
studies we use observational research for many purposes including designing better surveys
by knowing what choices need to be put on the questionnaire. Additionally, we use “casual
observation” to assist us in developing a “formal” observation process where the observer will
know what types of behaviours to watch for and how they should be recorded. In addition,
behavioural research also allows for the flexibility of recording things that were seen but were
not expected. Examples of using observational research in recent studies such as watching for
specific food safety behaviours as people prepared a meat dish in their own kitchen and
research where observational data was used to validate diet surveys will be presented. The
reliability of the observations can be increased by training observers, practicing recording
observations and reaching agreement between the observers during the training.

Key words: observational research; survey validation

1. INTRODUCTION
Consumer research is conducted to obtain many types of information including preference,
liking, attitudinal, and behavioural information. Behavioural information, such as what
people eat, how they store or prepare food, how they use particularly products, and other
types of data often are collected with surveys.
Research methods, such as surveys, that rely on self-reported data are valuable in many
situations. However, for behavioural data, self-reported information may be biased and show
incorrect information because there can be a substantial difference between what people say
they do and what they actually do. In addition, people forget what they do, people tend to
answer with what is considered appropriate, and people tend to say what they think the
SPISE2009 30

interviewer would like to hear (Herzog, 1996). Because self-reported data essentially are
second hand information and relevant behaviour is not being observed directly by the
researcher, the accuracy of the data may be questionable (Herzog, 1996; Pike & Agnew,
1991).
Taylor-Powell and Steele (1996) state that direct observation is a valuable and underused
technique for collecting data. In addition, those authors point out that observational data can
take various forms from simple check-lists and counting to complex anthropological and
sociological methods. Moskowitz, Beckley, and Resurreccion (2006) suggested that
observational research is the most common form of research in many fields, but it only
recently has become popular in food and product research. Lee and Broderick (2007) state
that technology can revolutionise observational research through such techniques as video,
internet, tracking, and neuroimaging.

2. OBJECTIVES OF OBSERVATIONAL RESEARCH

Observational research can be used to achieve many objectives. It can be used to:
 Understand what the product or service means in the lives of consumers
 Learn how a product actually is used
 Uncover in-use motivations
 Uncover unmet needs
 Understand selection and purchase behaviour
 Document real benefits in-use
 Understand the category
 Assess actions in various situations
Observational research is the careful, planned, systematic watching, with attention to detail, of
various types of phenomena in order to learn new facts or test theories. Notice that this
definition is, in fact, virtually identical to that of any type of research.

3. OBSERVATIONAL RESEARCH ADVANTAGES

Direct observation methodology has two potential advantages over self-report and anecdotal
methods: observation captures actual behaviour and behaviour is captured in context
(Gittelsohn et al., 1997). Redmond and Griffith (2003) suggest, “Observational studies
provide a more realistic indication of the food hygiene practices actually used in domestic
food preparation.” Thus, direct observation research is believed to yield valid and reliable
information upon which to base educational efforts, or decisions regarding acceptance of new
items.

Observational data are best used when you need direct information to determine who people
are, what people actually are doing, when trying to understand on-going behaviour or process
such as how people prepare certain types of foods or interact with various product types,
when physical evidence needs to be observed such as how products turned out when they
SPISE2009 31

were made or how much food people left on the plate, and when other types of data collection
are incapable of capturing the data needed such as eating patterns or shopping behaviour
(Taylor-Powell & Steele, 1996).

Observational data is essential when people cannot accurately give information. Chambers,
Godwin, and Vecchio (2000) and Godwin and Chambers (2003) noted in studies of dietary
intake that people cannot estimate what never went into memory in the first place. Thus, in
those situations, alternative methods, including observational research, may be necessary to
track how much people actually eat.
Similarly, in a recent study on food storage habits that we conducted (unpublished data), we
asked people what the temperature was in their home refrigerator. More than half the
respondents had no idea what the temperature was and another fourth of the respondents
indicated they were guessing. In addition, many people did not have a thermometer in their
refrigerator so even asking them to check the temperature would be impossible. To obtain that
type of information, in home observation, where thermometers can be placed in the
refrigerator and temperatures tracked is necessary.

4. EXAMPLES OF STUDIES CONDUCTED USING OBSERVATIONAL

RESEARCH
Anderson et al. (2004) assessed food-handling practices during food preparation in the home
and compared them with Fight-BAC recommendations. Before conducting a similar study at
Tennessee State University (Godwin, Chen, & Kilonzo-Nthenge, 2009), we used
observational research during the preliminary phases of the study to assist with development
of structured check sheets for use during the actual research project.

The accuracy with which people report what and how much they eat during dietary recalls
was investigated by Godwin, Chambers, and Cleveland (2004) in a study comparing recall to
actual consumption. In general, people were able to recall the foods they ate, but they did not
do as well estimating how much they ate. In fact, some people gave estimates that were off
by more than 50%. In that case, observational data showed that the methods used for portion
size estimation of food were inadequate and needed further development.
Shopping behaviour was examined by Sinha and Uniyal (2005). Those authors found that
observational data provided some similar segmentation of shoppers as did attitudinal or
psychographic surveys, but additional segments were noted in the observational research.
They suggested that this type of research could augment other methods and could provide
needed practical information that could be used by retailers.

5. COLLECTION OF OBSERVATIONAL DATA

Observational data can take many forms including information about the people, the physical
environment, ways of functioning, timing, amounts, behaviours, and reactions among other
things. Some of these types of data may be able to be measured quantitatively such as
SPISE2009 32

temperatures in refrigerators, amounts of food eaten, numbers of people or objects, or time in

shopping. Other information may need to be measured semi-quantitatively, such as
approximate ages of people, relative time people spend in lines or eating a particular food on
a plate, and estimates of amounts eaten. Still other information is qualitative in nature
including observations of people's reactions to products or services, ways in which people do
things, or clarity of explanations.

A key aspect of observational research is the requirement that the measurement be observable.
While this seems like an obvious aspect, the number of studies that indicate observation of
attitudes or liking is surprising. While it is possible to make inferences about those aspects
they generally are not observed phenomena. For example, a researcher decides that asking
about liking is too direct and instead decides to watch people as they eat at a buffet. Based on
eating patterns of various people at the buffet, the researcher concludes that green salad is a
well liked food. Unfortunately, the researcher fails to consider other explanations for why
salad was selected so frequently (habits, customs, health aspects, other options at the buffet).
The researcher infers liking when, in fact, many people selected the salad not because they
liked it, but because it was not disliked and they chose it for other reasons.

6. ISSUES ASSOCIATED WITH OBSERVATION

Observational research is not without problems. As with any type of research, a clear
objective is essential. Knowing the information that is needed and careful planning together
that information is essential. Observation is only as good as the consumers that you target to
observe and the behaviours or other targeted information that is noted. Selecting the wrong set
of consumers has the same problems as with any consumer research, a lack of ability to
project the results to a larger population. Observing the wrong things results in the same
issues as asking the wrong questions in a survey, and results in the inability to meet the
objectives of the study.
Unfortunately, biases are a well-known possibility in observational research and have resulted
in it sometimes being treated as unreliable. Grimes and Shulz (2002) state that “selection bias,
information bias, and confounding are present to some degree in all observational research.”
However, those authors go on to state that such biases can be overcome and may have little
effect on the information gathered. Considering they were discussing medical research, where
the potential for incorrect information truly can be a matter of life and death, the idea that bias
is not a reason for eliminating observational research from the repertory of research
techniques is an important concept.

Biases often are thought to be personal, that is to say noting behaviours through a screen
based on personal beliefs; perception vs. reality, where the behaviour is observed to be one
thing when, in fact, it is another; or objective vs. subjective, where interpretation (e.g. motives
SPISE2009 33

for purchasing) is necessary to the observation. All of those biases are potentially real, but can
be reduced with training.

A commonly forgotten aspect of direct observation is the time required to observe the
phenomenon. Godwin et al. (2007) reported on the direct observation of food storage habits
for ready to eat foods in 210 respondent homes. To collect the observational data, five
researchers spent more than 1000 hours in travel and interview time, in addition to the time
required to analyse the large volume of data. A survey of 210 people would require much less
time, but would not result in the same quality of data.
Myllyluoma and Buck (2009) found that there was excellent inter-observer reliability when
single quantitative variables had to be assessed, but there was less reliability when noting
“multiple complex features” or when semi-quantitative estimates were reported. This may
suggest that observers should attempt to track fewer details or that more training of observers
is important.

7. TRAINING OF OBSERVERS
Careful and thorough training of the observer(s) is critical for the success of observational
research; however it can also introduce bias into the observations if not designed correctly.
Explaining the purpose of the research, and taking the observer through the issues that need to
be explored during the observation are critical components of the training process. Taylor-
Powell and Steele (1996) state that training and practice are essential for standardisation of
observation. Further, they suggest that observers must be trained to look for the same
information, record data similarly, and use similar vocabulary and meanings. Training in
observation and the use of field notes can help observers become more consistent within
themselves and among other observers. The development and use of checklists or other
simplified observational data collection tools can help to standardise observation, but the
overuse of such systems can result in losing the spontaneous, unique information that can be
provided by observational research.

8. ANALYSING AND INTERPRETING DATA

Although recognised as a valid method for obtaining data, the interpretation of the data must
be completed with some caution. Other than the obvious potential biases of people knowing
they are being watched, it is essential that the analysis of the data used unbiased procedures
whether those analyses are qualitative or qualitative.
Myllyluoma and Buck (2009) used various estimates of inter-reliability depending on the data
that was collected. Scarsellone (1998) suggested that typical reliability estimates were
inadequate and proposed using generalisability theory as an appropriate framework for
evaluation.
SPISE2009 34

9. CONCLUSIONS
Observational research provides an alternative to surveys and other types of consumer
research that can enhance the information we gather about products and services. In some
cases observational research is the only way to collect valid data. Because observational
research is subject to potential biases it is essential that training be provided to those who will
collect the data and that the analysis and interpretation of the data be considered carefully.

REFERENCES
Anderson, J.B., Shuster, T.A., Hansen, K.E., Levy, A.S., & Volk, A. (2004) A camera’s view of
consumer food-handling behaviours. J American Dietetic Association 104(2), 186-191.
Chambers, E. IV, Godwin, S.L., & Vecchio, F.A. (2000). Cognitive strategies for reporting portion
sizes using dietary recall procedures. Journal of the American Dietetic Association, 100, 891-897.
Gittelsohn, J., Shankar, A.V., West, K.P., Ram, R.M., & Gnywali, T. (1997) Estimating reactivity in
direct observation studies of health behaviors. Human Organization. 56, 182-189.
Godwin, S., & Chambers, E. IV (2003). Estimation of portion sizes by elderly respondents. Family
Economics and Nutrition Review, 15(1), 47-55.
Godwin, S., Chambers, E. IV & Cleveland, L. (2004). Accuracy of reporting dietary intake using
various portion-size aids in-person and via telephone. J American Dietetic Association 104(4),
585-594.
Godwin, S.L., Chen, F.C., Chambers, E. IV, Coppings, R., & Chambers, D.H. (2007).
Comprehensive evaluation of temperatures within home refrigerators. Food Protection Trends, 27(3):
16-21.
Godwin, S., Chen, F.C., & Kilonzo-Nthenge, A. (2009). Assessing the potential for cross-
contamination in home kitchens when preparing a meatloaf. Presented at the International
Association for Food Protection 96th Annual Meeting, Dallas, TX.
Grimes, D.A. & Shulz, K.F. (2002). Bias and causal associations in observational research. Lancet,
359(9302), 248-252.
Herzog, T. (1996). Research Methods in the Social Sciences. New York, NY: Harper Collins College
Publishers.
Lee, N. & Broderick, A.J. (2007). The past, present and future of observational research in marketing.
Qualitative Market Research: An International Journal, 10(2), 121 - 129.
Moskowitz, H.R., Beckley, J.H., & Resurreccion, A.V.A. (2006) Sensory and consumer
research in food product design and development. Boston, MA: Wiley Blackwell.
Myllyluoma, J. & Buck, D. (2009) Measuring Quality in Observational Data Collection. Presented at
the American Association For Public Opinion Association, Miami Beach, FL, accessed:
http://www.allacademic.com/meta/p17012_index.html
Pyke, S.W. & Agnew, N.M.. (1991). The Science Game: An Introduction to Research in the Social
Sciences. 5th ed. Englewood Cliffs, NJ: Prentice Hall.
SPISE2009 35

Redmond, E.C. & Griffith, C.J. (2003) Consumer food handling in the home: a review of food safety
studies. J Food Protection. 66, 130-161.
Scarsellone, J.M. (1998). Analysis of observational data in speech and language research using
generalizability theory. J. Speech Language Hearing Res. 41, 1341-1347.
Sinha, P.K. & Uniyal, D.P. (2005). Using observational research for behavioural segmentation of
shoppers. J Retailing Consumer Services. 12(1), 35-48.
Taylor-Powell, E. & Steele, S. (1996) Collecting evaluation data: Direct observation. Publication G-
3658-5, Madison, WI: Wisconsin Cooperative Extension Publications.
SPISE2009 36

Part 2: Food choice and consumer behaviour studies 
SPISE2009 37

HEDONIC RESPONSE TO THE TASTES OF WINE IN VIETNAM:

DOES THE REGION OF ORIGIN OF CONSUMERS MATTER?
V.B. Do a*, B. Patrisa, T. D. Ha b, D.H Nguyenc and D. Valentina
(a) UMR 5170 CSG CNRS INRA-UB, France.
(b) Hanoi University of Technology, Hanoi, Vietnam
(c) Hochiminh-city University of Technology, Hochiminh city, Vietnam.
*Corresponding author: [email protected]. Tél: 33 (0)3 80 68 16 30

Abstract
Recent surveys showed that the consumption per capita of sugar in Vietnam is higher in the
South than in the North. Do these different food habits lead to differences in liking for a new
food or beverage? To address this question, a consumer test was conducted in Hochiminh city
in the South and Hanoi in the North of Vietnam. In the two cities, consumers evaluated three
series of red wine. In the first series, fructose was added to a base wine to manipulate
sweetness. In the second series, sourness was manipulated by adding tartaric acid into the
base wine. In the last series, bitterness and sweetness were manipulated by adding quinine and
fructose into the base wine. Results suggest that “frequency of consumption” matters more
than “city of origin.” Our initial hypothesis that people from Hochiminh city (South) might
like wine to be sweeter than people from Hanoi (North) did not hold. In contrast, our results
suggest that bitterness liking might be somewhat affected by city of origin and gender.
However further work is needed to confirm this last point.
Keywords: Vietnam, tastes, food habits, wine, liking.

1. INTRODUCTION
Consumer’s hedonic responses to tastes depend generally on environment except for
sweetness for which the liking may be innate (Beauchamp, Cowart, & Schmidt, 1991).
Environment concerns notably the effect of exposure and the familiarity of the tastes (e.g. ;
Bertino & Chan, 1986; Moskowitz et al., 1975; Prescott et al., 1992; Prescott & Khu, 1995;
Laing et al., 1994). For instance, Japanese consumers tend to give higher hedonic responses to
umami taste than Australians ones do, because this taste is much more familiar in Japan than
in Australia (Prescott et al., 1992). Differences in hedonic responses seem thus to exist
between consumers having differential food habits.

Consumers from the South and the North of Vietnam appear to have differential food habits
(Figure 1). For example, surveys of the Centre for International Economics of Canberra and
Sydney (2003) and General Statistics Office of Vietnam (2006) reported that the consumption
per capita of sugar in the South of Vietnam (including Mekong river delta and the South-East)
was higher than in the North of Vietnam (including the Red river delta, the North-East and the
North-West). Yet, there is little scientific understanding of how hedonic responses to a taste
component might vary across these two regions. The answer to this question is important for
SPISE2009 38

professionals who want to formulate a food for Vietnamese consumers from both the South
and the North, especially for consumers from the biggest cities of these two regions:
Hochiminh city and Hanoi.

Figure 1: Average consumption of sugar by ecological regions in g/capita/day (General Statistics

Office of Vietnam, 2006).

As sugar consumption is higher in Hochiminh than in Hanoi, consumers from Hochiminh are
more exposed to sweetness than consumers from Hanoi. Does that lead to a differential liking
for sweet foods? It is difficult to address this question within the context of a familiar food for
which people have already strong existing expectations. For this reason, our study was
performed with wine, a non-familiar drink in Vietnam (cf. Do, Patris, & Valentin, 2007). We
hypothesise that consumers from Hochiminh city appreciate sweet wines better than
consumers from Hanoi. To verify this hypothesis, two experiments were carried out. The first
experiment permitted us to select a base wine. In the second experiment, we added sweet,
sour and bitter tastants into the base wine to evaluate the difference of appreciation between
consumers from Hochiminh city and Hanoi.

2. EXPERIMENT 1: SELECTION OF THE BASE WINE

The objective of this experiment was to select a base wine having little overall in-mouth
intensity.

2.1 Material and methods

Participants. Nine staff members from the national university of Hanoi (four women, five
men, average age: 31.3) were trained during three two-hour sessions. During the training
sessions, the panellists performed the following tasks:
 Identify sweetness, sourness, bitterness, astringency alcohol and overall in-mouth taste
in water and wine mixtures.
 Rank the perceived intensity of the solutions and wine samples for each of the above
mentioned sensory attributes.
SPISE2009 39

 Rate the sweetness, sourness, bitterness, astringency and alcohol of wine samples
using a six-point scale (0 = absence of the sensation, 1 = very weak, 2 = weak; 3 = in
the middle, 4 = strong and 5 = very strong).

Stimuli. We first chose two Vietnamese wines that we judged, a priori, having a neutral
overall in-mouth taste. They were both red wines from the Dalat wine company, Vietnam.
The first wine was made exclusively from cardinal grapes. Cardinal is a variety of table grape
that Vietnamese producers often use to produce wine. The second wine was made exclusively
from syrah grapes. Syrah is a worldwide popular variety of grape which has been recently
imported in Vietnam from Australia. Both wines were produced using the same technology.
Procedure. During the final evaluation, the nine trained panellists rated respectively the
sweetness, the sourness, the bitterness, the astringency and the alcoholic sensation of each
candidate wine using six-point scales going from 0 (absence of the sensation) to 5 (very
strong).

2.2 Results and discussion

The mean intensity rating of each sensory attribute (sweetness, sourness, bitterness,
astringency and alcohol sensation) is shown Figure 2. Five bilateral student t-tests were
conducted to compare the intensity of the two wines in sweetness, sourness, bitterness,
astringency and alcohol sensation.
Cardinal red wine Syrah red wine

5.00

4.00

3.00

2.00

1.00

0.00
Sweet Sour Bitter Astringent Alcohol

Figure 2: Mean intensity of each sensory attribute obtained from the rating given by nine panellists.
Bars represent confidence interval at 95%

Results showed a significant difference between the two wines for sweetness, t(8) = –4.6,
p<.01), for bitterness, t(8) = –2.83, p < .05) and for astringency, t(8) = –2.68, p < .05) but no
significant difference for sourness and alcohol sensation. The cardinal wine was judged less
sweet, less bitter and less astringent than the syrah wine. By consequence, we selected the
cardinal wine as the base wine for experiment 2. This base wine contained no residual sugar,
3.98 g/l tartaric acid, and 11.8 v/v ethanol, according to information given by its producer.
SPISE2009 40

3. EXPERIMENT 2: CONSUMER TEST

3.1 Material and methods
Participants. The experiment was conducted in Hochiminh city in the South and Hanoi in the
North of Vietnam. A total of 211 consumers were recruited in the two cities; all of them were
volunteers. The recruitment was performed using the snowball method. First, five frequent
wine consumers and five non-frequent wine consumers from different ages were recruited in
each city. The five frequent consumers declared drinking wine more than two times per
month and the five non-frequent consumers up to two times per month. Then these ten
volunteers from Hochiminh city and ten volunteers from Hanoi were in charge of asking
people of their age at their work places to participate to the experiment. In both cities, the
proportion of men and women was roughly equal. Details about participants can be found in
Table 1.
Table 1: Characteristics of the consumer test participants in Ho Chi Minh City (HCMC) in the south
and Hanoi in the north. Frequent consumers (FC) declared drinking wine twice per month or more;
non frequent consumers (NFC) declared drinking wine less than twice per month.

HCMC Mean Women Men N Hanoi Mean Women Men N

age age
NFC 26.5 32 30 64 NFC 29.4 24 18 42

NFC 32.2 20 24 44 FC 30.6 29 32 61

TOTAL 28.6 52 54 108 TOTAL 29.9 53 50 103

Stimuli. The stimuli were made from the base wine selected in Experiment 1. To limit bias
due to between-bottles effect, before each session of the consumer test, the base wine was
obtained by mixture of all bottles needed for the preparation. Then, fructose, tartaric acid and
quinine (Rectapur, France) were added into the base wine to obtain three series of wine. The
first series (A) included the base wine (C0) and four samples of the same wine in which four
concentrations of fructose C1 = 2 g/l and C2 = 4 g/l and C3 = 8 g/l and C4 = 16 g/l were
respectively added to manipulate sweetness. In the second series (B), sourness was
manipulated by adding tartaric acid into the base wine (C0) at four concentrations C1 = 0.3 g/l
and C2 = 0.6 g/l and C3 = 1.2 g/l and C4 = 2.4 g/l, respectively. The last series (C) included the
base wine boosted in bitterness by addition of 20 mg/l of quinine and four samples of the
same wine with sweetness manipulated like in series A. Concentration steps were determined
by preliminary triangular tests performed on ten wine consumers to ensure that the difference
between two consecutive concentrations was perceived by more than 70% of the participants.
The preparation was carried out 24 h before the consumer test. Prepared samples were stored
at 21°C, temperature of the tasting room.
SPISE2009 41

Procedure. Consumers were requested not to smoke or eat and drink for 1 h prior to the
consumer test. The wines were served 15 min before the consumer test. Each sample
contained 20 ml of wine in a 40-ml plastic cup. During the consumer test, the 15 samples of
wine were tested. The samples of each series were presented in a randomised order. To
separate the two series of sweetened wines (A and C), the series A was presented first, then
the series B and the series C at last. The test was conducted in separate booths in 21°C air-
conditioned rooms, under the day light. Participants were requested to rate their liking for
each sample of wine using a nine-point scale going from “I totally dislike it” to “I like it very
much.” They could swallow or spit out the wines at their convenience but they had to rinse
their mouth between each sample with water and crackers.

3.2 Results and discussion

Our main hypothesis was that the appreciation of consumers from Hochiminh city and from
Hanoi differs according to the concentration in tastants. To verify this hypothesis, three two-
way ANOVAs with “city of origin” as a between-subject factor and “concentration in tastant”
as a within-subject factor were carried out, one ANOVA for each series of wine. Contrary to
our hypothesis, none of the ANOVAs showed a significant interaction “city of origin” 
“concentration in tastant” at the 5% level. The only significant effect observed at this level
was a main effect of “wines” for each series.

In the series A (Figure 3, left panel), the appreciation increased with the sweetness, F(4, 836) =
32.67, p < .001). The LSD test showed the order of the preference given to each concentration
of fructose added was C4 > C3 > C2 = C1 = C0. In the series B (Figure 3, middle panel), the
appreciation decreased with sourness, F(4, 836) = 7.53, p <.0001). The LSD test showed the order
of the preference given to each concentration of tartaric acid added was C0 = C1 = C2 = C3 >
C4. In the series C (Figure 3, right panel), the appreciation increased with the sweetness, F(4,
836) = 38.15; p < .0001). The LSD test showed the order of the preference given to each

concentration of fructose added was C4 > C3 > C2 = C1 > C0.

Liking: C4> C3> C2= C1= C0 Liking: C0= C1=C2 = C3> C4 Liking: C4> C3> C2= C1> C0
6,5 5,4
7,0

5,2
6,0
6,5
5,0

5,5 4,8 6,0

4,6
Rating

Rating

Rating

5,0 5,5
4,4

4,5 4,2 5,0

4,0
4,0 4,5

3,8

3,5 4,0
C0 =0 (Base wine) C1= 2 g/l C2= 4 g/l C3= 8 g/l C4= 16 g/l 3,6
C0= 0g/l C1= 2g/l C2= 4g/l C3= 8g/l C4= 16g/l
C0=0 (base wine) C1= 0.3g/l C2= 0.6g/l C3= 1.2g/l C4= 2.4g/l
(base wine with
Conc entration of fruc tos e added (g/l) Concentration of tartaric ac id added (g/l) 20g/l of quinine) Concentration of fructose added (g/l)

Figures 3: Effect of “wine” (p < 0.001) on hedonic responses to the series A, B and C respectively.
Bars represent confidence interval at 95%

To better understand the appreciation of the consumers we carried out a Principal Component
Analysis (PCA) with the 211 “participants” as individuals and the 15 “wines” as active
variables. The “city of origin,” “gender,” and “frequency of consumption” were projected as
 SPISE2009 42

nominal illustrative variables. Only the first two factors of the PCA having an eigen-value
higher than 1 were taken into account for interpretation (Figure 4). The first principal
component, which explains 36% of total variance, is positively correlated with all the
variables. This dimension opposes participants who tended to like the wines to participants
who tended to dislike them. This dimension seems to be linked to the frequency of
consumption. Consumers who liked the wines the best tend to be the most frequent
consumers. The second dimension, which explains 10% of the variance opposes participants
who tended to prefer the series C (wines boosted in quinine) over the series A (wines not
boosted in quinine) to participants who tended to prefer the series A over the series C. This
dimension seems to be somewhat linked with the participant’s “city of origin” and “gender”.
Participants from Hochiminh city (South) and women tended to prefer the bitter sweetened
wines (series C) over the sweetened wines (series A). Inversely, participants from Hanoi
(North) and men tended to prefer the sweetened wines (series A) over the bitter sweetened
wines (series C).

0.8

Bitter sweetened C4
Bitter sweetened C2
0.4 Bitter sweetened C1
Bitter sweetened C3
Bitter sweetened C0
Soured C2
Soured C1
Soured C0
0 Soured C1

Soured C1

Sweetened C4
Sweetened C3
-0.4 Sweetened C0
Sweetened C1
Sweetened C2

-0.8

-0.8 -0.4 0 0.4 0.8

Figure 4: Factors 1 and 2 of the PCA

4. CONCLUSION
All together our results, suggest that the “frequency of consumption” matters more than the
“city of origin” of the participants. Our initial hypothesis that participants from Hochiminh
city (South) might like wine to be sweeter than participants from Hanoi (North) is thus not
confirmed. In contrast, our results suggest that there might be a difference in bitterness liking
between Hochiminh city and Hanoi, and also between male and female consumers. However
further work is needed to confirm this trend. From an applied point of view, this experiment
suggests that when professionals formulate a wine for the Vietnamese market, they should
take into account the “frequency of consumption” of the consumers that they want to target.

Acknowledgment. We would like to express our special thank to Dalat wine company for
giving us the wines used in this experiment.
SPISE2009 43

REFERENCES
Beauchamp, G.K., Cowart, B.J., & Schmidt, H.J. (1991). Development of chemosensory sensitivity
and preference. In T. V. Getchell, R. L. Doty, L. M. Bartoshuk & J. B. Snow (Eds), Smell taste,
health and disease . New York: Raven Press, pp. 405-416
Bertino, M. & Chan, M.M. (1986). Taste perception and diet in individuals with Chinese and
European ethnic backgrounds. Chemical Senses, 11, 229-241.
Centre for International Economics of Canberra and Sydney (2003). Report of Vietnam sugar
program: Where next? Canberra: Ian Potter House.
Do, V.B., Patris, B. & Valentin, D. (2007) Using focus groups to study the representation of wine in
Vietnam. In D. Valentin, H.D. Nguyen & L. Pelletier (Eds), Proceeding of SPISE 2007: new
trends in sensory evaluation of food and non-food products. Ho-Chi-Minh city: National
University of Ho chi Minh City Publishing House, pp. 32-44.
General Statistics Office (2006). Statistical yearbook of 2005. Hanoi: Statistical Publishing House.
Moskowitz, H.R., Kumaraiah, V., Sharma, K.N., Jacobs, H.L., & Sharma, S.D. (1975) Cross-cultural
differences in simple taste preferences. Science, 190, 1217-1218.
Prescott, J., Laing, D., Bell, G., Yoshida, M., Gillmore, R., Allen, S., Yamazaki, K. & Ishii, R. (1992)
Hedonic responses to taste solutions: a cross-cultural study of Japanese and Australian panels.
Chemical Senses 17, 801-809.
Prescott, J. & Khu, B. (1995) Changes in preference for saltiness within soup as a function of
exposure. Appetite, 24, 302.
Laing, D.G., Prescott, J., Bell, G.A., Gillmore, R., Allen, S., Best, D. J., Yoshida, M., Yamazaki, K. &
Ishii,-M,R. (1994) Responses of Japanese and Australian panel to sweetness in the context of
different foods. Journal of Sensory Studies, 9, 131-155.
SPISE2009 44

COFFEE OR MARGARITA: IMPACT OF AMBIENCES ON BEVERAGE CHOICES

IN A BAR.
C. Dacremont *, A. Sutan , F. Galiab, J-F. Desmarchelierc, and D. Valentina
a b

a
AgroSup Dijon – CSG UMR5170 CNRS-Université de Bourgogne-INRA, Dijon, Fance
b
Groupe ESC Dijon Bourgogne - LESSAC, Dijon, France
c
EREBUS, Dijon, France
*Corresponding author: [email protected] Tel: +33 (0)3 80 68 16 64

Abstract
Two approaches are classically used to perform consumer tests: “central location” tests
favouring experimental condition control or “at home” tests favouring ecological validity. We
describe an alternative approach (based on immersion combining both approaches) which was
implemented to study beverage consumption in bars. In a cultural manifestation, two pub-like
ambiences were created with a few pieces of furniture: A traditional pub ambiance with raw
wood furniture and yellow lights and a modern ambience with blue translucent furniture. In
both ambiences, five video clips evoking different warmth levels were displayed successively
and participants had to choose what they would like to drink among a drink list. Hot drinks
were predominantly chosen when watching a clip depicting icebergs; Mexican drinks were
predominantly chosen when watching a clip depicting a desert, probably because of the
South-American style music that accompanied it. The three abstract clips (red, white and blue
on a black background) also impacted the drink choices but in a less extent.
Keywords: consumer test, context, drink choice, bar

1. INTRODUCTION
Hedonic tests are widely used in the food industry to get insight into consumers’ preferences
and identify the products that will perform best on the market. However, consumer tests still
need further developments. The difficulty of consumers’ testing originates from an inherent
contradiction: the objective is to gain information regarding the products whereas their
appreciation is context dependant by nature. Food behaviour is determine or influence by
numerous factors related to 1) consumers (physiologic, socio-economic or cultural factors), 2)
intrinsic properties (organoleptic characteristics) as well as extrinsic properties (origin,
nutritional claims, convenience, price, etc.) of products, and 3) contexts of consumption
(Rozin & Tuorila, 1993). For instance, the same food product will be more appreciated in a
restaurant than in a student cafeteria (Meiselman, Johnson, Reeve, & Crouch, 2000). Food
acceptability is related to food accessibility, time of consumption, eating location, social
interactions, and ambience (for a review, see Stroebele & de Castro, 2006). However, the
actual impact of these factors is not completely elicited and the effect of a given factor on a
specific food is still difficult to predict (King et al., 2007).
SPISE2009 45

For consumers testing, two approaches are classically used: central location tests (CLT) and at
home tests (ATH). CLT are run in a sensory room to standardise the contextual parameters; the
experimental conditions are strictly controlled to insure that all products are evaluated in
comparable ways. ATH favour ecological validity by allowing consumers to use or consume
products as usual. But then, experiential conditions may vary considerably from one product
to another decreasing the validity of products comparison. The few studies comparing CLT
and ATH did not demonstrate any superiority of one approach over the other (Boutrolle,
Arranz, Rogeaux, & Delarue, 2005; Bouterolle et al., 2007). An alternative approach to carry
consumers’ tests would consist in evoking a contextual situation in a controlled set up as
suggested by Köster (2003, 2009). This would combine advantages of both CLT and ATH by
insuring ecological validity of the tests through the evocation of a situation but still insuring
comparison among products by controlling testing conditions. However, an efficient
evocation may not be easy to implement. Petit and Sieffermann (2007) set up a sensory room
with some contextual element such as pictures, odour, and music to evoke a “warm
ambiance.” But, the appreciation of drinks tested in this situation was not different from
appreciation of drinks tested in a classic sensory room. This lack of effect may come from the
fact that a situation is not only a set of elements, but rather reflects the meaning that
consumers associate to the overall set up (Köster, 2009). A situation of consumption is related
to a specific pattern of elements including the food, the time of the day, the location, the
social environment, the activities in which consumers are involved, the mental processes
(emotions) and the recurrence of the episode (Bisogni et al., 2007). All these elements have to
be taken into account to evoke efficiently a situation of consumption.

The objective of the present study was to explore whether a consumption situation could be
evoked using contextual elements. We implemented this idea to study beverage consumption
in bars. We carried out the experiment during a cultural manifestation to match the
appropriate activities (going out with friends), the time of the day (evening), and the
recurrence (special occasion) of the situation. Two bars were materialised using a few pieces
of furniture. We manipulated the ambiance of the bars (traditional vs. modern) through the
furniture material (wood vs. plastic) and the lights (orange vs. blue). In each bar the warmth
of the ambiance was modified through short video-clips projected in the room. People were
allowed to participate with their friends and settle down at the same table. They received a
drink list offering a large choice of drinks comparable to what can be found in bars and their
task was to indicate which drink they feel like to drink for each ambiance. We hypothesised
that modification of the ambiance would lead to different choices from the participants.

2. MATERIAL AND METHODS

2.1 Participants
Individuals who attended a cultural manifestation dedicated to beer on the campus at the
University of Bourgogne took part in this experiment. They were allowed to attend both or
SPISE2009 46

only one of the two experimental bars; a majority of them attended both bars. Overall 92 and
82 participants completed the questionnaires in the traditional and modern bars respectively.

2.2 Material
Five video-clips combining video and music were specifically designed for this experiment.
Two figurative videos were chosen to evoke two contrasted ambiences: icebergs slowly
drifting in the sea and a road in a desert landscape. Three abstract videos pictured moving
coloured shapes on a black background. These videos were selected from a set of 20 videos
on the basis of their perceived warmth assessed by a group of about 30 students on a 10-point
scale anchored with cold at the left end and warm at the right end. The three selected videos
were white, blue, and red coloured leading to different levels of warmth from the coldest
(white) to the warmest (red). The same approach was adopted to select music from an array of
specifically composed tunes. Videos-clips combined videos and music according to their
warmth level and their tempo, especially for abstract videos were shapes move at a specific
pace. The duration of video-clips was from 1.5 to 2 min each.

2.3 Experimental setup

Two bars were settled in two adjacent rooms. In each room, one wall was white and was used
as a screen to project video-clips. The other walls were draped with black curtains and the
room was kept in dim light. A few pieces of furniture: three tables each surrounded by three
bar stools and a small counter, were dispatched in the room to evoke a bar. In one room, the
bar and tables were made from raw wood. Orange lights displayed on each table and on the
counter created a “traditional” ambiance. In the other room, the bar and tables were made
from a white translucent plastic material retro-lighted with blue light providing a rather
“modern” ambiance. The five video-clips associating video and music were displayed in loop
in the bar in a fix and predetermined sequence: blue, desert, white, iceberg, and red.
Participants who entered one of the bars where welcomed by an attendant, who explained the
task, provided them with the questionnaire and settled them down at a table. Participants were
also offered a drink to drink during their stay in the bar. This was aimed at enhancing the
situation by reproducing actual activity in a bar (i.e., having a drink). Participants were
offered either a glass of orange juice or of non-alcoholic beer, as they wished.

Participants were instructed to wait for the beginning of the next video clip, to take a moment
to enjoy the ambiance and then fill out the questionnaire. First, they were asked to choose
among a drink list (Figure 1) what they would like to drink at the moment. Second, they were
asked to assess the ambiance of the bar on 10-point scales regarding overall appreciation
(from disagreeable to agreeable), warmth (from cold to warm), arousal (from calm/relaxed to
excited/stressed) and harmony of the ambiance (from not harmonious to very harmonious).
Participants performed the same task (choosing a drink in the list and assessing the ambiance)
for each of the five clips consecutively.
SPISE2009 47

Hot drinks Fruit juices Cocktails

 Coffee  Orange juice  Bloody Mary
 Déca coffee  Grappefruit juice  Electric Blue
 Tea  Pineapple juice  Margarita
 Hot chocolate  Grappe juice  Pina Colada
 Apple juice
Water & Soft drinks
 Still water Beers Alcohols
 Perrier  Heineken  Whisky
 Coke  Desperado  Vodka
 Diet coke  Leffe blonde  Gin
 Seven up  Leffe brune  Cognac
 Ice tea  Hogaarden blanche
 Schweppes  Kreik
 Orangina  Chimay rouge
 Guiness
Figure 1: drink list. 

2.1 Data analysis

Ambiance ratings were converted into scores from 1 to 10. Liking, warmth, arousal, and
harmony scores were submitted to 3-ways ANOVA (furniture, clips, participants (in furniture)).
Whenever a furniture  video-clip interaction was observed, t-tests were performed to
compare the two furniture types for each video-clip. Drink choices were assessed by counting
the number of choices observed for each drink in each bar and each ambience. The
distributions of the choice in each condition were then analysed using 2 tests.
3. RESULTS AND DISCUSSION
3.1 Ambience assessments
The warmth of the ambience is driven by the video-clips only, F(4, 687) = 222.86, p <.0001)
as there is no significant furniture effect nor furniture  clips interaction. As expected the
iceberg landscape was assessed as the coldest and the desert landscape as the warmest. The
abstracts clips are assessed with intermediate levels of warmth and with significant
differences among the three clips.
Table 1: Mean warmth scores.
Video-clips Warmth scores
1-Blue 5.0 b
2-Desert 8.2 a
3-White 4.1 c
4-Icebergs 3.7 d
5-Red 7.9 a

Liking and arousal evoked by the ambiences are driven by the video-clips only as no
significant furniture nor furniture  clips interaction was observed. Table 2 shows the mean
scores obtained for each video-clip. For abstract videos there is a relation between liking and
SPISE2009 48

warmth scores. However, this link does not hold for figurative videos as icebergs and desert
landscapes are much contrasted in term of warmth but were equivalently liked.
Table 2: Mean liking and arousal scores.
Video-clips Liking scores Arousal scores
1-Blue 6.5 c 4.2 c
2-Desert 7.2 ab 5.9 b
3-White 5.0 d 6.6 a
4-Icebergs 6.8 bc 3.6 d
5-Red 7.4 a 4.1 cd

The furniture type did not impact the evaluation of warmth, liking nor arousal. But, the
furniture was not completely ignored by the participants. A significant furniture  clips
interaction was observed for harmony scores, F(4, 687) = 4.10, p < .01). It is worth noticing
that the interaction was significant for the abstract video-clips but not for the figurative ones
(Table 3).
Table 3: Mean harmony scores.
Video-clips Wood furniture Plastic furniture t-test (Pr)
1-Blue 5.79 6.38 1.60 (.11)
2-Desert 6.64 6.68 0.13 (.90)
3-White 4.36 5.39 2.88 (.004)
4-Icebergs 6.68 6.62 -0.18 (.86)
5-Red 7.16 6.37 -2.35 (.02)

3.2 Drink choices

Some beverages were pooled as they showed similar choice profiles: tea and hot chocolate
were pooled as “hot drinks”; still water and Perrier were pooled as “water”; coke, diet coke,
seven up, ice tea, Schweppes, and orangina were pooled as “soft drinks, and all fruits juices
were also pooled.
Figure 2 shows the number of choices for different drinks for each bar and each ambience.
The ambience and more specifically the video-clips had a clear impact on drink choices. The
video-clip depicting icebergs led to choose preferentially hot drinks; more than half of the
participants chose a hot drink in this ambience. It was even more pronounced in the wood bar
although the difference is not significant. The video-clip depicting a desert landscape led to
choose more desperado, margarita and pina colada than in other ambiences (2 = 96.6; p p <
.001). These three drinks are typically link to South-America at least for French people. These
choices were probably oriented by both the video and even more the music that evoked
South-America. Abstract video-clips also induced some specific choices. The white video-clip
led to choose significantly more spirits (vodka and in a lesser extent gin and whisky) than in
other ambiences (2 = 87.3, p < .001). The blue video-clip led to choose significantly more
fruit juices than in other ambiences (2 = 11.0, p < .001). Finally, the red video-clip led to
choose significantly more stout beer than in other ambiences (2 = 36.9, p < .001).
SPISE2009 49

Plastic furniture Wood furniture

20
Blue Clip
10

0
20
Desert Clip
10

0
30
White Clip
20

10

0
50
Icebergs Clip
40

30

20

10

0
20
Red Clip
10

0
Hot drinks

Coffee

Soft drinks
Water

Heineken

Desperado
Fruit juices

Leffe blonde

Leffe brune

Hogaarden_blanc

Kreik

Chimay_rouge

Guiness

Bloody Mary

Electric blue

Margarita

Pina Colada

Whisky

Gin
Vodka

Cognac
Figure 2: Number of drink choices according to bars and video-clips

4. CONCLUSION
This experiment showed that it is possible to evoke a bar situation and to impact the drink
choices by modifying the ambience inside the bar. As expected, figurative video-clips very
contrasted in term of temperature evocation clearly induce some specific drink choices. But,
abstract video-clips were also efficient to direct choices. This experiment is a preliminary
study demonstrating the potential of an immersive approach to credibly evoke a consumption
situation. However, in this experiment, we only recorded declarative choices, participant were
well aware they will not actually get the drink. A further step will be to confirm the interest of
this approach to study actual behaviours.

Acknowledgment: The authors express their thanks to C. Sester, A-R. Noël, and A. Pécourt
for their help in running the experiment.
SPISE2009 50

REFERENCES
Bisogni C.A., Winter Falk L., Madore E., Blake C.E., Jastran M., Sobal J., & Devine C.M.
(2007). Dimensions of everyday eating and drinking episodes. Appetite 48, 218–231.
Boutrolle I., Arranz D., Rogeaux M., & Delarue J. (2005) Comparing central location test and
home use test results: Application of a new criterion. Food Quality and Preference 16,
704–713

Boutrolle I., Delarue J., Arranz D., Rogeaux M., & Köster E.P. (2007). Central location test
vs. home use test: Contrasting results depending on product type Food. Quality and
Preference, 18, 490–499

King S.C., Meiselman H.L., Hottenstein A.W., Work T.M., & Cronk V. (2007). The effects of
contextual variables on food acceptability: A confirmatory study. Food Quality and
Preference 18, 58–65.
Köster E.P., (2003). The psychology of food choice: Some often encountered fallacies. Food
Quality & Preference, 14, 359-373
Köster E.P., (2009). Diversity in the determinants of food choice: A psychological
perspective. Food Quality & Preference, 20, 70–82.

Meiselman H.L., Johnson J.L., Reeve W., & Crouch J.E. (2000) Demonstrations of the
influence of the eating environment on food acceptance. Appetite 35, 231-237.
Petit, C. & Sieffermann J.M. (2007). Testing consumer preferences for iced-coffee: Does the
drinking environment have any influence? Food Quality and Preference, 18, 161–172.

Rozin P. & Tuorila H. (1993). Simultaneous and temporal contextual influences on food
acceptance. Food Quality and Preference, 4, 11-20.
Stroebele, N. & de Castro, J.M. (2006). Listening to music while eating is related to increases
in people's food intake and meal duration. Appetite, 47, 285-289.
SPISE2009 51

THE SOUPE DU JOUR EFFECT: LANGUAGE AS A COUNTRY-OF-ORIGIN CUE

AND ITS IMPACT ON PRODUCT PERCEPTION
T.X.U. Phana* and C-F Sheub
(a) HoChiMinh-city University of Industry, Vietnam
(b) National Cheng Kung University, Taiwan
*Corresponding author: [email protected], Tel: +84.8.8647256 (ext: 5696)
Abstract
This research investigated the effect of language as a proxy of country of origin on people’s
perception of a new food product and their purchase intention. A primary study has shown
that “to which country a food is thought to belong could influence whether or not people
would be willing to try it”. This effect was named the “soupe du jour” effect. It was verified,
firstly, by a survey conducted in Taiwan. In this study, participants read a cover story of a
new tea product and its advertising bilingual description (traditional Chinese Mandarin
combined with either simplified Chinese Mandarin or English or Japanese or Vietnamese and
Chinese traditional by itself). They were asked to indicate the prices and whether they were
willing to buy the tea products on a 5 point scale. The “Soupe du jour” effect was also
confirmed by an experimental study. The stimuli were five bona-fide tea boxes with labels
written in one of the five languages mentioned above. The participants were asked to choose
one and only one tea box to taste. They also indicated the price of each tea box in comparison
with the rest. The findings suggested that, depending on the types of products and the
cultures, language could serve as proxy of country of origin cue and elicit a positive or
negative “soupe du jour” effect.
Keywords: Country of Origin, product perception, language effect, culture difference
1. INTRODUCTION
Eating is a basic physiological need (Maslow, 1943), but its psychological, social and cultural
dimensions are undeniable (e.g., Rozin & Vollmecke, 1986; Rozin, 1996). During this time of
globalisation, foods from other countries are crossing the traditional borders of food cultures.
Even though encountering new foods is becoming all the more common, relating to novelties
in everyday life may be ambiguous. Relating to new foods evokes ambivalent thinking, as
people eat “with their mind as much as with their mouth” (Beardsworth & Keil, 1997). Food
choice entails not only decisions based on conscious reflection, but also automatic, habitual,
and subconscious decisions (Furst et al., 1996). Therefore, people are not as rational in food
choice as they think they are. Factors such as price, brand name, trademarks, promotion
appealing, or even foreign words in the commercials are able to drive people to consume
things.
Explanations and predictions of people’s choices are often founded on the assumption of
human rationality. In food context, it is often that one has to make a choice from many
different options of a food product category without tasting it first. This happens everywhere,
at anytime, in supermarkets, in restaurants, in food stores or in open markets. Without the
help of sensory tasting, people acquire other cues to justify their choice. Take food context in
a restaurant as an example, instead of “Seafood filet”, “Succulent Italian Seafood Filet” was
found to increase sales by 27 percent (Wansink et al., 2001). It is also common to find
restaurants with foreign names such as “Le Village” whose food has nothing to do with
French cuisine. Another example is the packed coffee “La gauche de La Seine - Café au Lait”
of Uni-President Enterprises Corp. – the biggest food company in Taiwan. It seems that
foreign languages have an effect on consumers; otherwise, restaurants and food companies
SPISE2009 52

would not have used them in their business. However, is it true that foreign language could
induce people to consume a new food product? What is the process underlying it? And how
does it work? Is the language effect a product-specific effect?
In previous work we explored the existence of a language effect, named “soupe du jour
effect”, on inducing people to try a novel food product. Our hypothesis was that the country to
which a food was thought to belong could influence whether or not people would be willing
to try it. We showed that Taiwanese and Vietnamese participants was influenced differently
by the information linked to the country to which the food was though to belong:
expensiveness was weighted more by Taiwanese, and familiarity was weighted more by
Vietnamese (Phan & Sheu, 2007; Phan, 2008)
The goal of the current study was to investigate further the effect of extrinsic cues, languages
and country of origin, on people perception of price and willingness to try (purchase) a new
food product. Our approach was twofold. First we used a survey to examine the effect of
country of origin images and its interaction with languages on people willingness to buy a
new tea product, and price perception. Then we carried out a behavioural experiment to
validate the results of the survey.
2. QUESTIONNAIRE SURVEY
2.1 Material and methods
Stimuli. A cover story of a new tea product was created. This cover story included a
description of the tea in the form of a commercial advertisement. The description gave
participants general information about the flavour of the tea without any specific information
such as the kind of tea, health information, and so on. In the description, the country where
the tea leaves came from was explicitly stated. Two factors were explicitly manipulated: the
country where the tea leaves were supposed to come from, and the language in which the
cover story was presented.
 Country of Origin: Five country names were used: Taiwan, China, Vietnam, Japan and
England. Those countries were all tea producing countries but differed in term of
economic development: China and Vietnam are developing countries, Japan and
England are developed countries and Taiwan is a new developed country.
 Language: Five languages were used: traditional Chinese Mandarin, simplified
Chinese Mandarin, Vietnamese, Japanese, and English.
Participants. A total of 700 people participated in the survey: about 30 in each COO and
language combination. Participants were tea consumers (aged from under 25 to over 60,
sample’s ratio of female/male about 1/1, from different professions).
Procedure. Participants were first asked to rate their willingness to purchase the tea product
on a scale of 1 to 5 (1 - least likely to buy, and 5 – most likely to buy). Then they were asked
to select a market price for the tea product from five different prices provided. The price
options were chosen based on data from a previous survey of the market prices of different
kind of teas, both domestic and imported products, in several main supermarkets in Taiwan.
The survey was conducted in different places (university campus, public parks, offices,
department stores, hospitals). This was a self-administrated survey, and the participants were
approached by the interviewer directly. They were introduced to the survey and its scientific
purpose. Half of the participants did the task of rating their willingness-to-buy the new tea
first. The other half did the task of choosing the market price first. Participants were allowed
to take as much time as they wanted to complete the task.
Data Analysis. Analysis of variance (ANOVA) was used for comparing the mean scores of
price choices and “willingness to buy” (WTB) in different languages and COO conditions.
SPISE2009 53

Multiple comparison using Tukey’s adjustment were conducted on the mean scores of all
paired-wise combinations of languages and COO. The prices were converted to an ordinal
scale going from1 to 5.
2.2 Results and discussions
Taiwanese sample: Two two-way ANOVAs were used to test the effect of COO direct (country
name) and indirect (language) cues on price perception and WTB (Table 1 and 2). Country
name had a strong effect on both price-perception and WTB. Although the main effect of
language was just marginal for price-perception, its effect is reflected through the interaction
with country-name. In contrast, there was no main language effect on WTB and its interaction
with country-name cue was marginal.
Table 1: ANOVA of country and language effect on price perception
Source Df SS3 Mean square Value Pr > F
Country 4 21.72 5.43 4.82 0.0008
Language 3 8.92 2.97 2.64 0.0489
Country*Language 12 35.03 2.92 2.59 0.0024
Residuals 533 601.06 1.13

Table 2: ANOVA OF Country and Language Effect on WTB

Source Df SS3 Mean square Fvalue Pr > F
Country 4 19.75 4.94 4.29 0.0020
Language 3 8.05 2.68 2.33 0.0731
Country*Language 12 24.19 2.01 1.75 0.0533
Residuals 537 618.04 1.15

Simple effects were run to examine the country by language interaction for both price
perception and WTB. Results showed that the interaction effect of country name by language
came mainly from the country name “Taiwan” and language “Japanese”. Traditional and
simplified Chinese Mandarin did not have any interaction with the country names. Regardless
of where the tea was from, if its description was in Chinese, Taiwanese consumers perceived
its price in the same way. The same situation was found for the country names of Japan,
Vietnam, and China. If the tea was from these three countries, no matter what language was
used to describe it, people perceived the price of the product in the same way.
We assumed that participants perceived the tea described as Taiwan tea in traditional Chinese
Mandarin to be the tea product marketed in Taiwan only. At the same time, they thought of
Taiwan teas with foreign languages as a product marketed in foreign countries. Besides, if the
tea product was described only in traditional Chinese Mandarin, no matter where it was told
to be from, people would perceive it as a product sold only in Taiwan.
There was no difference in price perception between tea from Taiwan and teas from other
countries if they all were thought to be sold only in Taiwan (see Table 3). Meanwhile, Taiwan
tea should be advertised in Japanese to gain the best price in people perception (see Table 4).
Chinese simplified, and English could not make the product be perceived more expensive
than the one just advertised in traditional Chinese.
SPISE2009 54

Table 3. Difference in price perception between Taiwan tea and tea from other countries which were
described in traditional Chinese Mandarin
Difference between Country of Origin Difference between the
(Chinese traditional description) means of price choice
Taiwan – Vietnam - 0.35
Taiwan – China 0.01
Taiwan – England 0.07
Taiwan – Japan - 0.52

Table 4. Difference in price perception between Taiwan tea marketed in Taiwan and Taiwan tea sold
in different countries (* significantly different at =.05)
Taiwan tea, difference in languages of Difference between the mean of
description price choice
Chinese Traditional – Chinese simplified - 0. 48
Chinese traditional – English - 0. 47
Chinese traditional – Japanese - 1.09 *

Besides, as the simple effect analysis revealed strong effects of Taiwan COO and Japanese
language, it drew our interest in comparing Taiwan tea in different conditions of languages
with each other to reveal the effect of perception of markets, as well as comparing Taiwan tea
with other countries’ teas in the condition of traditional Chinese only to reveal the effect of
COO.

Table 5. Difference in price perception and WTB between Taiwan tea and other teas, with
description in Japanese (* significantly different at =.05)
Difference between Country of Difference between Difference between
Origin (Japanese description) Means of Price choice Means of WTB

Taiwan – Vietnam 1.007 * 0.88 *

Taiwan – China 1.02 * 0.53
Taiwan – England 0.40 0.13
Taiwan – Japan 0.70 0.74 *

Though Japanese was found to have a strong effect in inducing people to perceive the tea
product more high-end and show high purchase intention, it could suppress the COO effect
linked to countries with a low image. An advertisement in Japanese would not be enough to
bring Vietnam and China tea to the same class as Taiwan tea. However, Japanese language
was not found to increase people purchase intent for Japan tea in comparison with Taiwan tea.
Japanese gave the most effective effect on Taiwan tea, but not on Japan tea (Table 5).
2.3 Conclusion
The results of this survey were consistent with the findings of our previous study. Languages
could serve as COO to elicit price and familiarity perception in people about food from
different countries. The “made-in” effect was also found to have a strong effect on
participants’ perception of product. Above all, Taiwanese had strong preference for tea “made
in Taiwan.” This demonstrated a strong home-country selection bias, which was consistent
with many previous studies on COO effect, such as Levin et al. (1993), Good and Huddleston
SPISE2009 55

(1995), or Baker and Michie (1995). In Taiwan, drinking tea has developed to an art or tea
ceremony, which was adapted much from Chinese and Japanese tea cultures. Therefore, to
Taiwanese drinking tea is sort of sacred. Thus, a familiar taste may be desirable. An
unfamiliar imported tea could not satisfy people needs in this aspect.
However, Taiwanese people seemed to have desires of high-class products. If they could have
a similar taste in a high class tea product, it would be the best choice ever. Therefore, nothing
could be better than a tea made in Taiwan but supposed to sale in Japan. Despite all the fame
Japan earns in the world market for other kinds of products, just the fact that Japanese tea
ceremony has a worldwide reputation was enough to be a secure for a perception of a high
class level of any tea product which was sold in that market. This could be a reason why
Japanese language emerged as the most effective language to influence Taiwanese
consumers’ purchase intent to a new tea product. This finding stabilized the “Soupe du jour”
effect, for the fact that it existed even on a special product category like tea.
Yet this first study used a direct and self-administrated questionnaire survey method to
approach people perception. This methodology was based on the implicit idea that people are
“reasonable” and make rational choices. Unfortunately, psychological theories show that
many decisions are irreflexive, and that people simply rationalise because they do not know
the true reasons for their choice. It means that they think up an answer to any question, even
unanswerable question, instead of saying they don’t know. That’s why the fallacy of
conscious choice becomes particularly dangerous in sensory science and consumer research
(Köster, 2003). Knowing this fallacy is especially important in consumer survey research, in
which the investigators use questionnaires to interact with consumers. Several studies showed
that in that case, using indirect methods is more efficient than direct methods. Behavioural
frequency questions are used to imply liking. Asking indirect questions always gets more
reliable answers than direct questions, for instance, “Have you ever hesitated to invite friends
to your garden at home because of the pollution” prove to be a much better question to
estimate the pollution in an area than “How bad is the odour pollution in this
neighbourhood?” (Köster, 2003). Therefore, the validity of the findings of our first study
needs to be confirmed using behavioural data.
3. BEHAVIOURAL EXPERIMENT
3.1 Materials and methods
Stimuli. Fake tea boxes were made from ordinary bamboo-style tea boxes. The original labels
of the boxes were removed, and replaced by experimental labels. To control for unexpected
factors which could affect the results, the stimuli labels were made with the same image, logo,
expiration date, and distributor company. The content of the labels were kept similar, too.
Only languages of the labels were different. Five languages were adopted from the precedent
survey. They were traditional Chinese Mandarin, simplified Chinese Mandarin, Vietnamese,
English and Japanese. Figure 1 showed the stimuli and the English labels. Other labels had
the same picture and content, but different languages. Those designed tea boxes have gone
through a pre-test to verify their convincibility as real products before they were used in the
main experiment.
SPISE2009 56

Front Back

Figure 1. Tea stimuli used in behavioural experiment

Participants. A total of 148 college students (aged from18 to 25 years old) and 2 faculties (30
to 45 years old) from the Department of Baking Technology and Management in National
Kaohsiung Hospitality College in Kaohsiung participated in the experiment.
Procedure. The experiment included four tasks carried out in the following order (1) language
recognition test, (2) tea label forced choice, (3) price estimation and (4) demographic
information. The language recognition test was performed in the waiting room. The last three
tasks were conducted in the testing room, which was decorated as a tea shop. There were five
separate tables in the room. These tables were arranged to diminish the contact between
participants. A set of five experimental tea boxes was available on each table.
Participants received a cover story about a tea distributor who wanted to market five different
new tea products in Taiwan and foreign markets. Before officially doing the market
launching, the company wanted to conduct a consumer test to examine the promising success
of their new products. The distributor would like to invite them to join the test.
In the language recognition test, participants received a questionnaire containing five slogan
sentences written in either traditional Chinese Mandarin, simplified Chinese Mandarin,
Vietnamese, English or Japanese, and five names tags corresponding to those languages. They
were asked to match each sentence with one language. If they could not answer all correctly
in this task, they would be ineligible to the experiment.
SPISE2009 57

In the label forced choice task, participants were asked to examine the tea boxes in front of
them (label, logo) and decide which tea they wanted to taste later (Figure 2).

No: ………………………………………………

Please take a careful look at five different tea products in front of you.
Their labels are written in five different languages:

Mandarin Traditional Mandarin Simplified

Vietnamese Japanese English.

Indicate which one you like to try. Make a choice by circling the name
of the label above. The tea you chose will be served to you in 10 mins.

Figure 2. Label forced choice task

After completing the label choice task, participants moved on to the price estimation task. In
this task, subjects were randomly divided into five groups of 30 persons each. Each group was
assigned into a COO of the tea product condition. In the five different conditions of COO,
participants were given a scenario saying that all five tea products came from one of five
countries of Taiwan, China, Vietnam, England and Japan. Depending on what country the tea
was told to be from, the anchoring price corresponded to the tea box whose label was written
in the language of that country. Based on the anchoring price, participants had to estimate the
price of the other four tea products, indicated by their labels’ languages (Figure 3).

No:………………….

All of the tea products are manufactured in different parts of

Japan. The product with the label written in Japanese has the
market price of 550 NT for a box of 300 grams. Please estimate
the price of the other tea boxes at the same weigh.

Box with Vietnamese label: …………………………………

Box with Traditional Mandarin label ……………………….

Box with English label: ……………………………..............

Box with Simplified Mandarin label: ……………………….

Figure 3. Price estimation task in the condition of Japan COO. The tea was told to be made in Japan,
and the tea box with Japanese label was the anchor product to estimate the other prices
SPISE2009 58

When participants have completed this task, they received a questionnaire of 11 demographic
questions and a cup of tea. This questionnaire included questions on participants’ tea habit
and preference, personal information, and reason of their choice in the label choice task
Data Analysis. Data of label choice task was analysed using the proportion test (Wilson,
1927; Newcombe, 1998 a and b). Participants’ reasons of choosing the label were classified
into three main groups: extrinsic product characteristics (packaging, language), socio-cultural
factors (cultural, economical influence, trust in industry, habits) and psychological factors
(emotion, previous experiences, personality traits, belief, and perception). Price estimation
data were analysed by one-way ANOVA model for each COO. Other demographics information
was collected to help explaining the results of label choice and price estimation tasks.
Taiwan Vietnam
Vietnam

Taiwan
200 600
Vietnamese Simplified English Japanese Simplified Traditional English Japanese
150 500

100 400
Difference

50

Difference
300
0
200
-50
100
-100
0
-150

-200 -100

En gland
China
Ch in a England
300 200
Vietnamese English Traditional Japanese Vietnamese Simplified Traditional Japanese
250 150
200 100
150 50
Difference
Difference

100
0
50
-50
0
-100
-50
-150
-100
-150 Japan -200
Jap an
100
Vietnamese Simplified Traditional English
50

0
Difference

-50

-100

-150

-200

Figure 4. The differences between the estimated price scores of experimental language labels and the
anchoring price in the five countries of origin conditions.

3.2 Results
Label choices. The observed proportions of label choice were: 52% for Japanese, 36% for
Traditional Chinese, 6% for English, 4% for Vietnamese and 2% for simplified Chinese.
These observed proportions were significantly different, χ2(4) = 192.75, p < 2.2e-16. The
Japanese label was the most chosen among the five labels even when compared with the
SPISE2009 59

second-most-chosen label (traditional Chinese Mandarin). Significant differences were also

found between the Chinese traditional label and the other labels.
Price estimation. Figure 4 shows the differences between each condition of labels’ languages
with the anchoring language condition for each “made-in” country. Results showed that only
Japanese language increased price for a tea made in Taiwan. Japanese also increased the price
of Chinese and English tea. Traditional Chinese Mandarin, English and Japanese all increased
the price of Vietnam tea. In case of Japan tea, no increase in price was found in all conditions
of languages.
3.3 Conclusion
This behavioural experiment verified the language effect or “soupe du jour” effect on product
perception in two important indicators: price and choice. Japanese language was found to be
the most effective language to increase the price of tea. Vietnam gained the most benefit,
followed by China. Taiwan and England had the same gain from Japanese language for their
tea products.
All participants highly valued the tea with Japanese labels, regardless from the origin of the
tea. They associated Japan with price, high class and high quality, and they showed high
desire for Japanese label.
4. CONCLUSIONS
The survey and behavioural experiment demonstrated the existence and magnitude of a
“soupe du jour” effect on perception of two key characteristics of a tea product: price and
purchase intention. Language was shown to be an indirect way of communicating “made-in”
concept. First, language activates the image of the country it represents. Then, the country
image activates people’s perception of its representative products, national characteristics,
economic background, history, and tradition. Finally, this activated knowledge influences
how people evaluate the food product which is presented in that language.
We also found strong interaction effects between COO and language on people perception of
price and purchase intent of a new tea product. This result was inconsistent with Lim et al.
(1994). These authors have demonstrated that when other product information cues are
provided at the time of exposure to the COO information, such product information lessened
the effect of COO. But our findings showed the reinforcement of language cue and COO effect
when they were presented together. Furthermore, Taiwan COO and Japanese language proved
to be the best combination between COO and language to increase price perception and
purchase intent.
To sum up, this study has shown how irrational people could be. No one would realise that
they were easily manipulated just by a simple factor as language. People always tried to be
rational in their choice by acquiring information to justify their choice. To do that, people
need cues to guide them. The “soupe du jour” effect provides a language cue, which was
automatically associated to country image. This process was somewhat unavoidable. In other
words, foreign language in a description does more than just getting attention from
consumers. It is an alternative way to elicit COO effect and influence the perceived price of a
food product and people’s willingness to purchase the product.
REFERENCES
Baker, M.J. & Michie, J. (1995). Product country images: perceptions of Asian cars.
University of Strathclyde, Department of Marketing, Working Paper Series No. 95/3.
Beardsworth, A. & Keil, T. Sociology on the menu: An invitation to the study of food and
society. Routledge. 1997. 277pp.
SPISE2009 60

Furst, T., Connors, M., Bisogni, C. A., Sobal, J., & Winter Falk, L. (1996). Food choice: a
conceptual model of the process. Appetite, 26, 247-266.
Good, L.K. & Huddleston, P. (1995). Ethnocentrism of Polish and Russian consumers: are
feelings and intentions related? International Marketing Review, 12 (5), 35-48.
Köster, E.P. (2003). The psychology of food choice: Some often encountered fallacies. Food
Quality and Preference, 14 (5-6), 359-373.
Levin, I., Jasper, J., Mittelstaedt, J., & Gaeth, G. (1993). Attitudes towards ‘buy America
first’ and preferences for American and Japanese cars: a different role for country-of-
origin information. in McAlister, L., & Rothschild, M.L. (Eds), Advances in Consumer
Research, 20, 625-629.
Lim, J., Darley, W.K., & Summers, J.O. (1994). An assessment of country of origin effects
under alternative presentation formats. Journal of the Academy of Marketing Science, 22
(3), 274-282.
Maslow, A. (1943). A theory of human motivation. Psychological Review, 50, 370-396.
Newcombe R.G. (1998a). Interval estimation for the difference between independent
proportions: comparison of eleven methods. Statistics in Medicine, 17, 873–890.
Newcombe R.G. (1998b). Two-Sided Confidence Intervals for the Single Proportion:
Comparison of Seven Methods. Statistics in Medicine, 17, 857–872
Phan, T.X.U & Sheu, C.F. (2007). Would you try this? Soup of the day (Soupe du jour). Oral
presentation at SPISE2007 symposium: New Trends in Sensory Evaluation of Food and
Non-food Products, HoChiMinh City, Vietnam, 26-27 July 2007.
Phan, T.X.U (2008). The Soupe du Jour effect: Language as a Country-of-Origin cue and its
impact on product perception. Master thesis. Institute of Cognitive Science, National
Cheng Kung University, Taiwan.
Rozin, P. (1996). Towards a psychology of food and eating: from motivation to module to
model to marker, morality, meaning, and metaphor. Current Directions in Psychological
Science, 5, 18-24.
Rozin, P. & Vollmecke, T.A. (1986). Food likes and dislikes. Annual Review of Nutrition, 6,
433-456.
Wansink, B., Painter, J., & Van Ittersum, K. (2001). Descriptive menu labels’ effect on sales.
Cornell Hotel and Restaurant Administration Quarterly 2001; 42; 68
Wilson, E.B. (1927). Probable inference, the law of succession, and statistical inference. J.
Am. Stat. Assoc., 22, 209–212.
SPISE2009 61

ATTITUDES AND SPENDING BEHAVIOURS FROM A MIDDLE CLASS POINT OF

VIEW DURING THE CURRENT ECONOMIC RECESSION IN VIETNAM
T.M. Le(a*), D.T. Do(b), D.H. Nguyen(c)
(a) Food Industry College of HoChiMinh-City, Vietnam.
(b) SaoViet Sensory and Consumer Research Company, Vietnam.
(c) HoChiMinh-City University of Technology, Vietnam
*Corresponding author: [email protected]
Abstract
The objective of this study was to investigate attitude and spending behaviour of members of
middle class in the current economic recession period in Vietnam. Two experiments were
carried out: a focus group and a consumer survey. The focus group results provide a
description of the general profile of the economy of Vietnam: “this economic situation is not
in crisis period,” “it is a downturn as a consequence of domino effect from the recession of
capitalistic countries”; and describe the concept of recession as “difficult life”, “bankruptcy”,
“unemployment”, and “rise of social evils”. Facing the recession situation, the reactions of
the participants were to “work harder”, “study more”, “recruit/select superior sale
managers”, “satisfy the outside and inside clients”, and “manage expenses in a planned
way” The questionnaire survey focused on the consumer expense distribution patterns and on
forecasting consumer spending behaviours. The proportion of expenditures showed that the
five items on which consumers spent the most were food at home (20.56%), savings
(15.29%), food services (8.72%), payment for rent (5.61%), and party expenditure (5.34%).
Moreover, a new trend in spending behaviour was discussed. Eleven items of expenditures
were highlighted in this discussion: food at home, alcoholic beverage, clothing, jewellery,
travelling, private transportation, telephone, party expenditure, home equipment (with and
without motors), and payment for rent.
Keywords: Vietnam economic, recession, spending behaviour, income distribution

1. INTRODUCTION
Erupting in US financial declination, the current recession has spread rapidly and wrecked
many countries all over the world. The global recession was revealed via some obvious signs
such as: increasing prices of basic food and fuel, increasing unemployment, decreasing retail
turnover, decreasing actual revenue and the gradually weakening industrial growth index
(IMF, 2009, Shimelse et al., 2009). All of these extraordinary changes imply a global
economy shrinking. This recession has affected consumer spending. According to Jena
(2009), 73 % of Americans planned to cut back on spending in the next 12 months, and 26%
of Chinese were likely to trim their expenditures. Like consumers all over the world,
Vietnamese consumers also have to prepare for an economic emergency, with consequences
such as: losing jobs, cutting costs and changing spending habits. The objective of this study
was to explore the attitudes and behaviours of middle class consumers in Hochiminh city
SPISE2009 62

towards the current recession. Hochiminh city is a region that has Gross Domestic Product
(GDP) per capita at the highest level in Vietnam. Its GDP per capita reached 2,534 USD in
2008 (Song, 2009) and the middle class has income about 5-15 million VND/month (equal
294-882 USD/month).

2. METHOD
This research included two studies: a focus group study and a market survey. The objective of
the focus group study was to obtain information about concepts, attitudes and reactions of
Vietnamese consumers in a recession period. And the purpose of the market survey was to
investigate the behavioural changes in spending patterns.

2.1 Focus groups

A total of five focus groups were held in Hochiminh city. Numbers of participants in each
group were between 3 to 9 persons, with a total of 27. The main characteristics of focus group
are presented in Table.1.

Table 1. Main characteristics of focus group

G1 G2 G3 G4 G5
Participants 6 9 4 3 5
Age (years) 25-34 25-34 35-44 35-44 45-55
Participants were selected based on their income (5-15 million VND/month) and age (25-55
years old). They worked in banking, information technology, retailers (food, construction
materials), logistics, and bureaucratic government. All focus group sessions were conducted
in a room large enough to accommodate 10 persons comfortably. The room was equipped
with fluorescent lights, video and audio recorder, blackboard, and a round table to allow free
discussion. The protocol guide was based on a semi-structured interview (Barrios, 2008; Ives,
2003). All focus group sessions were operated by one moderator and two assistants. The
moderator was responsible for facilitating the group discussions and the assistants controlled
the audio equipment and took notes. When the participants sat around the table, the moderator
started the discussion by introducing himself. Then he informed participants that there were
no correct or incorrect answers to each question, but rather just personal point of views
concerning their professionals and social positions. Subsequently, the participants were asked
to briefly introduce themselves, and the discussion started around the following questions:
1. What do you think about Vietnam’s current economic situation? How to get this
information?
2. Thinking about economic recession, what is your first impression?
3. How does this recession affect your job?
4. Which solutions were found by your firm?
5. How does this recession affect your life?
SPISE2009 63

6. What do you do in recession period?

7. When will the recovery begin?
Each session lasted approximately 90 minutes, and the interviews were recorded.

2.2 Market survey

The survey was conducted in March 2009 on 200 inhabitants who have an income in the
range of 5-15 million VND/month in HoChiMinh city. The questionnaire focused on their
main spending categories (i.e. food, beverages and tobacco, clothing products, tools, health-
care services, cultural activities, sport, recreational facilities, transportation, post, medical
care, education, communicative relationship, home equipments, housing, building, materials
and saving). For each item, respondents were asked to indicate their expenditures at two
periods (before and during recession).

3. RESULTS
3.1 Focus group
The moderator and the two assistants analysed the notes taken and the transcript tapes.
Results were summarised by the following themes: a) Vietnam’s current economic situation;
b) Effect and reaction to recession on their jobs/lives; c) Forecasting the ending time of
recession.
Vietnam’s current economic situation: The focus groups revealed both positive and negative
aspects of the recession period. Most participants mentioned that Vietnam’s economy was
influenced by domino effect from global economic recession through opinions such as: “this
economic situation is not in crisis period”, “we are living in recession period, not crisis one”,
and “it is a downturn as a consequence of domino effect from the recession of capitalistic
countries”. for these participants, the concept of recession was defined by “difficult
investment environments”, “bankruptcy risks”, “slackness of import-export activities”,
“unemployment”, “price increase”, and “consumer spending declination”. Participants
expressed other causes of recession, as follows: “the financial management and the monetary
systems has not been received a good control”; “the policy system for investments is not
thorough”; and “our banking was not prepared to face recession.
Effect and reaction to recession: The four respondents working in bureaucratic government
(education and manufacture) and information technique indicated that they were not affected
by economic recession. However, they did not disclaim that it is hard to reach new targets like
the years before. According to most of the participants, with the rising of fuel and material
prices, they spent substantially more on manufacturing goods, and shipping merchandise.
Companies would prefer not to push those cost increases down the chain to customers. They
accept to reduce their incomes, cut down spending, or even add supported service; but, they
do not want to discount the product prices. The underlying reasons of this behaviour were
explored further in the focus groups. Respondents talked about: “Streamlining operations”,
SPISE2009 64

“Cutting back on inventories”, “Reducing part-time staff hours”, “Conserving energy”,

“Buying more local goods to save on transportation”, “Offering incentives to keep customers
engaged”, “Paying more attention in local market”, “Recruiting/selecting superior sellers”,
“Adding supported services”, and “losing customers if price hikes”. The two participants
who are retailers (in different focus groups), indicated another difficulty in rotating capital for
company. They mentioned that “most suppliers request cash in business”, and “have not
enough cash for commercial transaction”. Furthermore, in Vietnam, “Cash is King”. So,
companies/retailers prefer to keep their capital invested in bank rather than making risky
investments.
When asked about their daily lives, many participants recognised having changed their
behaviour during recession. They have to “work harder, study further and/or improve their
skill to keep jobs“ Besides, they mentioned cutting down spending on: “doing shopping, using
food away from home, using out of home entertainment, upgrading technology (home
equipments, cell phone, computers, motors), smoking and using luxury products.”

Forecasting the ending time of recession: The period at the end of 2009 was predicted to be
the most difficult period for the economy. That is a corollary of all the decisions from
sweeping job and changing spending habits that people made from the end of 2008 and first
quarter of 2009. However, participants’ belief in the Government’s economic stimulating
package was strong enough to lift up the economy. Three out of nine participants in group G2
showed a positive point of view in recession period: “This time is a good opportunity for
companies that have good strategy in management and business, This is a chance to reject the
business live from hand to mouth, this is also a good chance for small-scale companies
compete with large-scale ones about price and marketing strategies.”
Most participants are optimistic as to their market ability to ride out the current economic
recession. It is interesting that they appear more confident in their future. They mentioned that
the “Recession, will] just survive for a short time”, “we will overcome this time soon”, “our
government has stimulated the economy, we do not worry so far”, and “our economy will
recover in the next year”. So, Vietnam stays at the 14th position in the Nielson Global
Consumer Confidence Index (Nielson, 2008).

3.1 Marker survey

The survey data were analysed using a Wilcoxon t-test to determine whether there is a
significant change in expenditure between two periods. Besides, the percentage of spending
for each subcategory was calculated to forecast trends of spending habits.

The changes of expenditures in crisis time: The maximum and median values expenditure
costs before and during recession are presented Table 3.along with Wilcoxon t-test.
Statistically significant differences (p<0.05) were found for 11 subcategories. The expenses of
three out of eleven sub-categories have been increasing in recession period, including: food at
SPISE2009 65

home, party for communicative relationship, and payment for rent. In contrast, eight
expenditures have been declining, including: alcoholic beverages, clothes, jewellery,
travelling, private transportation, telephone, home equipments with/and without motors.
Table 3. Main expenditure categories and subcategories of consumers’ spending in recession period
comparing with the time before. The (-) sign indicate decreasing spending during recession period
compared with the time before and the (+) sign increasing spending
Items of Consume Max Median t-value of Result
Before After Before After wilcox.test Change/Not
1. Food Food for home 8000 7000 1000 1000 0.006 C (+)
consumption
Service: fast food, take- 4000 3600 500 400 0.5741 N
out, delivery, buffet and
cafeteria (away from
home)
2. Beverages Alcoholic beverages 2000 2000 50 50 0.0004 C (-)
and Tobacco Non-Alcoholic 2200 2300 200 200 0.2696 N
beverages
Tobacco & smoking 600 700 0 0 0.9546 N
products
3. Clothing Casual, sportswear, 2000 2000 167 150 0.0071 C (-)
products formal, undergarments
and sleep clothes
Hat, comforter, glasses 500 700 20 20 0.5256 N
Shoes, sandal, slipper 800 1500 50 50 0.5725 N
4. Tools & Private tools (tower, 800 900 70 80 0.4239 N
health-care paper handkerchief,
services shaver, tampon...)
Cleaning & beauty 800 800 50 50 0.1230 N
products (shampoo,
soap, detergent,
shower-cream,
perfume, lipstick…)
Jewellery 1500 1500 0 0 0.0495 C (-)
Beauty service (spa, 1000 1000 0 0 0.5020 N
beauty salon…)
5. Cultural Publication 800 600 90 90 0.1998 N
activities, Travelling 2000 2000 0 0 0.0039 C (-)
sport, Sport activities (tennis, 700 800 0 0 0.3263 N
recreational swimming,
facilities badminton…)
Recreational activities 1200 1200 70 50 0.0624 N
(cinema, theatre, game,
bar, karaoke, go
fishing…)
6. Public transportation 500 500 0 0 0.1142 N
Transportati Private transportation 1000 1000 300 250 0.0328 C (-)
on, post Letters, package 400 700 0 0 0.4017 N
SPISE2009 66

Telephone: home 1100 1100 200 200 0.0282 C (-)

phone, mobile, fax
7. Medical- Medical, hospital and 1500 1500 0 0 0.2551 N
care related services
8. Education Foreign Language 2000 3000 0 0 0.3559 N
Intensive-course to 5000 4000 0 0 0.1018 N
enrich your career
9. Gift 700 900 0 0 0.2905 N
Communicat Party (marriage, death 1500 1500 300 300 0.0003 C (+)
ive anniversary…)
relationship
10. Home With motors (washing, 2500 2700 50 50 2.3e-05 C (-)
equipments machine, fan,
television, freeze)
Without motors 5000 1500 30 10 0.0002 C (-)
(furniture, porcelain...)
11. Housing, Payment for rent 4500 5000 0 0 0.0001 C (+)
building Repair, construct 3500 3000 0 0 0.1563 N
materials (building arterials,
paint, lime)
Immovable (house, 6000 6000 0 0 1 N
apartment, land…)
12. Saving Saving money (in bank, 10000 10000 1000 1000 0.5725 N
at home)
Take out insurance 5000 5000 0 0 0.7865 N
Stock investment 4200 4200 0 0 0.4227 N

The spending distribution of each sub-category is shown in Figure 1. Three items account for
approximately 40% of participants’ incomes: food at home (18-21%), saving (15-16%) and
food away from home (7-8%). Besides, consumers spent about 2-5% of their incomes for the
following sub-categories: alcoholic beverages, clothes, jewellery, travelling, private
transportation, telephone (home phone, mobile and fax), home equipments (with and without
motors), payment for rent and party (communicative relationship).
The way consumers spend for their needs: The spending sub-categories were grouped to form
the five following groups of needs (Figure 2):
 Basic needs including expenditures of food, beverages & tobacco, and clothing
products;
 Luxury needs including tools & health-care services, and cultural activities;
 Future needs including medical care, education, and communicative relationship;
 Transportation needs, including transportation and post;
 Real properties need including housing and savings.
SPISE2009 67

Sub-categories 0,00% 5,00% 10,00% 15,00% 20,00% 25,00%

( %)
19,18%
Fo o d fo r ho me co nsumptio n 20,56%
Saving.M o ney 15,98%
15,29%
Fo o d away fro m ho me 8,86%
8,72%
5,21%
P ayment fo r rent 5,61%
P arty 4,94%
5,34%

Before
4,72%
P rivate Transpo rtatio n 4,67%
4,28%
P ho ne 4,16%
3,55%

During
Intensive co urses 3,94%
No n-A lco ho lic beverages 3,25%
3,43%
3,43%
Casual, spo rtswear ... 3,14%
2,68%
A lco ho lic beverages 2,46%
Insurance 2,14%
2,17%
Fo reign Language 2,01%
2,16%
2,72%
Ho me Equipments with M o to rs 1,75%
Recreatio nal activities 1,65%
1,54%
P rivate To o ls 1,48%
1,50%
1,45%
Cleaning & beauty pro ducts 1,49%
Sho es, sandal & slipper 1,33%
1,38%
1,41%
P ublicatio n 1,35%
Traveling 1,57%
1,34%
1,09%
Gift 1,13%
Ho me Equipments Witho ut M o to rs 1,26%
0,98%
Ho use: repairing/Co nstructing 1,02%
0,96%
0,88%
M edical & related-service 0,86%
Spo rt activites 0,69%
0,75%
Hat, co mfo rter & glasses 0,65%
0,63%
0,49%
To bacco 0,52%
0,48%
B eauty service 0,52%
Immo vables 0,42%
0,50%
0,43%
Jewelry 0,35%
P ublic transpo rtatio n 0,36%
0,35%
0,32%
Sto ck investment 0,32%
Letter, package 0,09%
0,13%

Figure 1. Distribution of consumers’ spending

On the whole, Vietnamese consumers spend the same amount of their income in basic needs
before and during the recession (39.87% before and 40.83% during the recession period). Yet
results from the focus group suggested a decrease in basic needs during the recession.
Participants declared cutting down on: food away from home, new clothes, alcoholic
beverages, out of home entertainment. So, the question is: why do consumers think they
spend less on basic need during recession when in fact they declare spending the same
amount in a questionnaire?
SPISE2009 68

0,00 5,00 10,00 15,00 20,00 25,00 30,00 35,00 40,00 45,00
% % % % % % % % % %
(%)
39,87%
Bas ic needs
40,83%

29,05%
Real property needs
27,59%
s ub-category

12,47% Before
Future needs
13,42% During
Trans portation 9,45%
needs 9,32%

9,16%
Luxury needs
8,85%

Figure 2. Spending distribution based on needs of Vietnam consumers in recession period

compared with the time before recession.
Global spending on “Luxury need” and “Transportation need” were also very stable (9.16%
and 9.45% respectively), and “Real property needs” declined slightly from 29.05% to
27.59%. This result was expressed in the focus group sessions by: items such as cutting down
on travelling, private transportation, jewellery, investing in real property. This suggests that
consumers will try to cut down spending as much as they can. Interestingly, despite saving
behaviour; “Future needs” does not decline; or even, it increases very slightly from 12.47%
to 13.42%. It showed that Vietnamese consumers are optimistic enough to invest in health and
education (i.e. invest for future).

4. CONCLUSIONS
Like consumers all over the world, Vietnamese consumers feel their lifestyle has been
impacted by global economic recession. They fear job loss and price increases. So, they have
to change their behaviour by cutting down spending on some sub-categories (alcoholic
beverages, new clothes, food away from home, entertainment, telephone expense, jewellery,
private transportation, travelling, telephone, home equipments). However, Vietnamese
consumers are optimistic. They invest in future (health and education). And they expect that
Vietnamese economy will recover at the end of 2010 with Government’s stimulating package.

REFERENCES
Barrios, X.E., Bayarri, S., Carbonell, I., Izquierdo, L., & Costell E., (2008). Consumer
Attitudes and opinions toward functional foods: a focus group study, Journal of Sensory
Studies 23, 514-525.
SPISE2009 69

International Monetary Fund (2009). World Economic Outlook: Crisis and Recovery, World
Economic and Financial Survey. Retrieved April, 2009 from http://www.imf.org/external
/pubs/ft/weo/2009/01/
Ives, W. (2003). Focus group in market research, FOCUS GROUP: Supporting Effective
Product Development. Edited by Joe Langford and Deana McDonagh, Taylor & Francis,
p.47-58.
McGregor J. (2009). Consumer Spending in a Recession, BusinessWeek. Retrieved April 30,
2009 from http://www.businessweek.com/careers/managementiq/archives/2009/04/
consumer_spendi.html
Nielson (2008). Consumer confidence, Concerns, Spending and Attitudes to Recession.
Retrieved December, 2008 from http://vn.nielsen.com/site/documents/NielsenGlobal
Consumer Confidence ReportDec08.pdf
Shimelse, A., Uri, D., & Lauren F. (2009). Financial Transmission of the Crisis: What’s the
Lesson, International Economic Bulletin. Retrieved June, 2009 from
http://www.carnegieendowment.org/publications/index.cfm?fa=view&id=23284
Song, N. (2009). City’s GDP per capita soars to over US$2,500, TheSaigonTimes. Retrieved
April 1, 2009 from http://english.thesaigontimes.vn/Home/business/vietnam-economy/
2704
SPISE2009 70

EFFECTS OF AGE AND GEOGRAPHICAL ORIGIN IN PREFERENCE OF

CONSUMERS FOR YAM TUBERS AND MIX WHEAT YAM FLOURS COOKIES
B. Ranaivosoaa* D. Valentinb V.H. Jeannodac and J.L. Razanamparanya
(a) LBASAN, Université d’Antananarivo. Madagascar
(b) UMR CSG 5170 CNRS INRA, Université de Bourgogne. France
(c) LBEV, Université d’Antananarivo, Madagascar
*Corresponding author: [email protected]

Abstract
Yams (Dioscorea spp.) are important root crops cultivated in Madagascar and constitute an
important source of calories. Yet, despite a 45% rate of malnutrition prevalence, yam is part
of dietary customs only in the oriental and southern regions of Madagascar. The main
objective of this study is to evaluate whether the consumption of yam and yam derived
products could be extended to other areas of Madagascar in particular to urban areas. In a first
experiment we showed that among the four varieties of yam the most cultivated in
Madagascar (D. esculenta, and three cultivars of D. alata spp.) D. esculenta and to a lesser
extend D. alata. ovilalaina had the best nutritional and sensory potentialities. In the main
experiment, we used D .esculenta and D. alata. ovilalaina to make cookies with different
wheat/yam proportion of dry flour going from 100% wheat 0% yam (reference) to 50% wheat
50 % yam. Preference tests were realised in an urban community and a rural community. For
each region, 60 adults and 60 children of school age evaluated 11 cookies on a 9-point
hedonic scale. A three-way ANOVA showed a significant cookies  origin interaction:
Participants from rural community liked best cookies with up to 20% yam flours whereas
participants in urban community appreciated more cookies with 0% yam flavour and an
interaction between age and origin: In rural community adults appreciated more cookies than
children and inversely in urban community.
Key words: Madagascar; Yam; Nutrition; Sensory evaluation; cookies.

1. INTRODUCTION
Yams (Dioscorea spp.) are an important source of carbohydrate for many people of the sub-
Sahara region. Globally, roots and tubers are the second staple food after cereals, with a total
annual production of about 626 million tones, including 153 million in Africa (Oluwatoyin et
al., 2007). In Madagascar, the flora of yams, Dioscorea (or “oviala” in Malagasy) is
particularly rich with probably more than 40 species, about one-tenth of the Flora of
Dioscoreaceae in the world. Several studies were conducted on domesticated varieties to
improve the availability and potential use in food yam tubers. In Madagascar, these studies
focused on the ecological diversity, the traditional use and the development of sustainable
management systems. They showed a large diversity of cultivars in the Eastern, humid and
Western dry regions of Madagascar. Yet, despite a 45% rate of malnutrition prevalence, yam
is part of dietary customs only in the Oriental and Southern regions.
SPISE2009 71

The main objective of this study is to evaluate whether the consumption of yam and yam
derived products could be extended to other areas of Madagascar in particular to urban areas.
More precisely we evaluated the effects of age and geographical origin on consumers’
preference for yam tubers and mix wheat yam flour cookies. Our approach was twofold. First,
we determined the nutritional and hedonic potentialities of the four varieties of yam, the most
cultivated in Madagascar: D. esculenta, and three cultivars of D.alata spp. Then, we selected
the two varieties with the highest nutritional and hedonic potential and used them to make
cookies with different wheat/yam proportion of dry flour going from 100% wheat 0% yam
(reference) to 50% wheat 50 % yam.
Proteins, carbohydrates and starches contents in floor largely influence the structural and
textural properties of derived products (Nindjin et al., 2006). The mix optimal threshold of
yam flours would thus be function of mainly proteins contents and other components which
contribute to form dough elasticity and firmness properties as well as overall taste. A partial
substitute of the wheat flour by starchy products (tubers and cereal) constitutes an interesting
alternative in terms of cost profit, reason for which FAO proposed flours program (FAO,
1999). To evaluate the effect of age and geographical origin on the optimal threshold of yam
flours that can be incorporated without decreasing hedonic responses, we carried out a series
of preference tests in an urban community and a rural community. For each community, 60
adults and 60 children of school age evaluated 11 cookies (the reference, 10 cookies
incorporating respectively 10%, 20%, 30%, 40%, and 50% of D.esculenta or D.alata.
ovilalaina dry flour) on a 9-point hedonic scale.

2. EXPERIMENT 1: NUTRITIONAL AND SENSORY POTENTIALITIES OF FOUR

VARIETIES OF YAM TUBERS
2.1 Material and method
Plants materials and samples harvest site: Brickaville, a rural region in the Middle Eastern of
Madagascar (18°48’S et 49°04’E), was the harvest site of D.alata spp, D.esculenta. Several
batches of each variety were harvested in five villages of producer region: Razanaka,
Fanasana, Andevoranto, Anivorano, Lohariandava (Figure 1).
Nutritional characteristics of yam tubers: Samples were peeled, washed, sliced into cubes
which were freeze-dried using an Edwards bench freeze-drier (Edwards Instruments Ltd.,
Hornchurch, Essex, UK) and ground in a Hammer mill (Christy and Norris Ltd., UK) into
flour to pass through a 250µm sieve. The samples were analysed in triplicate for moisture,
ash, crude fat, crude protein contents and minerals using Association of Official Analytical
Chemist Approved methods (AOAC, 2000). Carbohydrate content was estimated by
difference.
SPISE2009 72

Figure 1: Harvest site

Hedonic ratings of yam tubers: For each variety, tubers were peeled and cut into pieces of
approximately 0.05 Kg and cooked in water (P/V: 1/1), T°= 45°C during 20 to 25 min.
Hedonic tests were realised using a 9-point hedonic scale labelled in Malagasy and illustrated
with smiley (Figure 2) as well as four 7-point intensity scales (Tableau 1) Participants were
recruited in three regions: a producer region 18°48’S et 49°04’E, an urban community
18°52’S et 47°31’E and a rural community18°46’ S et 47°34’ E. For each region, 60 adults
and 60 children of school age participated in the tests. In each region, each age group was
balanced for gender.

1 2 3 4 5 6 7 8 9

Tsy ankafizina Tsy tena TsyAnkafizina Tsy Antonony Ankafizina Ankafizina Tena Ankafizina
mihitsy ankafizina (Désagréable) Ankafizina (Ni agréable, kokoa (Agréable) ankafizina indrindra
(Extrêmement (Très kokoa ni (Plutôt (Très (Extrêmement
désagréable) désagréable) (Plutôt désagréable) agréable) agréable) agréable)
désagréable

Figure 2: Liking scale

Table 1: four intensity scales used in the yam consumer test.
Attributes left anchor right anchor
Faharenesana tsiro mamy Tsy mamy mihitsy Mamy be
Sweetness not sweet very sweet
Toetoetra Tena mafy Tena malemy
Firmness very firm not firm
Loko Tena matroka Tena mazava
Colour very dark very light
Fanatelomana Tena sarotra atelina Tena mora atelina
Ease to swallow difficult to swallow easy to swallow
SPISE2009 73

2.2. Results
Results were analysed using a Multiple Factor Analysis (MFA, Escofier & Pagès, 1998) with
yam varieties as observations and nutritional and consumer data as variables. MFA is a
generalisation of principal component analysis (PCA) that allows for the simultaneous analysis
of multiple data sets. Its main goal is to compare the different data sets and to combine them
into a common structure called a compromise. The subspace defined by the first two principal
components explains around 90% of the variance (52% and 38 %, for PC 1 and 2,
respectively) and therefore we limited our interpretation to these dimensions (Figure 3). The
first PC is strongly correlated with liking and opposes the two varieties D. Ovilalaina and D.
esculenta which were preferred by all consumers for their sweetness and firmness as well as
characterised by high calcium, proteins, carbohydrates, starch and calories contents to the D.
Ovy fotsy and D. Ovibe varieties. These last two varieties were richer in fibres and in humidity
with a lighter colour but were not as appreciated as the other ones.

PC 2 - 38.00 % easy to swallow PC 2 - 38.00 % nutritional data

0.8 nutritional data

1 D.Ovy fotsy
D.Ovilalaina

0.4 Ca H consumer data

nutritional data
consumer data
PT 0
D.esculenta
consumer data
firm consumer data
0 light color
sweet
liking -1

-0.4 GT D.ovibe
AMIDON FIBERS
-2
CAL PT
-0.8
ASH MG nutritional data
Na
-0.8 -0.4 0 K Mg 0.4 0.8 -1.50 -0.75 0 0.75 1.50
PC 1 - 52.25 % PC 1 - 52.25 %

Figure 3: Projection of the nutritional and consumers data sets (left panel) and yam varieties (right
panel) onto the MFA first two principal components.

A finer analysis of the raw data indicates that D. esculenta is characterised by higher contents
in proteins (6, 7 g100g-1db), in carbohydrates (90.6 g100g-1db) and is less fibrous (7.4 g100g-
1
db) than all other varieties. For all consumers, D. esculenta and in a lesser extend D.alata
ovilalaina obtained the best liking scores. Additionally a three-way analysis of variance
(ANOVA) with origin and age as within-subject variables and yam variety as between-subject
variable revealed a significant effect of origin and age on the global appreciation of yams.
Adults coming from a region where yams are cultivated (Brickaville) liked more the four
varieties used in this test than other consumers.
SPISE2009 74

3. EXPERIMENT 2: EFFECT OF AGE AND ORIGIN ON PREFERENCE FOR

WHEAT AND YAM COOKIES
3.1 Material and method
Formulation and preparation of cookies: D. esculenta and D. alata. ovilalaina were used to
make cookies with different wheat/yam proportion of dry flour going from 100% wheat 0%
yam (reference) to 50% wheat 50 % yam (Table 1). Ingredients used were dry yam wheat
composite flours (45 %), caster sugar (27%), shortening (20%), powdered milk (2.25%),
NAHCO3 (0.50%), NaCl (0.42%) and 8.5ml of water. Cookies were baked at 200°C for 10 to
15 min.
Table1: Wheat/yam flour formulation
cultivar Botanical name Code Formulation
Triticum aestivum B Wheat flour (100%)
D.esculenta E10 Wheat flour (90%) / yam flour (10%)
D.esculenta E20 Wheat flour (80%) / yam flour (20%)
Mavondro D.esculenta E30 Wheat flour (70%) / yam flour (30%)
D.esculenta E40 Wheat flour (60%) / yam flour (40%)
D.esculenta E50 Wheat flour (50%) / yam flour (50%)
D.alata A10 Wheat flour (90%) / yam flour (10%)
D.alata A20 Wheat flour (80%) / yam flour (20%)
Ovilalaina D.alata A30 Wheat flour (70%) / yam flour (30%)
D.alata A40 Wheat flour (60%) / yam flour (40%)
D.alata A50 Wheat flour (50%) / yam flour (50%)

Preference tests: Two tests were realised, one in an urban community and one in a rural
community. For each region, 60 adults and 60 children of school age evaluated 11 cookies,
(the reference and 10 cookies incorporating respectively 10%, 20%, 30%, 40%, and 50% of
D. esculenta or D. alata. ovilalaina dry flour on the same 9-point hedonic scale as in
experiment 1.

3.2 Results and discussion

A three-way ANOVA with origin and age as between-subject variables and type of cookies as
within-subject variable was carried out. Results showed a significant effect of age and origin:
 children (M = 6.34) appreciated more cookies than adults (M = 6.23), F (1, 236) =
31.74, p<.001.
 rural consumers (M = 6.86) appreciated more cookies than urban consumers (M =
5.71), F (1, 236) = 3368.24, p < .001.
However, the main effect of age should be interpreted with some caution since this variable
interact significantly with consumers’ origin, F (1, 236) = 229.87, p < .001. In rural
community, adult consumers (M = 6.96) appreciated more cookies than children (M = 6.77)
whereas in urban community, children (M = 5.91) appreciated more cookies than adult
consumers (M = 5.49). Moreover, these two variables interact significantly with the type of
cookies, F (10, 236) = 15.18, p < .001. To analyse further this three ways interaction we
carried out LSD mean comparison tests on each consumer group separately (see, Figure 4).
SPISE2009 75

The main point that can be noted on Figure 4 is that the optimal threshold of yam flours that
can be incorporated without decreasing hedonic responses depends on the origin of the
consumers and to a lesser degree on their age. All urban consumers significantly preferred the
100% wheat cookie (reference) whereas rural consumers significantly preferred the cookies
with 20% of yam flour. Although small significant differences appear between the two
varieties of yam at all incorporation levels, for adult consumers this effect of yam variety
remains smaller than that observed between successive incorporation levels. In contrast,
children, and in particular urban children seem to be more sensitive to the difference between
the two varieties of yam than between two successive levels of incorporation. The higher
preference for cookies made from the D. esculenta variety might be due to the smaller contain
in fibres and the higher content in sugar of this variety compared to the Ovilalaina variety.
However further analyses are needed to support this interpretation.

Adult - Rural community Adult - Urban community

10 10
Esculenta
8 e ba c d ef f g h 8 a b c Ovilalaina
i j de f g
6 6
hi i j
4 4

2 2

0 0
Ref 10% 20% 30% 40% 50% Ref 10% 20% 30% 40% 50%

Children - Rural community Children - Urban community

10 10
a b a b
8 c c c 8 c
de e f g h d e d
6 6 f g h hi
4 4

2 2

0 0
Ref 10% 20% 30% 40% 50% Ref 10% 20% 30% 40% 50%

Figure 4: average liking scores for the cookies as a function of the proportion of yam flour
incorporated in wheat flour. The letters represent the results of the LSD test at = .05.

The fact that consumers from rural communities used to consume yam tubers have a higher
optimal threshold of yam flours that can be incorporated in cookies than consumer from urban
communities rarely exposed to yam tubers can be put in perspective with cross cultural
studies on food preference. For example, Prescott (1998) in a program comparing Japanese
and Australian consumers highlighted that familiarity with a product influenced the
assessment of individual characteristics of the product such as sweetness and saltiness.
Japanese and Australian consumers differed in their liking of salty and sweet products: They
tended to find too salty or too sweet products from the other country when these products
SPISE2009 76

were considered about right by consumers from the country. A more recent study, conducted
in France and in Pakistan, showed that biscuits coming from a country are better scored by the
panellists coming from this country than biscuits coming from the other country (Pagès,
Bertrand, Ali, Husson, & Lê, 2007).

IV. CONCLUSION
This study has allowed evaluating the effects of age and geographical origin in preference of
consumers for mix wheat yam flour cookies. D. esculenta is characterised by higher contents
in proteins, in carbohydrates and is less fibrous than other varieties. Preference tests realised
in a producer region, an urban community and a rural community with 60 adults and 60
children of school age in each region showed that although consumers in producer region and
in rural community appreciated more yam tubers than urban consumers. All consumers liked
best D. esculenta and to a lesser extend D. alata. ovilalaina. Participants from rural
community liked best cookies with up to 20% yam flours whereas participants in urban
community appreciated more cookies with 0% yam flavour and in rural community adults
appreciated more cookies than children and inversely, children appreciated more cookies than
adults in urban community. In conclusion, age and origin effects were observed in global
preference of consumers: Rural consumers appreciated more yam tubers and cookies with mix
wheat yam flours than urban consumers. The present study inscribes itself in the frame of
preliminary studies that aim to develop a better knowledge of quality components for yam
flours derived products and of consumer expectations. The tool presented here using
consumer tests could eventually facilitate the design and evaluation of improved and market
oriented production for yam tubers and dry yam flour derived products.

Acknowledgments: The authors are grateful to CORUS 6020 project for financial support.
Thanks are also due to the inhabitants of Brickaville, Talatan’i Volonondry, Antananarivo
Centre, for making preference tests possible.

REFERENCES
A.O.A.C. (2000). Official Methods of Analysis, 17th ed. Association of Official Analysis
Chemists (A.O.A.C.). Washington, DC,USA.
Escofier, B. & Pagès, J. (1998). Analyses factorielles simples et multiples: objectifs, méthodes
et interprétation. Paris: Dunod.
FAO (1999). Protein quality evaluation report of a joint FAO/WHO expert consultation. FAO
Food and Nutrition Conference. Rome: FAO.
Nindjin, C., Otokore, D.A., & Hauser, S. (2006). Determination of relevant sensory properties
of pounded yams (Dioscorea spp.) using a locally based descriptive analysis
methodology. Food quality and preference,18, 450-459.
Oluwatoyin, A.,Odeku, Katharina, & M.,Picker-Freyer (2007). Analysis of the material and
tablet formation properties of four Dioscorea starches. Starch/Stärke, 59, 430-444.
SPISE2009 77

Pages, J., Bertrand, C., Ali, R., Husson, F., & Le, S. (2007). Sensory analysis comparison of
eight biscuits by French and Pakistani panels. Journal of Sensory Studies, 22(6), 665-686.
Prescott J. (1998). Comparison of taste perceptions and preferences of Japanese and
Australian consumers: Overview and implications for cross-cultural sensory research.
Food Quality and Preference, 9, 393-402.
SPISE2009 78

Part 3: Application in the food industry / Product development 
SPISE2009 79

GAINING INSIGHT INTO MARKETING STRATEGIES AND RETAILER

PERCEPTIONS OF US BEEF IN VIETNAM: A FOCUS GROUP APPROACH
T. T. N. Dinha*, D. H. Nguyenb, S. S. Harpa, L. D. Thompsona, J. C. Brooksa, M. F. Millera,
J.C. Boycea, D. B. Reeda, K. C. Lea, S. F. Maxnera, and J. L. Luskc
(a) Texas Tech University, Lubbock, TX, U.S.A. 79409
(b) HCMC University of Technology, Ho Chi Minh City, Vietnam
(c) Oklahoma State University, Stillwater, OK, U.S.A. 74078
*Corresponding author: [email protected] ; Tel: (806) 742 2805 (ext 265)

Abstract
Vietnam is considered an emerging market for the United States (US) beef industry. However,
little is known about the perceptions of Vietnamese retailers and consumers regarding US
beef. As a result, the US Department of Agriculture Emerging Markets Program funded a US
Beef Export School with the dual purpose of (a) enhancing awareness of US beef agribusiness
“from-farm-to-table” in the Vietnam retail sector as well as (b) assessing marketing strategies
and ascertaining retailer perceptions regarding US beef using a focus group approach. A
multi-disciplinary team from Texas Tech University, Oklahoma State University and Ho Chi
Minh City University of Technology with support from the US Meat Export Federation and
the National Cattlemen’s Beef Association collaborated in forming strategic alliances between
the US and Vietnam with the purpose of enhancing export of US beef to Vietnam.
Data were obtained from a convenience sample of US Beef School participants. Two focus
group interviews (N = 11) were conducted with executives representing the food retail sector
in the northern and southern regions of Vietnam. A discussion guide was produced in English,
translated into Vietnamese, and back-translated into English. A bilingual member of the
research team led the participants through each 90-minute discussion session. The
professional audio recording of each session was transcribed into English, and a note-CD-
transcript-based content analysis was prepared. The focal point of this presentation is the
summary of the major themes resulting from the content analysis: (a) factors affecting beef
import decisions, (b) perceptions of US beef, (c) purchase of US beef by Vietnam retailers for
importation, and (d) implications for the US beef industry.
Keywords: US beef, export, focus group, Vietnam, retailer perception.

1. INTRODUCTION
Penetration of United States (US) agricultural exports typically depends on high-income
markets such as Japan and the European Union (USDA/ERS, 2007). Recently, significant
and steady economic growth in developing countries such as Vietnam has created a number of
emerging markets that offer opportunities for growth of US agricultural exports, especially
high quality meat products. Identified as one of the world’s rapidly developing markets,
Vietnam has an increasing demand for high quality beef. Drivers of this demand include:
SPISE2009 80

increasing population, rising income and spending on food, limited resources for live cattle
production, and consumer concerns regarding safety and quality of domestically produced
beef. From 2005 to 2007 export statistics indicate more than a 100% increase in importation
of US beef to Vietnam (USDA/ERS, 2008).
In 2008, about 8,000 Vietnamese retailers, restaurant operators, chefs, and distributors
sampled US beef at a food show in Ho Chi Minh City and most participants expressed an
interest in purchasing beef imported from the US (USMEF, 2008). A review of the scholarly
and industry literature revealed a void in the knowledge of Vietnamese retailer and consumer
perceptions regarding US beef. Further, research has shown social and cultural differences
can create varied perceptions of and demand for foods including factors such as source, price,
quantity, and quality of meat (Resurreccion, 2004; Kizilaslan et al., 2008; Glitsch, 2000;
Ackerman & Tellis, 2001; Grunert, 2006).
It has been well documented that product quality is determined not only on intrinsic and
extrinsic cues at the point of purchase, but also on opinions formed during product
preparation and/or consumption when quality expectations are confirmed or rejected (Bredahl,
2004; Banovic et al., 2009). As in all cultures, Vietnamese have distinct methods of meat
preparation and cooking. Many traditional dishes utilise tougher cuts of beef such as flank
steaks, shanks, or tendons and wet cooking methods are more common in Vietnam than the
dry cooking techniques (Nguyen, 2006). In this context, targeted product selection, and
fabrication and packaging strategies for US beef are essential in the development of marketing
plans that enhance the export of value-added components of US beef. Additionally, it is
customary for Vietnamese consumers to purchase meat on the day of preparation at local
fresh markets or retail food outlets, resulting in a “just in time” marketing strategy for US beef
in the wholesale and retail market sectors in Vietnam.

2. OBJECTIVES
In an effort to narrow the gap of understanding in the US beef industry regarding the Vietnam
retail meat market, the US Department of Agriculture Emerging Markets Program
(USDA/EMP) funded a US Beef Export School to enhance awareness of US beef agribusiness
“from-farm-to-table” in Vietnam retail sectors as well as to assess marketing strategies and to
ascertain retailer perceptions regarding US beef using a focus group approach. As part of the
US Beef Export School, this preliminary research was conducted to determine: (a) the
perceptions of Vietnam retailers regarding US beef, and (b) factors important in the
development of viable marketing strategies for imported US beef.

3. METHODOLOGY
Focus group interviews were selected as an appropriate method for generating a breath of
insight regarding US beef and import potential in a short time frame. Qualitative research
methods for focus group sessions followed guidelines described by Krueger and Casey
SPISE2009 81

(2009). The study was approved by Texas Tech University (TTU) Institutional Review Board
for Protection of Human Subjects. All instruments and forms (i.e., moderator guide, consent
form, orientation guide) were developed in English, translated to Vietnamese, and back-
translated to English.

3.2 Participants and group formation

Two focus group interviews were conducted in May 2009. The participants were executives
representing four companies in the food retail sector in the northern and southern regions of
Vietnam attending the US Beef Export School. Six participants comprised the first focus group
and five other participated in the second focus group session. Participants were assigned to
one of the focus group sessions based on company affiliation and position within their
company to avoid the possibility of undue influence or dominance of higher ranking co-
workers.

3.3 Procedures and questions

The focus group interviews were conducted in a conference room on the TTU campus and
lasted 90 minutes with an intermission halfway through. The sessions were audio recorded by
a professional audio technician using state-of-the-art, industry quality digital technology to
ensure accurate transcription of individual comments and accurate recognition of voice
inflection. The same bilingual moderator, the first author, conducted the two focus group
interviews. Prior to conducting the two focus groups, the moderator received training by
observing and assisting in two focus group interviews conducted by a professionally trained
and experienced focus group moderator. The second, third, and fourth authors assisted in the
focus groups and observed to ensure that similar procedures were followed across the two
interviews.
A two–part structured moderator guide was developed by the research team based on the
research objectives. A questioning route approach was used in the development of the
questions to be asked by the moderator in the focus group sessions to ensure a practical
structure for organising the discussion sequence and to facilitate the analysis of results
(Kruger & Casey, 2009). The questioning route in each section included uncued and cued
questions in open-ended and sentence format as suggested in Krueger and Casey (2009).

In the first 40 minutes, each group was led through a series of opening, introductory, and
transition questions. Opening questions were formulated to engage individuals in a
conversational dialogue and gain insight into the participants’ industry experience, company
ownership, and potential customers for imported beef. Introductory questions were designed
to obtain participants’ thoughts and opinions about product attributes as well as trading
policies and procedures considered important in beef import decisions. Transition questions
were developed to discover the participants’ perceptions and beliefs regarding major
SPISE2009 82

implications for their company associated with offering US beef to Vietnam consumers, and
specific characteristics ascribed to US beef.

Following a 15-minute intermission, the moderator addressed key and ending questions with
the participants in the final 50 minutes of each session. Key questions were constructed to
generate specific information regarding the form in which they anticipate purchasing product
(i.e., whole carcass, primals, subprimals, portion cuts), fabrication and packaging; and to
explore company and consumer procedures on distribution, wholesale and consumer storage,
and retail display of US beef. Ending questions were designed to gain closure to the
discussion, ensure no critical aspects related to export of US beef to Vietnam had been
overlooked in the discussion, and to brainstorm about future in-country activities the research
team should consider implementing to enhance export of US beef to Vietnam.
The focus group questions and procedures were reviewed by a panel of three members of the
research team experienced in teaching, research, and outreach in meat science as well as three
members of the research team experienced in planning, conducting, and publishing focus
group research. Based on comments and recommendations from the expert panel, questions
were clarified and procedures adjusted prior to conducting the two focus group interviews.
Prior to the beginning of each focus group sessions, participants were asked to sign a consent
form agreeing to be audio recorded, verifying they understood participation was voluntary
and that no penalties would be imposed if they choose not to participate. At the beginning of
each session, a member of the research team provided an overview of the discussion sequence
and topic focus. At the intermission the moderator consulted with the research team to
confirm the coverage of topics. At the conclusion of each session, the research team thanked
the participants for their time and input.
At the completion of the two focus groups, saturation of responses had been achieved: no new
perceptions or insights associated with the importation of beef from the US were mentioned
by the participants. Throughout the interviews, special care was taken to ensure that responses
were offered from all participants and that no single individual dominated the discussion.
Opinions were specifically solicited from any participant who appeared reticent to offer
suggestions.

3.4 Data analysis

Bilingual members of the research team prepared a verbatim transcript of each focus group
using detailed notes taken at the interview sessions and by listening to the audio CDs. The
content analysis followed a systematic process described by Krueger and Casey (2009).
When conducting the note-CD-transcript-based qualitative content analysis, certain words,
consistency, similarities, intensity and frequency of comments, and stand-alone responses
were taken into consideration. Instead of using numbers, explicit qualifiers were used for
interpreting agreement, disagreement, or level of consensus: “A Couple” = one or two
SPISE2009 83

participants; “Several” = at least 3, but less than 1/3 of participants; “Some” = at least ¼, but
not much more than 1/3 of participants; “A Fair Number” = at least 1/3, but less than 1/2 of
participants; “Evenly Divided “ = 1/2 of participants; “Many” = more than 1/2, but less than
2/3 of participants; “Most” = at least 3/4, but less than 90% of participants; “Almost All” = at
least 90%, but less than 100% of participants; and “Participants” = 100% of participants
(Harp et al., 1998; Assante et al., 2007; Templeton, 1996). When analysing the interviews,
the researchers looked to identify broad concepts or themes from the responses to questions.

4. RESULTS AND DISCUSSION

Two focus groups were conducted with 11 participants consisting of 7 males and 4 females.
The participants were Vietnamese executives in the food retail sector and meat industry with
position titles ranging from general director and CEO to production and quality manager. Job
responsibilities included general supervision, management, production supervision, and
quality implementation and assurance. All participants significantly contribute to
organisational policy-making and import decisions.

Participants represented five state-owned and one privately-owned company. Nine

participants were employed by the state-owned companies and two participants were
employed by the privately-held company. These companies represent a significant market
share of meat products in both the southern and northern regions of Vietnam. Four state-
owned companies and the privately owned company have marketed imported non-US beef for
domestic retailing or value-added processing. Participants from five companies (four state-
owned, one privately-owned) expressed a desire to import US beef in the near future. The one
remaining state-owned company wanted to obtain additional information on consumer
demand before committing to importing US beef. Most companies intend to market imported
US beef to high-income consumers or up-scale restaurants. One state-owned company has the
capacity and desire to sell US beef to wholesalers or other food retailers.
The content analysis which was based on the opinions, ideas, perceptions, and purchase
motives disclosed by the participants, evolved to include four major themes: (a) factors
affecting beef import decisions; (b) perceptions of US beef; (c) purchase of US beef by
Vietnam retailers for importation; and (d) implications for the US beef industry. These themes
are expanded upon in the discussion that follows.

4.1 Factors affecting beef import decision

Most participants thought that price and quality of beef were important criteria in their
decision to import beef. Across cultures, consumers expect an affordable price in relation to
quality. As retailers, in order to make profit selling meats, their opinions should reflect those
of their consumers. This finding is consistent with prior studies using consumer surveys and
focus group discussions, in which consumer perceptions of meat and other food products were
found to be dependent on intrinsic food properties, as well as the interaction between intrinsic
SPISE2009 84

properties and immediate extrinsic factors such as price and information (Dransfield et al.,
1998; Acebron & Dopico, 2000). Price is perceived as relevant cue to quality when
consumers do not have adequate information about intrinsic food properties. Although the
results may vary across studies, most indicated that price and quality are positively related
(Acebron & Dopico, 2000). Moreover, the term of “quality” expressed by the participants in
the present study often included safety. Scientifically, safety is a separate issue from quality.
However, safety is a critical attribute of food products across cultures. Consumers invariably
expect foods to be safe, and without safety, consumers will not purchase foods (Verbeke et
al., 2006).
Most participants indicated that in order to forge a desirable relationship between importer
and exporter, the exporter should provide marketing assistance, service, and technical support
to the importer. Many of the participants desired information about prospective exporters,
beef cuts, assurances of product availability, and information on consumer demand. It is
understandable that retailers and importers are cautious and need to be well-informed
regarding the purchase of beef because beef has a relatively higher price, requiring a larger
investment than other meats (Resurreccion, 2004).
Participants revealed that country-of-origin is an important factor in the import decision. The
term “county-of-origin” and “brand” were used interchangeable by almost all of the focus
group participants. Further in the discussion, the participants indicated that to Vietnamese
consumers US products are generally a symbol of reliability, quality, and safety. As such,
Vietnam consumers would be willing to purchase US beef at a higher price than domestic beef
or beef imported from countries other than the US. These observations were also noted by
Sepulveda et al. (2008), who found that region or origin of production was perceived as a
quality attribute. Banovic et al. (2009) also reported that quality cues related to beef brands
were highly and positively correlated with product origin; and brand of beef was found to be
the major extrinsic quality cue. Consumers used brand and origin as a determinant for both
perception of intrinsic quality attributes and quality expectations (Grunert, 2006). However, a
couple of participants noted that the country-of-origin influences purchasing, provided the
consumer has an awareness of the beef’s origin and related quality. Also relevant to the
importer are logistical and cost concerns associated with the geographic proximity of the
exporter to Vietnam.
Aging of meat, a common practice in the US beef industry, was not initially perceived as a
factor influencing importation decisions simply because all of the participants lacked an
understanding of the aging process and the impact on meat quality. Participants indicated
they need more information on meat aging before they could evaluate the importance of aging
in the decision making process. Additionally, participants felt that they would need to educate
the consumer about the benefits derived from the aging process. After the moderator
explained the aging process and its advantages, a fair number of participants were concerned
SPISE2009 85

that aging might affect safety, shelf life, fresh appearance, and cost. However, several did not
think aging would be relevant to the import decision, and indicated that they would decide
whether or not to import beef based on quality traits rather than on utilisation of a specific
technology or process.
When asked about grass-versus grain-fed beef, many participants thought it was an important
factor that could influence beef flavour. Again, a fair number of participants thought feed was
not important as long as the beef quality meets or exceeds consumer expectations.
Additionally, in the context of grass- vs. grain-feeding, a couple of participants said they
preferred the flavour of domestic beef, while a couple of others thought that US grain-fed beef
was very flavourful and that beef flavour would be attractive to Vietnam consumers. It is also
important to recognise that Vietnamese typically consume highly seasoned grass-fed beef
containing little marbling prepared by wet cooking methods. During the US Beef School and
prior to the focus group sessions, the participants consumed meals featuring highly-marbled,
dry-cooked (grilling) beef entrées. Grilling or dry cooking causes thermal oxidation of lipids
resulting in an intensified cooked beef flavour (Wood et al., 2004). Therefore, sensory
attributes of beef consumed at the US Beef School may have impacted their perception of the
flavour of US beef.
At this point, it was apparent that the greater the knowledge the participants had on a
particular factor that influences a quality attribute, the more important the factor became in
the decision-making process. Consequently, an increase in knowledge leads to more informed
decision-making. Quality perceptions of meat have been traditionally based on intrinsic
attributes such as colour and visible fat. Over time, fuelled with debate and information
regarding issues such as health and safety, consumers have relied increasingly on extrinsic
attributes to evaluate the perceived meat quality. Consumers in focus group discussions were
usually confident that they could judge the sensory quality of the meat without additional
information (Grunert, 2006). However, if provided with additional information, and the more
they understood the information, the more important the information became in their quality
perception (Grunert & Grunert, 1995). Once an extrinsic cue is established in the customer’s
mind, the effects on quality perception can be exceptional, as in the case of country-of-origin
as previously discussed. However, it is interesting to note that information is perceived as
least credible coming directly from producers (Poppe & Kjaernes, 2003).
All participants agreed that Vietnamese consumers evaluate the quality of beef on the
appearance (bright red colour) and texture (softness); therefore, prefer to purchase fresh rather
than frozen beef. Hence, beef imported in the frozen state must be thawed for retail display.
Meat colour has been well-documented as a predominant intrinsic factor indicating freshness
(Glitsch, 2000; Grunert, 2006; Grunert, 1997; Lynch et al., 1986). Lynch et al. (1986)
concluded that while the bright red colour of ground beef was important, informed consumers
were more likely to have positive purchase intent toward vacuum-packaged ground beef even
SPISE2009 86

though it was purple in colour and were as willing to buy the purple vacuum-packaged ground
beef as the bright red product. These findings support the belief of some participants that the
utilisation of educational materials at point-of-purchase could modify beef purchasing
patterns of consumers in Vietnam.

4.2 Perceptions of US beef

Participants immediately ascribed high quality to US beef. Most participants characterised US
beef as reputable and reliable in terms of quality and safety. Many thought US beef was
expensive although some participants considered US beef to be reasonably priced or
affordable (some cuts) in relation to value and some did not think about price. A fair number
of participants considered US beef to be tender, frozen, dry-cooked, and produced in
abundance. Almost all participants encountered US beef for the first time while attending the
US Beef Export School. Therefore, it is reasonable to conclude that their first impression of US
beef came primarily from eating high quality beef prepared at up-scale restaurants and by
caterers specialising in beef preparation, through experiential learning including field trips to
one of the largest meat packing plants and cattle feed yards in the US, and classroom activities
conducted by nationally-recognised industry representatives and academic beef specialists.
Participants agreed that offering US beef would have a significant positive impact on their
companies and retail stores. Because the US is recognised as a good trademark, the presence
of US beef might positively impact company image and increase the value of other goods
offered in their retail stores.

4.3 Purchase of US beef by Vietnam retailers for importation

Participants indicated they would purchase primals or subprimals depending on price.
However, some participants would also import quarters and special orders such as shanks for
traditional Vietnamese dishes. The special demand for low-value, underutilised cuts such as
shanks provides an opportunity for the creation of value-added products destined for meat
markets in Vietnam. All the participants would further fabricate and package imported beef
cuts. Participants immediately agreed that fresh beef would be their first choice because of
customer preference as previously discussed. The primary target markets in Vietnam for US
beef would be high-income consumers and up-scale restaurants. Participants recognised the
importance of having sufficient freezer and cooler systems for meat storage. Most
participating companies have central warehouses with freezer and cooler systems for storage,
and refrigerated trucks for meat distribution. Some companies depend on suppliers to deliver
meat to their retail stores. Most wholesale customers have their own freezer and cooler
systems while the ultimate consumer utilises refrigerator/freezer units for in-home meat
storage. The USMEF has expressed concern about the cold chain in Vietnam meat retail
markets. However, as reported by the participants, imported US beef would be sold in
supermarkets under conditions that would ensure proper preservation of quality. Further,
SPISE2009 87

some participants reported that their company provides in-store consumer information
regarding proper refrigeration of meat products.

Vietnamese customers shop for meat several time a week, purchasing single-meal portions.
They are not in the habit of buying meat in bulk or storing meat for long periods of time.
Instead, they prefer to buy meat when it is still “fresh at store.” Therefore, successful
retailers in Vietnam plan inventory assortments according to these shopping orientations. As a
result, Vietnamese retailers are more likely to buy portioned primals and subprimals than
retailers in the US. By customising the type and size of cuts retailers are in a better position to
achieve higher sell-through margins. In an interview on Meatingplace.com (Gabbett, 2009),
John Niemann, vice president for beef pricing, sales and business management at Cargill, one
of the largest meat packers in the US, revealed the company boosted beef sales by marketing
half-cuts and third- cuts to help retailers offer steaks at consumer-desirable price points.
Additionally, price is usually the extrinsic factor that is processed fastest when consumers
make a decision of whether or not to buy a package of meat (Grunert, 2004; Grunert, 2006).

4.4 Implications for the US beef industry

Most participants agreed that the US beef industry should provide technical support to
Vietnamese importers by providing adequate information on product quality, export
companies, and assist in organising experiential marketing, promotional events, and point-of-
purchase informative materials. In this context, participants repeatedly suggested that the US
beef industry could learn from the successes of US apple exporters in Vietnam. Most of
participants agreed that the initial introduction of US beef at the retail level should saturate the
market with a large-scale, national campaign. Additionally, some participants suggested that
the US beef industry should advise Vietnam importers regarding regulations and procedures.
Finally, some of the participants concluded that the US beef industry could best enhance
export opportunities by assisting Vietnamese retailers in conducting marketing and consumer
research focused on the purchase and consumption of beef in Vietnam.

5. IMPLICATIONS AND RECOMMENDATIONS

This research was an initial effort to explore retailer perceptions of US beef and the potential
for import penetration. Preliminary results suggest a positive business environment for
marketing US beef export to Vietnam. Further, the research team believes that the focus group
participants’ suggestions are industry-relevant and should be implemented. To validate these
findings, more focus groups with a sensory component in retail and foodservice sectors
should be conducted. Future quantitative research is needed to examine the qualitative
findings with a larger sample of retailers and consumers.

Acknowledgements. We would like to express our appreciation to the US Department of

Agriculture Emerging Markets Program for funding the project; and to the US Meat Export
Federation, the National Cattlemen’s Beef Association, the Texas Tech University
SPISE2009 88

International Center for Food Industry Excellence and participating US and Vietnam
producers, processors, wholesalers and retailers for supporting these efforts to enhance the
export of US beef to Vietnam.
REFERENCES
Acebron, L.B. & Dopico, D.C. (2000). The importance of intrinsic and extrinsic cues to expected and
experienced quality: an empirical application for beef. Food Quality and Preference, 11(3), 229-
238.
Ackerman, D. & Tellis, G. (2001). Can culture affect prices? A cross-cultural study of shopping and
retail prices. Journal of Retailing, 77(1), 57-82.
Assante, L.M., Huffman, L., & Harp, S.S. (2007). Conceptualization of quality indicators for U.S.
based four-year undergraduate hospitality management programs. Journal of Teaching in Travel
& Tourism, 7(2), 51-71.
Banovic, M., Grunert, K.G., Barreira, M.M., & Fontes, M.A. (2009). Beef quality perception at the
point of purchase: A study from Portugal. Food Quality and Preference, 20(4), 335-342.
Bredahl, L. (2004). Cue utilisation and quality perception with regard to branded beef. Food Quality
and Preference, 15(1), 65-75.
Dransfield, E., Zamora, F., & Bayle, M.C. (1998). Consumer selection of steaks as influenced by
information and price index. Food Quality and Preference, 9(5), 321-326.
Gabbett, J. (2009). Cargill executive discusses merchandising meat in the current economy.
Retrieved March 27, 2009 from http://www.meatingplace.com/MembersOnly/ webNews/
details.aspx ?item=11804.
Glitsch, K. (2000). Consumer perceptions of fresh meat quality: cross-national comparison. British
Food Journal, 102(3), 18.
Grunert, K.G. (1997). What's in a steak? A cross-cultural study on the quality perception of beef. Food
Quality and Preference, 8(3), 157-174.
Grunert, K.G. (2006). Future trends and consumer lifestyles with regard to meat consumption. Paper
presented at the 52nd International Congress of Meat Science and Technology (52in ICoMST),
Dublin, Ireland. Aug 13-18
Grunert, K.G., Bredahl, L., & Brunso, K. (2004). Consumer perception of meat quality and
implications for product development in the meat sector - a review. Meat Science, 66(2), 259-272.
Grunert, K.G., & Grunert, S.C. (1995). Measuring subjective meaning structures by the laddering
method: theoretical considerations and methodological problems. International Journal of
Research in Marketing, 12, 17.
Harp, S.S., Hoover, L.C., Crockett, K.L., & Wu, C.K. (1998). Development of a beef appetizer
concept for casual dining restaurants: application of focus group interviews and consumer sensory
evaluation. Journal of Hospitality & Tourism Research, 21(3), 18.
Kizilaslan, H., Goktolga, Z.G., & Kizilaslan, N. (2008). An analysis of the factors affecting the food
places where consumers purchase red meat. British Food Journal, 110(6-7), 580-594.
SPISE2009 89

Krueger, R.A., & Casey, M.A. (2009). Focus groups: a practical guide for applied research. (4nd ed.).
Thousand Oaks, CA: Sage Publications.
Lynch N.M., Kastner, C.L., & Kropf, D.H. (1986). Consumer acceptance of vacuum packaged
ground-beef as influenced by product color and educational-materials. Journal of Food Science,
51(2), 253.
Nguyen, A.Q. (2006). Into the Vietnamese Kitchen: Treasured Foodways, Modern Flavors. Berkeley,
CA: Ten Speed Press.
Poppe, C. & Kjaernes, U. (2003). Trust in food in Europe. Oslo: National Institute for Consumer
Research.
Resurreccion, A.V.A. (2004). Sensory aspects of consumer choices for meat and meat products. Meat
Science, 66(1), 11-20.
Sepulveda, W., Maza, M.T., & Mantecon, A.R. (2008). Factors that affect and motivate the purchase
of quality-labeled beef in Spain. Meat Science, 80(4), 1282-1289.
Templeton, J.F. (1996). The focus group: a strategic guide to organizing, conducting and analyzing
the focus group interview. Chicago, IL: Probus Publishing Company.
U.S. Meat Export Federation (USMEF). (2008). International markets – ASEAN. Retrieved February
10, 2008 from http://www.usmef.org/TradeLibrary/Asean.asp.
United States Department of Agriculture, Economic Research Service (USDA/ERS). (2007). Global
growth, macroeconomic change, and U.S. agricultural change. Retrieved February 10, 2008 from
http://www.ers.usda.gov/Publications/ERR46/.
United States Department of Agriculture, Economic Research Service (USDA/ERS). (2008).
Livestock and meat trade data - Beef and veal: Annual and cumulative year-to-date U.S. trade
(carcass weight, 1,000 pounds). Retrieved February 10, 2008 from
http://www.ers.usda.gov/Data/MeatTrade/ BeefVealYearly.htm.
Verbeke, W., Frewer, L.J., Scholderer, J., & De Brabander, H.F. (2006,). Why consumers behave as
they do with respect to food safety and risk information. Paper presented at the 5th International
Symposium on Hormone and Veterinary Drug Residue Analysis, Antwerp, Belgium. May 16-19
Wood, J.D., Richardson, R.I., Nute, G.R., Fisher, A.V., Campo, M.M., Kasapidou, E., et al. (2004).
Effects of fatty acids on meat quality: a review. . Meat Science, 66(1), 21-32.
SPISE2009 90

FORMULATION OF FLOUR BASED PEANUT SNACK USING MIXTURE DESIGN

N. Pengboon, S. Punsuwan and P. Siriwongwilaichat*
Department of Food Technology, Faculty of Engineering and Industrial Technology,
Silpakorn University, Amphur Mueng, Nakhon Pathom, Thailand
*Corresponding author: [email protected], Tel: +66.34.219361

Abstract
Flour based peanut snack is a well-known local fried product widely consumed in Thailand.
Its textural characteristic is one of the key determining factors influencing consumer
acceptance. The aim of this study was to formulate three flours combinations using mixture
design facilitated by Design Expert 6.0.5. Rice flour, wheat flour, and tapioca flour were
chosen for the study. A quadratic model was assumed for the variable design, resulting in 14
flour combinations so that the total flour was controlled at 25% w/w. Hedonic scoring and
Just-About-Right scale were employed for sensory evaluation of the product. The results
indicates that a flour combination consisting of 13% rice flour and 12% tapioca flour achieved
the highest liking score, satisfactory attributes and minimum cost.
Keywords: Flour based peanut snack, mixture design, hedonic scoring, just-about-right-scale

1. INTRODUCTION
Peanut (Arachis hypogaea L.), despite its name, is not a nut but belongs to the Leguminosae
family. It is planted and consumed worldwide. The main nutrients in peanut seeds include
protein (30%), fat (47%), and carbohydrate (12%) (http://th.wikipedia.org, 2008). In
Thailand, it is cultivated nationwide and used as ingredient for various snacks or main dishes.
Flour based peanut snack is of these locally consumed products. This product is flat-round in
shape with around 6-7 cm in diameter. The number of peanut seeds in each piece varies (e.g.,
from 8-20 seeds). It is mostly produced and sold daily by small business sellers. Nevertheless,
its recipe and process have not yet been widely studied and adapted for industrial production.

The main ingredients of flour based peanut snacks include flour, sugar, salt, egg, water, and
coconut milk. The brief processing steps include flour and sugar mixing, addition of salt, egg,
water, and coconut milk, filtration, addition of peanut followed by deep oil frying, cooling
and packing, respectively. Similar to any other crispy snacks, the textural characteristic of the
flour based peanut snack is one of the key attributes influencing consumer acceptance.
Numerous studies have been reported on flour blend effect, mostly focusing on soy flour and
corn flour incorporation in fried wheat flour based snack (Senthil, Ravi, Bhat, &
Seethalakshmi, 2002; Ward, Resurreccion, & McWatters, 1998; Ahamed, Singhal, Kulkarni,
& Pal, 1997). In this context, the amylose-amylopectin ratio of flour or flour blend plays a
major role because it determines the volume expansion and the texture of the finished snack.
Amylose content has also been reported to be associated with the extent of oil uptake in deep
fat-fried snack (Ahamed, Singhal, Kulkarni, & Pal, 1997). These quality attributes influence
SPISE2009 91

consumer perception and acceptance of the final product. In flour based peanut snack,
however, rice flour (83% amylopectin), wheat flour (72% amylopectin), and tapioca flour
(83% amylopectin) (Hizukuri, 1988) are normally used as mixed ingredients. Therefore, the
purpose of this study was to find the combination of these three flours which will result in a
satisfactory flour based peanut snack product. Hedonic and Just-About-Right (JAR) scales
were employed for sensory evaluation of the product.

2. MATERIALS AND METHODS

2.1 Product ingredients
The product contained dried raw peanut seeds, sesame, rice flour, wheat flour, tapioca flour,
water, coconut milk, egg, fine salt, sugar. Palm oil was used for frying.
2.2 Flour based peanut snack preparation
Flour base was prepared from a mixture of flours and sugar to which was incorporated salt,
egg, water and coconut milk. The combination was well mixed until the sugar was completely
dissolved before sieving through a strainer to remove large particles. Peanut seeds and sesame
were then added, and the mixture was transferred into the round-flat metal mould (diameter of
5.3 cm) to control the product shape and size. The product was subsequently deep fried in
palm oil at 180 °C for around 40 seconds, then at 140 °C for around 15 seconds. The fried
product was then let cool down to ambient temperature before being packed in laminated
aluminium foil bag.
2.3 Experimental design
Except for flours of which total amount was controlled at 25% w/w, all other ingredients were
fixed in product formulation. According to a preliminary experiment, the proportion range
was 3 to 13% for tapioca flour, 13 to 23% for rice flour and 5 to 13% for all purpose wheat
flour. Design Expert 6.0.5 (Stat-Ease, Inc.) was employed for mixture design with assumed
quadratic model, resulting in 14 flour combinations (Table 1).
Table 1. Flour combinations from mixture design.
Flour Combination Rice flour (%) Wheat flour (%) Tapioca four (%)
1 15.50 3.35 6.15
2 18.00 - 7.00
3 23.00 2.00 -
4 13.00 - 12.00
5 18.00 7.00 -
6 13.00 7.00 5.00
7 19.25 3.85 1.90
8 15.50 7.00 2.50
9 13.00 3.50 8.50
10 23.00 - 2.00
11 13.00 - 12.00
12 18.00 - 7.00
13 13.00 7.00 5.00
14 18.00 7.00 -
SPISE2009 92

2.4 Product evaluation

All product samples were evaluated in order to obtain liking scores (7-point hedonic scale)
and JAR scores (1 = Much too weak, 2 = Too weak, 3 = Just about right, 4 = Too strong, 5 =
Too much strong). Each of 40 panellists was served with one piece of product. Product
hedonic attributes included appearance, colour, flavour, sweetness, saltiness, and crispness.
Selected samples were evaluated for attribute intensity employing JAR. Colour, sweetness,
saltiness, and hardness were also evaluated.

3. RESULTS AND DISCUSSIONS

Regarding the hedonic sensory evaluation, the results indicate that combinations 1, 2, 4, 5, 6,
8, 9, 11, 12, and 13 did not differ significantly (see Table 2). We also found that crispness
contributed significantly to the overall hedonic score (see Table 3). As indicated in Table 2,
the flour combination 11 (rice flour 13%, tapioca flour 12%). received the highest score
which was not significantly different from the score of combination 13 (rice flour 13%, wheat
flour 7%, tapioca flour 5%). The results also suggest that rice flour at 13% constitute the right
proportion. The largest proportion of rice flour at 23% (combinations 3 and 10) tended to
lower the crispness hedonic score. This result could be explained by the influence of amylose-
amylopectin ratio of flour or flour blend as reported in the literature (Senthil, Ravi, Bhat, &
Seethalakshmi, 2002; Ward, Resurreccion, & McWatters, 1998; Ahamed, et al., 1997).
However, further investigation was made on flour combinations 11 and 13 employing JAR to
profile the product so that a specific improvement could be suggested.
Table 2. Hedonic scores (1 = Extremely dislike, 5 = Neither like nor dislike, 9 = Extremely like)
of flour based peanut snacks from 14 flour combinations. Letters indicate significant difference
amongst values in the same column (p < .05).
Mean liking score
Flour combination
Overall Appearance Colour Flavour Sweetness Saltiness Crispness
1 6.4ab 6.4bc 6.4a 6.2a 6.1ab 5.9a 6.0bc
2 6.4ab 6.4bc 6.4a 6.2a 6.1ab 5.9a 6.0bc
3 6.1bc 6.3c 5.9ab 5.8ab 5.6b 5.6a 5.8bcd
4 6.4ab 6.4bc 6.4a 6.2a 6.1ab 5.9a 6.0bc
5 6.4ab 6.4bc 6.4a 6.2a 6.1ab 5.9a 6.0bc
6 6.4ab 6.4bc 6.4a 6.2a 6.1ab 5.9a 6.0bc
7 5.9bc 6.1c 6.1ab 5.5ab 5.7ab 5.8a 5.8bcd
8 6.2abc 6.2c 6.7a 6.0ab 6.1ab 5.9a 5.3cd
9 6.4ab 6.9ab 6.1ab 6.1ab 6.4a 6.1a 6.0bc
10 6.1bc 6.5abc 6.5a 5.4b 5.7ab 5.7a 5.6bcd
11 6.4ab 5.9c 5.5b 6.1ab 6.38ab 6.2a 7.4a
12 6.4abc 6.4bc 6.4ab 6.4a 6.34ab 6.1a 5.7bcd
13 6.6ab 7.2a 6.9a 6.1ab 6.38ab 6.1a 6.7ab
14 5.5bc 6.6abc 6.3ab 5.7ab 5.54b 5.5a 4.8d
SPISE2009 93

Table 3 Linear model describing relationship between overall liking score and product attribute liking
scores (Overall liking score = Y = 2.18247901352456-9.22815493277694E-02*X1+.228769258786121*
X2+6.34402086385718E-02* X3+.994663018412948* X4-.769223467650164* X5+.234665755839502* X6)

Source Value Standard error t Pr > |t|

Intercept 2.182 1.623 1.345 0.221
Appearance (X1) -0.092 0.106 -0.867 0.415
Colour (X2) 0.229 0.121 1.893 0.100
Flavour (X3) 0.063 0.208 0.306 0.769
Sweetness (X4) 0.995 0.513 1.941 0.093
Saltiness (X5) -0.769 0.482 -1.595 0.155
Crispness (X6) 0.235 0.079 2.966 0.021*

According to the JAR test (Figure 1), over 50% of the panellists were satisfied with the
product attributes of colour, sweetness, saltiness and hardness. However, the product obtained
from flour combination 11 seemed to be too strong in colour and hardness for some tasters
(about 40%). This result suggested that two formulas of the product could be developed as
hard texture and soft texture to satisfy two different consumer groups of preference. However,
when cost is considered, flour combination 11 should be selected.

100 100
90 90
Pan elis t pro po rtio n (%)

Panelis t proportion (%)

80 80
70 70 Too strong
Too strong
60 60 Just-about right
50 Just-about right
50 Too weak
40 Too weak 40
30 30
20 20
10 10
0 0
Hardness
Saltiness
Sweetness

Hardness
Sweetness

Saltiness
Color

Color

Figure 1 Product attribute intensity justified by panellists using JAR scale a) Flour combination 11
(left panel) and b) Flour combination 13

4. CONCLUSIONS
According to this study, the most suitable flour combination for flour based peanut snack
production consists of 13% rice flour and 12% tapioca flour because this combination
obtained the highest hedonic score and had minimum cost.

REFERENCES
Ahamed, N.T., Singhal, R.S., Kulkarni, P.R., & Pal, M. (1997). Deep fat-fried snacks from
blends of soya flour and corn, amaranth and chenopodium starches. Food Chemistry,
58(4), 313-317.
Hizukuri, S. (1988) Recent advances in molecular structure of starch. Jpn. Soc. Starch Sci.
31:185.
Matz, S.A. (1991). The chemistry and technology of cereals as food and feed. Springer. 751 p.
SPISE2009 94

Senthil, A., Ravi, R., Bhat, K.K., & Seethalakshmi, M.K. (2002). Studies on the quality of
fried snacks based on blends of wheat flour and soya flour. Food Quality and Preference,
13, 267-273.
Ward, C.D.W., Resurreccion, A.V.A., & McWatters, K.H. (1998). Comparison of acceptance
of snack chips containing cornmeal, wheat flour and cow pea meal by US and West
African consumers. Food Quality and Preference, 9(5), 327-332.
SPISE2009 95

APPLICATION OF GREEN TEA EXTRACT TO BISCUIT CREAM

M.N. Dang*
Danang University of Technology, Vietnam
*Corresponding author: [email protected],

Abstract
Tea, is well-known for its health benefits, and also plays an important role in domestic and
foreign trade of Vietnam. In Vietnam, tea is used mainly as a beverage in various forms,
while its utilisation in other fields is limited. In addition, only young tea leaves and buds are
used for these purposes, leaving large amount of tea leaves unused. The aim of our current
work was to examine the possibility to use freeze dried extract from “unused” tea leaves as
ingredient for biscuit cream to improve oxidation stability of this product.
In the present work, the radical scavenging activity of the extract from older leaves of local
tea was evaluated using DPPH assay. We measured the perception thresholds of some sensory
attributes associated with green tea extract in samples of biscuit cream. We found that
changes in brightness, astringency, sweetness, and bitterness were detected at level of 0.5%,
0.75%, 1% and 1% respectively. Finally, green tea extract was used as an ingredient for
biscuit cream at three levels: 500ppm, 1000ppm and 1%. Samples were stored at 40oC and the
oxidation stability of products with addition of BHT, green tea extracts, and without any
antioxidant were evaluated by measuring the peroxide and TBA values. We found that green
tea extracts showed good inhibition effect on lipid oxidation of the cream. The inhibition
effect of green tea extract at level of 500ppm was equivalent to the effect of BHT at 200ppm.
The higher the concentration of tea extract applied, the greater the effect.

Keywords: green tea, biscuit, antioxidant, sensory analysis, lipid oxidation.

1. INTRODUCTION
Green tea is a popular drink in Vietnam as well as in many parts of the world. Its consumption
has increased in many countries since the discovery of its health benefits. It has been well
documented that green tea possesses antioxidant, antimicrobial, anticarcinogen effect and
others. These effects are supposed to be linked to bioactive substances present in tea such as
catechins, methylxanthine. Among these substances, epigalocatechin galate (EGCG) is very
important because of its high concentration and high activity (Gramza & Korczak, 2005;
Shahidi & Nuczk, 2003; Wang et al., 2000; Hara, 2001). The chemical characterisation and
bioactivity of green tea depends on many factors, such as geographical location, cultivar
species, season, age of the leaves, climate and horticultural practices (Fernandez, Pablos,
Martín, & González, 2002; Lin, Tsai, Tsai, & Lin, 2003). In most studies, green tea was from
Chinese, Japanese or Indian origin, and tea of Vietnamese origin was rarely studied, this is
particularly the case for tea from central Vietnam.
SPISE2009 96

Vietnam is one of the biggest tea producers of the world. Tea here is made mainly from the
buds and young leaves. Local people use older leaves to make drink as well, but using older
leaves for other purpose is unpopular in the country, including for extending food shelf-life.
Biscuit is a favourite food which is usually stored for extended time before consumption.
During the storage, we observe oxidation of the lipid component of the biscuit, particularly
from the cream which has a large proportion of shortening. This slow oxidation process
deteriorates the nutritional and sensory value of the product, making it unacceptable for
consumers. In order to delay this process, synthetic antioxidants such as BHA, BHT have been
used, but the safety of these compounds is questionable and there is a trend to restrict their
use.

This paper presents a study on the possibility to use the extract from older tea leaves for
preventing oxidation of biscuit cream.

2. MATERIALS AND METHODS

2.1 Chemicals:
DPPH (2.2-diphenyl 1-picryl hydrazyl) from Sigma, Singapore; TBA (Thiobarbituric acid) from
Merck, Germany; Methanol in DPPH assay from Merck, Germany; other chemicals of Chinese
origin.
2.2 Preparation of green tea extract
Fresh green tea was collected from the tea field in the countryside near Danang City,
Vietnam. Young tea leaves and the buds were left, and only the older leaves were used for
experimental purposes. Tea enzymes were inactivated by exposing the tea leaves to steam for
45 seconds. The moisture of the tea measured after steaming was 63.97% (analysed by Ohaus
MB35 Moisture Analyser).
The green tea was extracted with water as follows: steamed tea leaves were cut into to 2-3
mm width bands and soaked in water (20 ml/1 g dry mass). All the materials were contained
in a 2-necked round bottom flask, connected with a condenser and a thermometer. The
extraction took place in water bath at 90 oC for 30 minutes. The extract was then filtered and
concentrated in vacuum oven at 60oC, 200 mbar. The concentrated extract was stored at -20
o
C before freeze-drying for 12 h, yielding brown yellow powder.

2.3 Evaluation of free radical scavenging capacity of tea extract with DPPH assay
The free radical scavenging capacity of the tea extracts was determined and compared with
that of BHT using the DPPH discoloration method (Lu & Chen, 2008). Four concentrations of
the tea extract in methanol 100ppm, 300ppm, 500ppm, 700ppm, were prepared. Dilutions of
the tea extracts or BHT in methanol (1 ml) were added to 1 ml of DPPH (0.1 mM in MeOH) and
mixed thoroughly and allowed to stand for 40 min before absorbance was measured at 517
nm using a Biorad spectrophotometer. The mixture of DPPH solution (1 ml) and methanol (1
SPISE2009 97

ml) were used as a negative control. Results were expressed as Inhibition Capacity (IC) which
is calculated as follows:

where ADPPH represents the absorbance of the negative control sample and Asample the
absorbance of analysed sample.

2.4 Preparation of biscuit cream:

The formula of basic biscuit cream was as follows: shortening, 41.03 g; lecithin, 0.54 g; icing
sugar, 50.74 g; milk powder, 7.13 g (Manley, 2000). These materials were provided by Quang
Ngai Confectionary, Vietnam. The shortening was agitated intensively by hand-mixer, so that
it became soft. Then the melted shortening was added with other ingredients and mixed
thoroughly again. The product was smooth without any grittiness.

2.5 Measurement of peroxide value

According to AOAC 965.33

2.6 Measurement of thiobarbituric value (Alexander, 1996)

Ten gram of cream was added with 2.5 ml HCl 4N and 97.5 ml distilled water in 500 ml
round bottom flask which was connected to a simple distillation unit. The first 50 ml of
distillate were collected, from which 5 ml were transferred to a glass tube. After addition of 5
ml TBA solution (2mg/ml butanol) the content of the tube was boiled in a water bath for 2 h,
quickly cooled in tab water to reach ambient temperature and the absorbance of orange colour
of the solution was measured at 532 nm against the control sample which was prepared in the
same way with exception that 10 ml of distilled water was used instead of the 10 g of cream
for distillation.

2.7 Sensory analysis:

We assessed, using rated difference scales (Lawless & Heymann, 1998), the perception
thresholds of three attributes: bitterness, astringency and sweetness. Six different samples
were prepared; among them there were two control samples without tea addition (one coded
and one used as reference) and four cream samples to which were added tea extract at levels:
0.25%, 0.50%, 0.75% and 1% with regard to cream weight. All samples were coded with
randomised 3-digit numbers except the control used as reference. For each attribute, the
difference between the coded sample and the reference was rated. The rating scale ranged
from 0 (no difference) to 9 (extreme difference). The brightness liking of control and four
samples with tea addition were also determined using a 10-point scale (0: dislike a lot and 9:
like a lot). The attributes were evaluated in the following order: astringency, bitterness,
sweetness and brightness.
SPISE2009 98

The evaluation took place in separate booths equipped with red and white lamps, except for
the analysis of brightness which was carried out under white light. The evaluation of other
attributes took place under red bulbs to mask the brightness difference.

The panel consisted of 30 students from Department of Food Technology, Danang University
of Technology, who knew well sensory evaluation techniques. In order to reach an accurate
result, panellists were provided with bread and distilled water to clean their palates after every
tasting.

2.8 Statistical analysis

Statistical assessment was carried out with the software system of Statgraphics Plus for
Windows 4.0. The rated difference of each attribute at all levels were compared using LSD test
with significant level of  = .05.
3. RESULTS AND DISCUSSION
3.1 Radical scavenging activity of the tea extract
The antioxidant effect of green tea is well known. However, in most studies only commercial
tea products or young tea leaves have been used, and therefore very few studies have focused
on the antioxidant property of older leaves. In order to examine the anti-radical effect of older
leaves of the local tea, we employed the DPPH assay mentioned above and determined the
Inhibition Capacity (IC) of the tea extract at various concentrations and of the BHT at 200ppm.
Results are shown in Table 1.
Table 1: Inhibition Capacity (IC) of the tea extracts and BHT.
Samples BHT(200ppm) Tea extract
700ppm 500ppm 300ppm 100ppm
IC (%) 95.3 96.4 95.4 94.3 93.2

The result showed that the tea powder possessed very good anti-radical activity at all
concentrations applied. Higher concentration scavenged the free radical DPPH better. Tea
powder at concentration 500ppm had nearly the same radical scavenging activity of the
synthetic antioxidant BHT at 200ppm.
Tea catechins are believed to be the main compounds responsible for the antioxidant activity
of tea. It is also known that the content of tea catechins in older leaves is lower than in young
leaves (Hara, 2001). However, the results of our measurements indicate that the extract from
old leaves of the local tea could be a good source of antioxidants and could replace BHT to
inhibit the oxidation of food lipid and because tea extract is natural, it could overcome the
problem rising from concerns about synthetic antioxidants.

3.2 Evaluation of recognition threshold of astringency, bitterness, sweetness and

brightness change in biscuit cream.
The tea extract had brown yellow colour along with a bitter and astringent taste and this could
affect the sensory properties of biscuit cream. Therefore, it is necessary to measure the
SPISE2009 99

recognition threshold of astringency, bitterness, sweetness, and brightness change of biscuit

cream because these properties could guide how much tea extract can be applied without
changing much the sensory properties of the product.

Five samples of cream were prepared and analysed according to sensory analysis method
mentioned earlier. Mean values and standard evaluation were calculated for each attribute.
LSD test at significance level 5% was employed using software Statgraphics Plus 4.0. Results
were shown in table 2 and table 3.
Table 2: Score of rated difference between coded and reference samples.
Attribute Samples
Control 0.5% 0.75% 1.00% 1.25%
Astringency 0,03±0.18a 0,93±1.57ab 1,33±2.15b 1.40±2.11b 2.70±2.71c
Bitterness 0,03±0.18d 0,83±1.58de 0,90±1.99de 1,17±2.09e 2,87±2.89f
Sweetness 0,01±0,31g 0,73±1,05gh 1,17±1,32gh 1,53±1,50h 1,83±1,42i
Table 3: Brightness liking score of cream samples.
Samples
Control 0.5% 0.75% 1.00% 1.25%
8.53±0.63k 6.83±0.83l 6.03±1.25m 5.60±1.50m 4.57±1.38n

Scores were expressed as mean ± standard deviation. Index above the score showed
significant difference at level 5% by LSD test. Scores with the same letter were not
significantly different each other.

The results showed that addition of tea powder affected the sensory properties of the samples
of biscuit cream. The brightness of product was less favoured even at the lowest level of tea
extract application (0.5%) in comparison with the control. The LSD test showed significant
difference in brightness liking at this level. When tea was added at a concentration of 1.25%,
the average hedonic score was less then five, meaning that people do not like this colour of
cream, this suggest that colorant may be added, in this case, in order to improve the colour
appearance of the cream. The brightness change was due to the brown yellow colour of tea
extract, this change was more evident at high concentration of the extract. Because of this
change in appearance, the evaluation of other sensory attributes was carried out under colour
mask (i.e., under red light).
Tea possesses a characteristic astringency and bitterness therefore we can expect that, the
addition of the extract to the cream will create a change in the perception of astringency and
bitterness of the cream. However, when the tea extract was added at a level of 0.5%, only few
panellists could recognise astringency and bitterness, as indicated by the LSD test which
showed no difference between tested and reference samples. At a level of 0.75%, astringent
taste in the cream samples was detected by more panellists but still few panellists detected the
bitterness of the products. The difference in astringency between samples with a 0.75%
extract addition and the reference sample was statistically significant, while the difference in
SPISE2009 100

bitterness was not significant. It could be conjectured that the sweetness of the product
masked the bitterness better than the astringency. The bitterness of the samples was more
evident at level of 1.00% when LSD test showed that the samples with a 1.00% tea addition
were significantly different from the reference samples.
Sweetness is also a characteristic property of the biscuit cream because it contains a large
amount of sugar. Results from Table 2 also showed that the addition of tea powder could
change this characteristic. The LSD test showed that this change was significant only at the tea
concentration above 1.00%.
Therefore, it is possible to conclude that the recognition threshold of astringency, bitterness,
and sweetness of the tea extract in biscuit cream is 0.75%, 1% and 1% respectively, while
brightness liking of the cream decreases at a level of 0.5%.

3.3 Effect of green tea extract on the oxidation stability of biscuit cream
Biscuit cream contains large amount of lipid that could readily oxidize during storage. To
examine if the extract from old leaves of local tea could replace BHT to protect biscuit cream
from oxidation an experiment was carried out as follows.
We prepared six lots of biscuit cream with the formulation mentioned earlier: one was a
control sample without any addition; one was treated with 200ppm BHT; tea extract was added
to the other samples, with tea extracts having the following concentrations: 500ppm,
1000ppm and 1%. Thirty grams of cream from each lot was weighted into 100 ml beaker.
Samples were prepared for six measurements in duplicates (12 beakers). The beakers were
stored in the oven at 40oC during 16 days. Every three days, the samples were taken for
analysis of peroxide value and TBA value; mean values were calculated. Results are shown in
Figures 1 and 2 which display the relationship between measured values and storage time.
Peroxides are primary by-products of oxidation process of lipids and their concentration in
lipids is characterised by their peroxide value. It could be noted from Figure 1 that this
parameter increased continuously during the storage of all samples at elevated temperature
(40oC). Addition of the synthetic antioxidant BHT or tea extract lowered the peroxide value.
The analysis of BHT samples and samples with 500ppm tea extract indicated that the BHT at
200ppm and the tea extract at 500ppm had the same inhibiting effect on the formation of
peroxides. The results also showed that higher concentration of tea extract decreased PV levels
more. We also found that application of the tea extract at a concentration level of 1% was the
most effective. This suggested that tea extracts did not show prooxidant effect at high level, in
contrast with what is observed for vitamin E which has prooxidant effect at high
concentration (Pokorny, 2001).
Similarly, the increase of TBA Value (TBAV) during storage was observed from Figure 2. TBAV
reflected the concentration of malonaldehyde in samples. It is a typical secondary product of
lipid oxidation. The courses of TBAV changes during the storage showed that all extract
SPISE2009 101

treatments could delay the formation of malonaldehyde and that this effect could be rank \-
ordered as: 1% > 1000ppm > 500ppm  BHT.

Figure 1: Changes of peroxide value during storage.

Figure 2: Changes of thiobarbituric acid value during storage.

Results from the analysis of PV and TBAV in cream samples during the storage showed that the
extract from old leaves of local tea could delay the oxidation of biscuit cream. Green tea
extracts at concentration 500ppm were as effective as BHT at 200ppm. This concurs with the
conclusions from the DPPH assay. The antioxidant effect of the tea extract could be explained
by the radical scavenging activity, but, the chelating ability of catechins could also contribute
to this effect (Pokorny, 2001; Hara, 2001).
SPISE2009 102

In a similar study on antioxidant property of tea extract, Wanasundara and Shahidi (1998)
reported that an aqueous ethanol extract of green tea showed prooxidant effect in edible oil
possibly due to catalytic effect of chlorophyll. In our work, we used water as solvent, so the
content of chlorophyll is negligible and the extract exhibits strong antioxidant property.
Our study suggests that application of green tea to biscuit cream does not have a prooxidant
effect and this could be an advantage if the extract is needed to be applied at high
concentration (e.g. in production of functional food). As an aside, some authors also
suggested to use some spices in prevention of lipid oxidation in biscuits (Reddy, Vrooj, &
Kumar, 2005).
In summary, the extract from old leaves of the local tea near Danang, Vietnam has been
proven to be a potential natural antioxidant that could replace BHT in application to biscuit
cream. Its recognition threshold for astringency, bitterness, sweetness change and brightness
change in biscuit cream is 0.75%, 1%, 1% and 0.5% (with regard to cream weight)
respectively.

4. CONCLUSIONS
Using natural antioxidants to replace synthetic ones is a current trend in food industry.
However, only few natural antioxidants and few applications have been successful. This work
evaluated the possibility to use older leaves of local tea in central Vietnam to inhibit oxidation
of lipids in biscuit cream. In order to do so, antiradical activity of the extract was evaluated
and the effect of the tea extract on sensory properties of the cream was also studied.
The results from the sensory and chemical analysis provided useful information on the
threshold value of green tea extract in biscuit cream and on the oxidation stability of the
cream fortified with tea extract. Below the threshold value, the sensory quality of green tea
extract fortified cream is not significantly compromised. Also the cream is proven to be more
stable against oxidation.

So it seems that green tea extract fortified biscuit could be a functional food product with
additional health benefits. This provides a good guide for those biscuit manufacturers who
want to pursue the production of functional biscuits with green tea extract fortification.

REFERENCE
Fernandez, P.L.; Pablos, F.; Martín, M.J.; González, A.G. (2002). Study of catechin and xanthine tea
profile as geographical tracers. J. Agric. Food Chem, 50, 1833–1839.
Gramza, A. & Korczak J. (2005). Tea constituents (Camellia sinensis L.) as antioxidants in lipid
systems. Trends in Food Science & Technology 16, 351-358.
Hara, Y. (2001) Green tea: Health benefits and applications, Marcel Dekker Incorporated.
Lawless, H.T. & Heymann H. (1998), Sensory analysis of Food: Principles and Practices. Chapman
& Hall.
SPISE2009 103

Lin, Y.S., Tsai, Y.J., Tsay, J.S. & Lin, J.K. (2003) Factors affecting the levels of tea polyphenols
and caffeine in tea leaves. J. Agric. Food. Chem, 51, 1864–1873.
Lu, M.J. & Chen C.(2008), Enzymatic modification by tannase increases the antioxidant activity of
green tea. Food Research International, 41, 130–137.
Manley, D. (2000) Technology of biscuits, Crackers, and Cookies.Woodhead Publishing, Cambridge.
Pokorny, J. (2001). Antioxidants in food. Woodhead Publishing Limited, Cambridge.
Príbela, A. (1996), Analyza potravin: Cvicenie [Food analysis: practice], STU, Bratislava.
Reddy, V., Vrooj, A., & Kumar A. (2005) Evaluation of antioxidant activity of some plant extract and
their application in biscuits. Food chemistry, 9, 317-321.
Shahidi, F. & Nuczk, M. (2003). Phenolics in food and nutraceuticals. CRC Press.
Wanasundara, U.N. & Shahidi, F. (1998) Antioxidant and pro-oxidant activity of green tea extracts in
marine oils. Food Chemistry, 63(3), 335-342.
Wang, H. & Provan, G.J, Helliwell, K. (2000). Tea flavonoids: their functions, utilization and analysis,
Trends in Food science & Technology 11, 152-160.
SPISE2009 104

Part 4: Sensory‐instrumental relationship 
SPISE2009 105

EXPLORING THE OPTIMISATION MODEL OF VIETNAMESE CONSUMERS

FOR STERILISED MILKS
T. B. Nguyena,b*, T. M. Lec, and D.H. Nguyen,b
(a) HoChiMinh City University of Technology, Hochiminh-city
(b) HoChiMinh University of Industry, Hochiminh-city
(c) Food Industry College of Ho Chi Minh-city, Vietnam
*Corresponding author: [email protected], Tel: +84.8.8647256 (ext: 5696)

Abstract
The objective of this study was to describe the sensory and physico-chemical properties of
seven commercial strawberry sterilised milk products bought in local supermarkets and to
evaluate consumer preferences for these products. We performed three types of analyses: (1)
descriptive analysis, (2) physico-chemical analysis, and (3) consumer analysis. For the
descriptive analysis, six trained panellists identified and evaluated 13 sensory attributes
(involving taste, aroma, texture) using 7-point structured line scales. In the second analysis,
the gross physico-chemical compositions of the milk products were determined for total
protein, fat content viscosity, and colour using the CIE L*a*b tristimulus method. In the third
analysis, we performed a consumer analysis, in which 140 consumers from a panel were
asked to evaluate the overall-liking of the samples using a 7-point hedonic scale.

The results differentiated the milk samples based on sensory attributes (e.g., pink colour,
greasiness, creaminess) and physico-chemical properties (fat content and viscosity),
respectively. We found that health and sensory appeal were the most important factors when
choosing products and that sweetness and greasiness were the key sensory attributes driving
consumer preference. Multiple factor analysis (MFA) showed that consumer preference was
more correlated to sensory attributes than to physico-chemical properties. Preference mapping
revealed four clusters of consumers.

Key words: dairy products, Vietnamese consumer, preferences, food choice

1. INTRODUCTION
Until recently cow milk was not easily available in the Vietnamese market and so dairy
products were not part of Vietnamese food habits. It is only for the last twenty years that
Vietnamese companies have been producing fresh milk. Today Vietnam has a total of 22 milk
factories. Vinamilk with nine plants has a productivity equal to 1.2 billion litres per year
(Dinh, 2009). Recent surveys indicate that the consumption of fresh milk and dairy products
is increasing (USDA, 2007). Consumers consider dairy products as high value processed food
products which also have nutritional and health benefits. Among the dairy products currently
available in the market, strawberry sterilised milks are the most popular. Yet, consumer
motivations for choosing this type of products are still not well understood. So the purpose of
SPISE2009 106

this research was to analyse (1) the sensory attributes of strawberry sterilised milk, (2) its
physico-chemical parameters, and (3) consumer preference.  

2. MATERIALS AND METHODS

2.1 Stimuli
We used seven strawberry sterilised milks produced by Tetrapak technology: DutchLady,
Vinamilk, Nutifoods, Daisy, Izzi, Milky, and Ancomilk. The strawberry sterilised milk
samples were stored in refrigerated conditions (4 to 5oC); they were taken out of the
refrigerator one hour before the experiment and maintained at room temperature (25oC).
Samples were served in 30 ml white plastic cups coded with three-digit numbers.

2.2 Methods
Descriptive test. Six panelists (4 females and 2 males) were screened and trained under ISO
13300–1:2006 and ISO 5492, 2008 standard. During training, panelists tasted products and
agreed on a list of attributes, definitions, references, and scale measurements. The training
lasted for one week with one session of two hours per day. At the end of training, panellists
evaluated the seven strawberry milk samples on 7-point structured line scales with one
repetition. Thirteen scales were used to perform the evaluation. These scales were 1)
astringent, 2) butter, 3) brown, 4) cream, 5) film, 6) greasiness, 7) oral viscosity, 8) sourness,
9) sour flavour, 10) sweetness, 11) strawberry aroma, 12) strickiness, and 13) pink. Samples
were presented in a randomised balanced block design within each session.
Consumer test. One hundred and forty consumers from Hochiminh City (Vietnam) agreed to
participate in this study. All consumers were adults and most were between the age of 18 and
65 year old and their distribution roughly matched the Vietnamese population age
distribution. The numbers of men and women were roughly equal. Participants had to
evaluate their overall liking of strawberry sterilised milk products on 7-point scales. Samples
were served in random orders. Participants were also asked to fill in a small questionnaire on
their food habits and preferences for dairy products.

Gross composition parameters. The Color of products was determined with a Konica Minolta
CR-410 colormeter using color scale CIE L*a*b (Gámbaro, Ares, Giménez, & Pahor, 2007).
The apparent viscosity of products was measured with a Brookfield VI rotary viscometer
(Brookfield Engineering Laboratories Inc., Stoughton, MA); using spindle no. 1 at 60 r.p.m.
Measurements were made at ambient temperature. To avoid thixotropic effects, the samples
were thoroughly stirred just before measurements. The Kiendahj method was used to measure
total protein content of strawberries sterilised milks (TCVN 5537:1991) and the gravimetric
method to was used to determine fat content in milk products. (TCVN 6508:1999)
Data analysis: The packages SensomineR and FactomineR (with R version 2.8.0, Lê &
Husson, 2006) were used to analyse sensory and consumer preferences data. Descriptive test
data were analysed with analysis of variance (ANOVA). Preference mapping (PREFMAP) was
SPISE2009 107

used to combine sensory properties and consumer data to explore consumer-preferences.

ANOVA was also used to analyse the difference between the physico-chemical properties of
the milk products. Consumer data, physio-chemical properties, and sensory properties were
combined using multiple factor analysis (MFA, see Morand & Pagès, 2006) to describe the
relationship between sensory, physio-chemical properties, and consumer preferences.

3. RESULTS AND DISCUSSION

3.1. Physico-chemical properties
Fat content, total protein, viscosity, and color of milk products are shown in Table 1. The
results indicate that Ancomilk and Izzi have high fat content, as opposed to Milky, Nutifood,
Daisy, and Dutchlady which all have a low fat content. We found no significant differences
about total protein amount (p >.05). Nutifood has the highest viscosity.

3.2. Sensory properties

The thirteen scales rated by the six trained panellists were analysed using principal
component analysis (PCA). The first two principal components (PCs) of the analysis explained
62.4% of the variance.
Table 1. Physio-chemical analysis and colorimetry of products:L*. a*. b*: colour parameters
(Gámbaro et al., 2007)

Fat Total Protein L*a*b colormetric Viscosity

Products
[g/100ml] [g/100ml] L* a* b* [Cp]
Milky US 0.804±0.03 1.364±0.03 107.300±2.82 37.97±2.50 06.54±0.78 11.30±2.50
Nutifood 1.110±.003 1.348±0.04 104.157±2.63 05.54±2.90 23.48±1.50 25.10±1.87
Daisy 0.911±0.03 1.218±0.03 108.963±2.54 37.43±2.80 05.80±0.95 08.69±1.10
Vinamilk 1.726±0.05 1.422±0.05 109.937±2.80 34.62±2.75 04.49±0.75 10.16±1.55
DutchLady 1.084±0.06 1.466±0.04 110.943±2.74 24.59±2.76 12.52±1.09 10.87±1.78
Ancomilk 1.726±0.04 1.494±0.06 101.450±2.53 35.39±2.54 09.79±0.75 14.13±1.96
Izzi 2.004±0.04 1.417±0.04 110.455±2.64 41.06±2.55 -1.25±0.05 09.12±1.09

Figure 1b shows the plot of the correlations between the sensory scales and the first two PCs
along with the “correlation circle.” As can be seen from this graph, the sensory scales form
two distinct clusters. The first cluster includes texture properties such as: oral viscosty,
greasiness, stickiness; and the second cluster includes aroma properties such as: sweetness,
strawberry aroma, sour flavour. The brown color was negatively correlated with pink color (r
= –.85; p < .05) and strawberry aroma (r = –.80, p < .05).

Figure 1a displays the projections of the products on the first two PCs. Ancomilk and Nutifood
are clearly differenciated from the other products. Ancomilk was associated to stickiness,
greasiness, stickiness, and strawberry; Nutifood was associated to the brown color and at the
opposite side of PC-1, Izzi was characterised by sweetness, pink color, and strawberry aroma.
SPISE2009 108

Figure 1. Correlation cirle (a) and product projection (b) of the seven sterilised milks studied.

3.3. Consumer preferences

For all products studied, the average of overall liking score was around 4 exept for Daisy. We
analysed the data using a repeated measurement design ANOVA with product being the within
subject factor. We observed a significant difference between product preferences (p < .05).
The most liked product was Vinamilk and the least liked product was Daisy (Figure 2).

Figure 2. The overall liking score of 140 consumers

The results of survey on milk consumer habit showed that more than 87% of the consumers
reported that they often drink milk, at least “1 to 3 times per week.” Among the products
studied, Vinamilk and Dutch Lady are the most familiar products as more than 90%
participants reported usually drinking them. When chosing milk, consumers were interested in
health (71%), sensory appeal (64%), origin (41%), trademark (40%), and price (26%). Only
5% of consumers reported being concerned by package (see Figure 3).
SPISE2009 109

Figure 3. The principal factors influencing milk choice of consumer

A preference mapping analysis confirmed the results of the hedonic test. About 70% of
consumers liked Vinamilk and Izzi products and about 60% liked Dutch Lady, Daisy,
Nutifood, and Ancomilk. From the preference mapping analysis, we can derive a description
for an optimum product which would be close to Vinamilk and Izzi and which would be
characterised by pink color, strawberry aroma, and sweetness (Figure 4).

Figure 4. Preference mapping of strawberry sterilised milks

The analysis also indicate that there were four clusters of consumers which were cluster 1 (46
consumers), cluster 2 (37 consumers), cluster 3 (28 consumers), and cluster 4 (29 consumers)
as shown Figure 5.
SPISE2009 110

25
20
15 Cluster 2 Cluster 1 Cluster 4 Cluster 3
Height

10
5

01
10

6
Figure 5. Cluster consumers from preference mapping

3.4. Relationship between sensory, physico-chemical, and consumer data

Multiple factor analysis (MFA) was used to analyse a data matrix obtained by the
concatenation of three data sets: sensory profile, physico-chemical parameters, and average
score of each of the four clusters of consumers obtained from the preference mapping analysis.
The first three PCs of the MFA explained 75.59% of the variance. Figure 6 shows the plot of the
correlation of these PCs with variables used in the analysis.” Consumer clusters 1 and 2 are
associated with sweetness, protein, and greasiness; consumers cluster 3 is associated with
strawberry aroma, sour flavor and pink colour, and consumer cluster 4 was associated with
film, strawberry sourness and oral viscosity.

Figure 6. MFA: Correlation of chemical, sensory, and consumer variables with PCs1 and 2 (left) and
PCs 1 and 3 (right).
SPISE2009 111

4. CONCLUSION
This study examined thirteen sensory properties of strawberry sterilised milk products and
their relationships with the physico-chemical properties of these milks. Among the products
studied, the preferred product was Vinamilk, which was characterised by strawberry aroma,
sweetness, and pink color. Finally, sensory appeal and health are very important when
choosing milk products.

Acknowledgement. This study was funded by Hochiminh City University of Technology. The
authors thank volunteers from Food Industry College of Hochiminh city, Hochiminh
University of Industry participanted in this research.

REFERENCES
Dinh., V.C. (2009). Cow milks: research and development in Vietnam. Retrived from
www.dairyvietnam.org.vn/vi/publications.php?mnu=5&nid=124
Gámbaro, A., Ares, G., Giménez, A., & Pahor, S. (2007). Preference mapping of color of uruguayan
honeys. Journal of Sensory Studies, 22, 597–519.
ISO 5492-2008, Sensory analysis- Vocabulary
ISO 13300-1:2006, Sensory analysis- General guidance for the staff of a sensory laboratory- Part 2:
Recruitment and training of panel leaders
Lê, S. & Husson, F., (2006) SensomineR: a package for sensory data analysis, Journal of Sensory
Studies, 23, 14–25.
Morand, E., & J. Pagès, (2006). Procrustes multiple factor analysis to analyse the overall perception of
food products. Food Quality & Preference, 17, 36–42
Statistical office of HoChiMinh City (2005), Average population by sex and by urban, rural. Retrived
from http://www.pso.hochiminhcity.gov.vn/so_lieu_ktxh/2005/Dan_so_ va_ ao_dong/0202.htm
USDA, Vietnam Food Detail report (2007) Gain Report Number: VM7086, 2007. Retrived from
http://www.stat-usa.gov/agworld.nsf/505c55d16b88351a852567010058449b/
90801935d5515d56852573a70057f725/$FILE/VM7086.PDF
Vietnamese standard, TCVN 5537:1991. Total protein content in milk products.
Vietnamese standard, TCVN 6508:1999. Gravimetric method determined fat content in milk products.
SPISE2009 112

THREE COMPOUNDS WITH POTENT -GLUCOSIDASE INHIBITORY

ACTIVITY PURIFIED FROM SEA CUCUMBER STICHOPUS JAPONICUS
H.T. Nguyenab & S.M. Kima*
(a) Faculty of Marine Bioscience and Technology, Kangnung-Wonju National University
Gangwondo, Republic of Korea
(b) Faculty of Aquatic Product Processing Technology, Nha Trang University, Vietnam
*Corresponding author: [email protected], Tel: +82-33-640-2343

Abstract
Diabetes mellitus is a worldwide health problem which is increasing every year. One
therapeutic approach to decrease postprandial hyperglycemia is to slow down the absorption
of glucose through inhibition of -glucosidase. Here we explore the potential α-glucosidase
inhibitory activity of compounds extracted from sea cucumber, Stichopus japonicus. We
found that the aqueous methanol extract of sea cucumber, Stichopus japonicus inhibited yeast
-glucosidase activity by 68% at 0.5 mg/ml. The following hexane fraction was the most
potent (98% inhibition at 10 g/ml). Three compounds with potent -glucosidase inhibitory
activities were purified from S. japonicus. IC50 values of compound 1, 2 and 3 were 1.22, 0.17
and 0.36 g/ml against Saccharomyces cerevisiae -glucosidase, and 2.49, 0.24 and 0.21
g/ml against Bacillus stearothermophilus -glucosidase, respectively. Both compound 1 and
2 inhibited yeast -glucosidase activity non-competitively (Ki value of 0.98 and 0.06 g/ml),
while compound 3 showed a mixed type inhibition (Ki value of 0.61 g/ml). In addition,
compound 1, 2 and 3 were very stable under thermal and acidic conditions up to 30 and 60
min. We conclude that compounds of S. japonicus have potential as natural nutraceuticals and
that these compounds could be use to prevent diabetes mellitus because of their high α-
glucosidase inhibitory activity.

Keywords: -Glucosidase, -Glucosidase inhibitors, Stichopus japonicus, Diabetes mellitus.

1. INTRODUCTION
Diabetes has become an alarming global problem in recent years. According to the
International Diabetes Foundation (IDF), the number of people diagnosed with diabetes in the
last twenty years has risen from 30 million to over 246 million. It is also a lifestyle-related
disease known to trigger many complications, nephropathy, retinopathy, neuropathy,
cardiovascular diseases, and so on (Saijyo et al., 2008). Type-2 diabetes is the most common
form of diabetes, accounting for 90% of cases, and it is usually characterised by an abnormal
rise in blood sugar right after a meal, called postprandial hyperglycemia (Apostolidis, Kwon,
& Shetty, 2006). Mammalian -glucosidase (-D-glucoside glucohydrolase, EC 3.2.1.20)
located in the brush-border surface membrane of intestinal cells, is the key of enzyme which
catalyses the final step in the digestive process of carbohydrates. Hence, -glucosidase
inhibitors can retard the liberation of D-glucose of oligosaccharides and disaccharides from
dietary complex carbohydrates, delay glucose absorption, and, therefore suppress postprandial
SPISE2009 113

hyperglycaemia (Lebovitz, 1997). Such inhibitors, including acarbose and voglibose, are
currently used clinically in combination with either diet or other anti-diabetic agents to
control blood glucose levels of patients (Van de Laar et al., 2005). A main drawback of the
current -glucosidase inhibitors (such as acarbose) is the presence of side effects such as
abdominal bacterial fermentation of undigested carbohydrates in the colon (Bischoff et al.,
1985). To either avoid or decrease the adverse effects of current agents and also to provide
more candidates of drug choices, it is still necessary to search for new α-glucosidase
inhibitors for further drug development (Lam et al., 2008).
The sea cucumber (a cylindrical marine invertebrate) has long been used as a tonic food in the
Ido-pacific region as well as a traditional medicine in East Asia due to its high nutraceutical
value (Zhong, Khan, & Shahidi, 2007). Several papers published in the last two decades
support these medical claims and document multiple biological activities of sea cucumber
extracts such as wound healing promoter and exhibiting antimicrobial, anticancer and
immunomodulatory properties (Mamelona et al., 2007).

The sea cucumber, Stichopus japonicusa widespread species in East Asia, including Korea,
China, Japan and Far Eastern Russia (Kanno, Li, & Kijima, 2005)contains antifungal
triterpene glycosides, holotoxins A, B, and C (Kitagawa, Sugawara, & Yosioka, 1976).
However, its -glucosidase inhibitory activities have not been reported. The objective of this
study was to investigate the inhibitory effects of sea cucumber on -glucosidase activity.
2.. MATERIALS AND METHODS
2.1 Materials
Live specimens of sea cucumbers Stichopus japonicus were purchased from a fishery market
in Kangnung, South Korea. B. stearothermophilus and S. cerevisiae α-glucosidase, rat
intestinal acetone powder, p-Nitrophenyl-α-D-glucopyranodase (pNPG), and glucose assay kit
(GAGO-20) were purchased from Sigma Chemical Co. (St. Louis, MO, USA). Sephadex LH-20
and Sephacryl HR-100 columns were purchased from Pharmacia Biotech Ltd. (Uppsala,
Sweden). The other chemicals used in this study were of analytical grade.

2.2 Sample preparation

Fresh sea cucumber specimens were washed with tap water. The body wall was cut into small
species. All samples were frozen at -80 0C for 24 h, and vacuum dried at 30 0C for 72 h. The
dried samples were ground to fine powder and sieved with a  600 μm sieve and kept frozen
at -20 0C until used.

2.3 Extraction and purification of -glucosidase inhibitors

The powder of sea cucumber (200 g) was extracted with refluxing 80% methanol (2  5 l,
each for 2 h). After the solvent was removed under reduced pressure at <400C, the residue
(70.5 g) was suspended in water and then successively partitioned with n-hexane, CHCl3,
EtOAc, and n-BuOH. The n-hexane fraction (17.6 g) was chromatographed over silica gel (2.0
SPISE2009 114

 15.0 cm) eluting with n-hexane, hexaneCHCl3, and CHCl3MeOH (in order of increasing
polarity) to give nine main fractions (F1-F9). Fraction F1 and F4 were further purified by
Sephadex LH-20 (3.0  30.0 cm) to obtain three fractions (compound 1, 102 mg; compound 2,
67 mg; compound 3, 82 mg).

2.4 Assay for -glucosidase inhibitory activity

-Glucosidase inhibitory activity was performed following the modified method of Kim,
Nam, Kurihara, and Kim (2008). A reaction mixture containing 2.2 ml of 0.01 M phosphate
buffer (pH 7), 0.1 ml of 0.25 U/ml -glucosidase in 0.01 M phosphate buffer (pH 7), and 0.1
ml of sample was pre-incubated for 5 min at 37 0C, and then 0.1 ml of 3 mM pNPG as a
substrate in the same buffer was added to the mixture. After further incubation at 37 0C for 30
min, the reaction was stopped by adding 1.5 ml of 0.1 M Na2CO3. Enzymatic activity was
quantified by measuring the absorbance at 405 nm. The percentage of -glucosidase enzyme
inhibition by the sample was calculated by the following formula: % inhibition = [AC –
AS]/AC×100, where AC is the absorbance of the control and AS is the absorbance of the tested
sample. The concentration of an inhibitor in the reaction mixture required to inhibit 50% of
enzyme activity under the foregoing assay conditions is defined as the IC50 value.

2.5 Purification of rat intestinal α-glucosidase

Rat intestinal α-glucosidase was purified according to the method of Kim, Nam, Kurihara, and
Kim (2008). Commercial rat intestine acetone powder (10 g) was dissolved in buffer A
(100 ml) (0.1 M potassium phosphate buffer containing 5 mM EDTA, pH 7.0), sonicated at
4 °C for 15 s, and then centrifuged at 27,000g at 4 °C for 60 min to obtain supernatant A. The
precipitate was dissolved in buffer A (100 ml), sonicated, and then centrifuged (32,000g,
60 min, 4 °C) to obtain supernatant B. The combined supernatant was dialysed against buffer
B (0.05 M potassium phosphate buffer containing 0.4 mM EDTA, pH 7.0) for 48 h. This was
concentrated to 20 ml by ultrafiltration (cut off membrane 10 kDa) and then loaded onto a
Sephacryl HR-100 column (2.6 × 60.0 cm) equilibrated with buffer B in advance. Rat
intestinal α-glucosidase was eluted with buffer B at a flow rate of 0.2 ml/min.

2.6 Inhibitory assay for rat intestinal α-glucosidase (sucrase and maltase) activity
Rat intestinal α-glucosidase inhibitory activity was determined according to the modified
method of Kurihara, Mitani, Tawabata, and Takahashi (1999). Sucrase activity was
determined in a mixture of 500 mM sucrose (0.1 ml), the isolated compound in MeOH
(0.05 ml), and 0.1 M maleate buffer (pH 6.0, 0.75 ml). The mixture was preincubated at 37 C
for 5 min, and reaction was initiated by adding rat intestinal α-glucosidase (0.1 ml) to the
reaction mixture. The mixture was incubated at 37 C for 60 min. The reaction was terminated
by adding 2.0 M maleate–Tris–NaOH buffer (pH 7.4, 1.0 ml). To measure maltase activity,
maltose (500 mM) was used instead of sucrose. The glucose release in the solution was
determined using a glucose assay kit based on the glucose oxidase/peroxidase method. One
SPISE2009 115

unit of α-glucosidase activity was defined as the amount of enzyme that liberated 1.0 M of
substrate per min. One unit of α-glucosidase inhibitory activity was defined as a 1 unit
decrease in α-glucosidase activity.

2.7 Stability under thermal and acidic conditions

The stability of purified compounds under thermal and acidic conditions assays were
performed following the method of Kim et al. (2005). Compounds 1, 2 and 3 were treated at
100 0C for 30 min or 37 0C for 40 min and 60 min at pH 2. Each compound was then used for
inhibition assay against B. stearothermophilus and S. cerevisiae -glucosidases. The relative
inhibitory activity of bromophenol at pH 2 was calculated based on the inhibitory activity of
the control.

2.8. Kinetics of -glucosidase inhibitor

For kinetic analyses of S.cerevisiae -glucosidase by compounds 1, 2 and 3, the enzyme and
test compounds were incubated with increasing concentration of pNPG. Inhibitory kinetics of
compounds 1, 2 and 3 for yeast -glucosidase was determined by Lineweaaver-Burk plot
analysis of the data (Lineweaver & Burk, 1934), which were calculated from the result
according to Michaelis-Menten kinetics.

2.9 Statistical analysis

SPSS for windows (version 10.0, SPSS Inc., Chicago, IL) was used for statistical analysis. Each
value is expressed as the mean  standard deviation (SD). Differences among groups at
various times of the experiment were subjected to a one-way analysis of variance (ANOVA)
followed by Duncan’s multiple-range t-test. A value of p < .05 is considered to indicate a
statistically significant effect.

3. RESULTS AND DISCUSSION

3.1 Extraction and isolation of -glucosidase inhibitors
The powder of S. japonicus (200 g) was reflux extracted using 80% MeOH. The aqueous MeOH
extract inhibited yeast -glucosidase by 68% at 0.5 mg/ml. In order to isolate -glucosidase
inhibitors from sea cucumber, the 80% MeOH extract was fractionated by monitoring
S.cerevisiae -glucosidase inhibitory activity. Firstly, the concentrated extract was suspended
in water and successively partitioned with n-hexane, CHCl3, EtOAc, and n-BuOH. The n-hexane
fraction showed the most potent inhibitory activity effects on yeast -glucosidase
(approximately inhibition 98% at 10 g/ml), followed by EtOAc (20.7%), even though the H2O
fraction resulted in the highest solute yield (68.1%), followed by the n-hexane (23.2%), BuOH
(4.0%), , CHCl3 (3.3%), and EtOAc (1.4%) fractions (Table 1). A large amount of the solute in
the aqueous MeOH (80%) extract was shifted over to H2O fraction, but this fraction showed no
α-glucosidase inhibitory activity at 0.5 mg/ml. Lipids, chlorophyll, refined oil, and sterols
were dissolved and fractioned in n-hexane; , CHCl3 is a excellent solvent for extracting resin.
In addition, polyphenols such as flavonoids and tannins are typically extracted in EtOAc and
SPISE2009 116

water-soluble components are shifted over to BuOH fraction (Suffness, Newman, & Snader,
1989). In a previous study, we found that the highest α-glucosidase inhibitory activity was
observed in the EtOAc fraction of the red alga, Grateloupia elliptica, followed by BuOH, and
H2O fractions, while n-hexane and , CHCl3 fractions did not show any α-glucosidase inhibitory
activity at 0.1 mg/ml (Kim, Nam, Kurihara, & Kim, 2008). Because the n-hexane fraction
appeared to be the most potent α-glucosidase inhibitory activity, the main α-glucosidase
inhibitor of S. japonicus was provisionally considered to be a fatty acid compound.

The hexane fraction in S. japonicus was further purified to isolate the compounds with α-
glucosidase inhibitory activity using silica gel, Sephadex LH-20 chromatographic separations,
respectively. Two fractions among nine fractions eluted from silica gel column showed α-
glucosidase inhibitory activity. Finally, three fractions from Sephadex LH-20 chromatographic
separation were pooled, and then confirmed and classified as compound 1 (102 mg), 2 (67
mg) and 3 (82 mg). We are currently analysing the structure of these compounds.
Table 1. -Glucosidase inhibitory activities of the solvent-partitioned fraction of sea cucumber, S.
japonicus at different concentrations.

-Glucosidase inhibitory activity (%)

Fractions Yield (%)
0.01 mg/ml 0.1 mg/ml 0.5 mg/ml
n-Hexane 23.2 98.2  0.2a 98.3  0.1a 98.8  0.8a
CHCl3 3.3 NI3) NI 8.0  2.3c
EtOAc 1.4 20.7 ± 4.5b 92.5 ± 4.7b 98.5 ± 0.1a

n-BuOH 4.0 17.9 ± 5.3b 18.5 ± 3.6c 42.9 ± 1.4b

Water 68.1 NI NI NI
1
Values are expressed as mean  SD. Mean in the same column with different superscripts are significantly
different (p  0.05). 2 The final concentration in the reaction mixture. 3 No inhibition.

3.2 Assay for -glucosidase inhibitory activity

Isolated compounds inhibited Saccharomyces cerevisiae and Bacillus stearothermophilus -
glucosidase activity in a dose-dependent manner (Figure 1). The IC50 values for compounds 1,
2 and 3 were determined to be 1.22, 0.17 and 0.36 g/ml against S. cerevisiae -glucosidase,
and 2.49, 0.24 and 0.21 g/ml against B. stearothermophilus -glucosidase, respectively
(Table 2). Compound 2 had strongest yeast α-glucosidase inhibitory activity among these
compounds, and IC50 value (0.17 g/ml) of 2 was about seven and two-fold lower than those
of compounds 1 and 3, respectively. Compounds 2 and 3 showed more potent inhibitory
activity against both yeast and bacterial α-glucosidases than compound 1. Nacao et al., (2004)
reported that two sulphated fatty acids from a marine sponge Penares sp., sulphates A1 and
B1, inhibited yeast α-glucosidase inhibitory activity (IC50 values of 1.2 and 1.5 g/ml,
respectively). It can be seen that the IC 50 of compounds 2 and 3 were lower than those of
SPISE2009 117

sulphates fatty acids, A1 and B1. In contrast, the commercial inhibitor, acarbose exhibited no
α-glucosidase inhibitory activities against S. cerevisiae and B. stearothermophilus -
glucosidases (Table 2). This is in agreement with other reports that either described a very
weak inhibitory activity of acarbose (Oki, Matsui, & Osajima, 1999) or no inhibition against
S. cerevisiae α-glucosidase (Kim et al., 2004; Kim, Nam, Kurihara, & Kim, 2008). A similar
result was observed for voglibose and glucono-1,5-lactone, which strongly inhibited
mammalian α-glucosidases, whereas no inhibition was observed in yeast and bacterial α-
glucosidases (Oki, Matsui, & Osajima, 1999; Kim, Nam, Kurihara, & Kim, 2008). In contrast,
Oki, Matsui, & Osajima (1999) showed that (+)-catechin, an inhibitor of S. cerevisiae α-
glucosidase, had no inhibitory effect on enzymes from mammalian species.

100 100 100

80 80 80
Inhibitory activity (%)

Inhibitory activity (%)

Inhibitory activity (%)

60 60 60

40 40 40

Compound 3 Compound 2 Compound 1

20 20 20

0 0 0
0 1 2 3 4 0 0.5 1 1.5 2 2.5 0 1 2 3 4 5
Concentration (g/ml) Concentration (g/ml)
Concentration g/ml)

Figure 1. Dose-dependent inhibition of S.cerevisiae () and B. Stearothermophilus () -glucosidases

by sea cucumber compounds.
Table 2. IC50 values of isolated compounds from S. japonicus against -glucosidases

Inhibitors IC501,2
S. cerevisiae B. Stearothermophilus Sucrase Maltase
-glucosidase -glucosidase (mg/ml) (mg/ml)
a a
Compound 1 1.22  0.02 2.49  0.07 3.52  0.15a 4.21  0.31a
Compound 2 0.17  0.01c 0.24  0.02b 2.30  0.11b 3.47  0.07b
Compound 3 0.36  0.01b 0.21  0.01c 0.08  0.02c 0.10  0.03c
Acarbose NI3 NI 0.02  0.01d 0.05  0.01d
1
Values are expressed as mean ± SD. Means in the same column with different superscripts are significantly
different (p < .05). 2 The IC50 value is defined as the inhibitor concentration to inhibit 50% of its activity under
assayed conditions. 3 NI: no inhibition.

3.3 Inhibitory activity against rat-intestinal sucrase and maltase

The inhibitory activities of S. japonicus compounds against rat-intestinal sucrase and maltase
were also compared with those of acarbose (Table 2). The IC50 values of compounds 1, 2 and
3 were 3.52, 2.30 and 0.08 mg/ml against rat-intestinal sucrase, and 4.21, 3.47 and 0.10
SPISE2009 118

against maltase, respectively. The IC50 value of acarbose was 0.02 mg/ml against sucrase and
0.05 mg/ml against maltase, respectively. Hence, acarbose was a better inhibitor against
mammalian α-glucosidase than the S. japonicus compounds. In previous studies, most
mammalian α-glucosidase inhibitors did not effectively inhibit microbial α-glucosidases,
whereas catechin, an inhibitor of S. cerevisiae α-glucosidase, did not inhibit mammalian α-
glucosidases (Oki, Matsui, & Osajima, 1999). Therefore, the higher mammalian α-
glucosidase inhibitory activity of acarbose compared to sea cucumber compounds is likely
due to the substrate specificities that depend on the source of α-glucosidases. This suggests
that the binding of compounds is less specific to the enzyme because these compounds can
bind to various proteins included in the crude enzyme solution. Bacterial, yeast, and insect
enzymes, called α-glucosidase I, show higher activity toward heterogeneous substrates such
as sucrose and pNPG, and either less or no activity toward homogeneous substrates such as
maltooligosaccharides; this implies that α-glucosidase I recognises the “glucosyl structure” in
the substrate (Kimura, 2000). The mould, plant, and mammalian enzymes, called α-
glucosidase II, hydrolyze homogeneous substrates more rapidly than heterogeneous
substrates, indicating that this class of α-glucosidases recognises the “maltostructure”
(Kimura et al., 2004). The hydrolysis of p-nitrophenol 2-deoxy-α-D-arabino-hexopyranoside
was catalysed by α-glucosidase II (Nishio et al., 2002), but no such reaction was observed
with α-glucosidase I, suggesting that the 2-OH groups in the glucose moiety are essential for α-
glucosidase I (Kimura et al., 2004). The α-glucosidase II catalysed the hydration of D-glucal
to produce 2-deoxy-α-D-arabino-hexose, but α-glucosidase I yielded no detectable hydration
product (Chiba et al., 1988). The strong enzymatic inhibitory activity against microbial α-
glucosidases shown by sea cucumber compounds is clearly better than the activity of
commercial inhibitors such as acarbose and voglibose at low concentration. It is likely that
sea cucumber compounds can decrease blood glucose level but with fewer, if any, adverse
gastrointestinal effects, and abdominal discomfort than acarbose and voglibose (Tewari et al.,
2003; Iwai, 2008).

3.4 Kinetics of enzyme inhibition

The inhibition mode of isolated compounds against S. cerevisiae α-glucosidase was analysed
from the data derived from enzyme assays containing difference concentrations of pNPG,
ranging from 2 to 10 mM at each different of the compounds. The data indicate that both
compounds 1 and 2 showed non-competitive inhibition against S. cerevisiae α-glucosidase,
while compound 3 displayed a mixed type inhibition mode (Figure 2). The different inhibition
kinetics of these compounds seemed to be due to structural differences formed by the origins
of the enzymes (Kim et al., 2005). The Ki (inhibition constant) values of compounds 1, 2, and
3 were 0.98, 0.06, and 0.61 g/ml, respectively (Table 3), which shows that compound 2 was
a more effective inhibitor than compounds 1 and 3.
SPISE2009 119

600 500 200

3 ㎍/ml

1 ㎍/ml 0.5 ㎍/ml 0.5 ㎍/ml

400
160

1/v (mM /min) -1

400 0 ㎍/ml 0.1 ㎍ mg/ml 0.25 ㎍/ml

1/[v] (mM /min) -1

0 ㎍/ml 0 ㎍/ml
300
120

1/v (mM /min) -1

200 200
80

100
0 40
-0.4 -0.2 0 0.2 0.4 0.6 0
0
1/[S] (mM )-1 -0.2 0 0.2 0.4 0.6
-200 -0.4 -0.2 0 0.2 0.4 0.6
-100 1/[S] (mM)-1
-40 1/[S] (mM )-1

Figure 2. Lineweaver-Burk plot of S. cerevisiae -glucosidase inhibition of isolated

compounds from S. japonicus at different concentrations of pNPG.

Table 3. Ki values and inhibition mode of isolated compounds against yeast -glucosidase.( 1 Ki value
was determined by Dixon plot analysis.2 Inhibition mode was determined by Lineweaver–Burk plot).

Compounds Ki (g/ml)1 Inhibition mode2

1 0.98 Non-competitive
2 0.10 Non-competitive
3 0.61 Mixed

3.5 Stability of purified compounds under thermal and acidic conditions

To determine the potential for industrial usage and stability in digestive organs of the sea
cucumber compounds, we investigated the stability of extracts at high temperature and low pH
by measuring inhibitory activity on bacteria B. stearothermophilus and yeast S. cerevisiae α-
glucosidases of sea cucumber compounds. Table 4 shows that the inhibition activity of all
compounds retain more than 99% of its value against both bacteria and yeast α-glucosidases
after standing for up to 30 min at 37 0C. The acidic condition (pH 2) was used to mimic the
pH encountered by the compound in the stomach. The inhibitory activity of sea cucumber
compounds against S. cerevisiae -glucosidase was very stable at pH 2 for 40 min and
remained from 80% (compound 1) to 88% (compound 3) for 60 min, respectively, while these
compounds remained moderately against B. stearothermophilus -glucosidase, the inhibitory
activities were remained from 75 to 83% and 67 to 69% for 40 and 60 min, respectively. This
finding was similar to the result of pine bark extract, where its inhibitory activity against yeast
S.cerevisiae -glucosidase was very stable under thermal condition and mildly stable under
acidic condition (Kim et al., 2005). This result suggests that the purified compounds from S.
japonicus are stable under thermal and acidic conditions and can be a candidate for
development of a -glucosidase inhibitor. Moreover, S. japonicus may be useful for potential
usage of industrial use or as food additive.
SPISE2009 120

Table 4. Effects of thermal and acidic conditions on inhibitory activities of B. sterothermophilus and
S. cerevisiae -glucosidase.Results are expressed as means  SD, n = 5

Relative inhibitiona (%)

Compounds
Conditions B. sterothermophilus S. cerevisiae
-glucosidase -glucosidase
100 0C for 30 min 1 99.6  0.4 99.4  0.2
2 99.8  0.1 99.8  0.2
3 99.7  0.4 99.8  0.2
pH 2 for 40 min 1 78.4  2.4 92.8  1.2
2 75.3  2.5 90.5  2.1
3 82.7  3.3 99.3  0.1
pH 2 for 60 min 1 69.5  2.5 85.2  3.2
2 67.5  1.7 80.6  1.4
3 67.0  5.7 87.8  0.2

4. Conclusion
One of the therapeutic approaches for preventing diabetes mellitus is to retard absorption of
glucose via inhibition of α-glucosidase. Hence, the search for α-glucosidase inhibitors in
marine organisms is important because these inhibitors could control the postprandial
hyperglycemia of diabetic patients. In this study, three compounds with strong α-glucosidase
inhibitory activity were purified from sea cucumber S. japonicus collected from the eastern
coastal area of the Korean peninsula. Therefore, compounds of S. japonicus can potentially be
developed as a novel natural nutraceutical to prevent diabetes mellitus because of their strong
α-glucosidase inhibitory activity.
The search for α-glucosidase inhibitors in marine organisms is important because these
inhibitors can help control the postprandial hyperglycemia of diabetic patients. Although it is
still not clear whether sea cucumber -glucosidase inhibitors can suppress hyperglycemia, we
found that sea cucumber compounds greatly inhibited yeast and bacterial α-glucosidase

Acknowledgements. This research was supported by the Regional Research Centers

Program of the Ministry of Education, Republic of Korea. T.H Nguyen is the recipient of a
graduate fellowship provided by the Brain Korea (BK21) program sponsored by the Ministry
of Education, Science and Technology, Republic of Korea.

REFERENCES
Apostolidis, E., Kwon, Y.I., & Shetty, K. (2006). Potential of select yogurts for diabetes and
hypertension management. Journal of Food Biochemistry, 30(6), 699-717.
Chiba, S. (1997). Molecular mechanism in α-glucosidase and glucoamylase. Bioscience Biotechnology
& Biochemistry, 61(8), 1233-1239.
SPISE2009 121

Bischoff, H., Puls, W., Karause, H.P., Schutt, H., & Thomas, G. (1985). Pharmacological properties of
the novel glucosidase inhibitors BAY m 1099 (miglitol) and BAY o 1248. Diabetes Research and
Clinical Practice, 1, 53-62.
Iwai, K. (2008). Antidiabetic and Antioxidant Effects of Polyphenols in Brown Alga Ecklonia
stolonifera in Genetically Diabetic KK-Ay Mice. Plant Foods for Human Nutrition, 63(4), 163-
169.
Kanno, M., Li, Q., & Kijima, A. (2005). Isolation and characterization of twenty microsatellite loci in
Japanese sea cucumber (Stichopus japonicus). Marine Biotechnology, 7(3), 178-183.
Kim, K.Y., Nam, K.A., Kurihara, H., & Kim, S.M. (2008). Potent α-glucosidase inhibitors purified
from the red alga Grateloupia elliptica. Phytochemistry, 69(16), 2820-2825.
Kim, Y.M., Jeong, Y.K., Wang, M.H., Lee, W.Y., & Rhee, H.I. (2005). Inhibitory effect of pine
extract on α-glucosidase activity and postprandial hyperglycemia. Nutrition, 21(6), 756-761.
Kimura, A. (2000). Molecular anatomy of α-glucosidase. Trends in Glycoscience and
Glycotechnology, 12(68), 373-380.
Kimura, K., Lee, J.H., Lee, I.S., Lee, H.S., Park, K.H., Chiba, S., & Kim, D. (2004). Two potent
competitive inhibitors discriminating alpha-glucosidase family I from family II. Carbohydrate
Research, 339(6), 1035-1040.
Kitagawa, I., Sugawara, T., & Yosioka, I. (1976). Saponin and sapogenol. XV. Antifungal glycosides
from the sea cucumber Stichopus japonicus SELENKA. (2). Structures of holotoxin A and
holotoxin B. Chemical & Pharmaceutical Bulletin, 24(2), 275-284.
Kurihara, H., Mitani, T., Kawabata, J., & Takahashi, K. (1999). Inhibitory potencies of bromophenols
from Rhodomelaceae algae against α-glucosidase activity. Fisheries Science, 65(2), 300-303.
Lam, S.H., Chen, J.M., Kang, C.J., Chen, C.H., & Lee, S.H. (2008). α-Glucosidase inhibitors from the
seeds of Syagrus romanzoffiana. Phytochemisty, 69(5), 1173-1178.
Lebovitz, H.E. (1997). Alpha-glucosidase inhibitors. Endocrinology and Metabolism Clinics, 26 (3),
539–551.
Lineweaver, H. & Burk, D. (1934). The determination of enzyme dissociation constants. Journal of
the American Chemical Society, 56, 658-666.
Mamelona, J., Pelletier, É., Lalancette, K.G., Legault, J., Karboune, S., & Kermasha, S. (2007).
Quantification of phenolic contents and antioxidant capacity of Atlantic sea cucumber,
Cucumaria frondosa. Food Chemistry, 104(3), 1040-1047.
Nakoa, Y., Maki, T., Matsunaga, S., Soest, R.W.M., & Fusetani, N. (2004). Penasulfate A, a new -
glucosidase inhibitor from a marine sponge Penares sp. Journal of Natural Products, 67(8), 1346-
1350.
Nishio, T., Hakamada, W., Kimura, A., Chiba, S., Takatsuki, A., Kawachi, R., & Oku, T. (2002).
Glycon specificity profiling of α-glucosidases using monodeoxy and mono-O-methyl derivatives
of p-nitrophenyl α-D-glucopyranoside. Carbohydrate Research, 337(7), 629-634.
Oki, T., Matsui, T., Osajima, Y. (1999). Inhibitory effect of -glucosidase inhibitors varies according
to its origin. Journal of Agricultural and Food Chemistry, 47(2), 550-553.
SPISE2009 122

Saijyo, J., Suzuki, Y., Okuno, Y., Yamaki, H., Suzuki, T., & Miyazawa, M. (2008). -Glucosidase
inhibitor from Bergenia ligulata. Journal of Oleo Science, 57(8), 431-435.
Suffness, M., Newman, D.J., & Snader, K. 1989. Discovery and development of antineoplastic agents
from natural sources. Bioorganic Marine Chemistry, 3, 131- 168.
Tewari, N., Tiwari, V.K., Mishra, R.C., Tripathi, R.P., Srivastava, A.K., Ahmad, R., Srivastava, R., &
Srivastava, B.S. (2003). Synthesis and bioevaluation glycosyl ureas as alpha-glucosidase
inhibitors and their effect on mycobacterium. Bioorganic & Medicinal Chemistry, 11(13), 2911-
2922.
Van de Laar, F.A., Lucassen, P.L., Akkermans, R.P., Van de Lisdonk, E.H., Rutten, G.E., & Van
Weel, C. (2005). alpha-Glucosidase inhibitors for type 2 diabetes mellitus, Cochrane Database of
Systematic Reviews 18 (2), p. CD003639.
Zhong, Y., Khan, M.A., & Shahidi, F. (2007). Compositional Characteristics and Antioxidant
Properties of Fresh and Processed Sea Cucumber (Cucumaria frondosa). Journal of Agricultural
and Food Chemistry, 55(4), 1188-1192.
SPISE2009 123

VOLATILE COMPONENTS AND SENSORY CHARACTERISTICS AND

CONSUMER LIKING OF COMMERCIAL BRAND OYSTER SAUCES
T.H.D Nguyen*, X.C. Wang, and Y.Z. Zhu
College of Food Science and Technology, Shanghai Ocean University, Shanghai, China
*Corresponding author: [email protected], Tel: +86 21 61900355

Abstract: The objectives of this study are 1) to compare the volatile compounds, the taste
compounds and the sensory attributes of four commercial oyster sauces, and 2) to determine
the relationship between these compounds and sensory attributes. In addition, a consumer test
was applied to evaluate the overall liking of these four sauces. A total of 74 volatile
compounds were identified, most of these being alcohols, furans, aldehydes and pyrazines .
Most of the samples showed a clear sensory organisation. Glutamic was the dominant free
amino acid in all samples. Canonical correlation analysis showed that sweet amino acids
including threonine, serine, glycine, and alanine were strongly associated with sweet taste.
Three sensory attributes (oyster, fishy and fired pork dour) were highly correlated to alcohols
(1-Penten-3-ol), aldehydes (propanal, butanal) and pyrazines, respectively. Cluster analysis of
hedonic scores showed that there were four clusters of consumers with similar preference.
According to preference mapping, consumers mostly like sample VN which has oyster odour
and sweet taste; By contrast, sample TL was unacceptable because of its soy sauce odour.

Key words: oyster sauce, SPME, sensory attributes, volatile compounds, preference mapping

1. INTRODUCTION
Flavour is one of the important factors in assessing the quality of oyster sauce as a savoury
product. Oyster sauce is a viscous dark brown sauce commonly used in Chinese, Philippines,
and Thai cuisine. In the past, to produce oyster sauce, fresh oysters were boiled, seasoned
with soy sauce, salt, and other spices and then preserved. Together, all these ingredients
created this ancient flavour. Although the cooking process did get rid of the fishy flavour,
some kinds of oyster sauce are on the salty side while others have much less salt than soy
sauce. In some countries, including the UK, the oyster content in some sauces is lower than its
Asian counterparts of the same brand due to laws regulating the import of seafood.
In order for a product to succeed, its flavour needs to be acceptable by the consumer. The
flavour of any food consists of both the aroma and the taste of the food. The taste components
of foods are generally non-volatile while their aroma constituents are volatile (Shahidi, 1998).
However, describing flavour of the product is not an easy task in sensory evaluation. It is
nearly impossible for everyone to interpret and define flavour terms, mainly because each
person has a unique experience of the sense of taste, smell, and ability to articulate this
experience (Park, 2005). However, even though the overlap of sensory attributes can be
challenging, it can, nevertheless, be readily grasped by trained assessors if the reference
chemical compounds are provided during training (Garcia-Gonzalez et al., 2008). In addition,
SPISE2009 124

since the 1960s, the ability of scientists to identify flavour components has expanded with the
use of gas chromatography, mass spectroscopy, and other methods that allow the isolation,
separation, and identification of the minor components found in foods, herbs, and spices
sometimes at concentrations as low as parts per billion and below. Moreover, variety of
traditional products such as blue type chesses, Reggianito chesses, red wine, and biscuits have
been analysed by solid phase microextraction (SPME) coupled with gas chromatography-mass
spectrometry (GC-MS) to study the aroma profile (Jurado et al, 2008, Panseri et al, 2008).
Nevertheless, a frequently asked question about flavour analysis in food industry is “Can an
instrument provide enough information?” In general, instrumental data are relevant issue for
quality control measurements because of instrumental data have advantages such as being
faster and simpler to obtain than sensory evaluation with humans.

The consumer, who is the ultimate user and the target of the food industry, speaks a louder
voice than any instrumental data. This is why consumer liking tests are conducted throughout
stages of product cycle such as: decision making of a new product development, product
maintenance, product improvement, product optimisation, and assessment of market potential.
Preference mapping, which combines description data and consumer acceptance, can be used
to uncover market segmentation and identifies drivers of liking for each segment.
Little research has been published using both preference mapping and instrumental techniques
to understand the relationship among volatile compounds, sensory characteristic, and
consumer acceptability. A product such as oyster sauce can be ideal for performing this type
of integrative research because oyster sauce flavour is characterised by complex components
which associated with specific manufacture methods. Moreover, different demographic
groups may have different attitudes relating to consumer acceptability of the oyster sources.
Consequently, the objectives of the present study are to compare the volatile compounds, the
taste compounds, and the sensory attributes. Specifically, we want to analyse the relationships
between these compounds and attributes as well as consumers hedonic scores in four
commercial brands of oyster sauce.

2. MATERIALS AND METHODS

2.1. Materials
Samples of oyster sauces belonging to four certified brands were obtained from Vietnam,
Thailand, and China (Table 1). Two of the samples were collected from the same company in
China (which is considered the birth place of the oyster sauce). In order to obtain homogenous
samples, all the sauces were obtained from the same batch and the same launch day.
SPISE2009 125

Table 1: Manufacture and ingredients of four kinds of oyster sauce.

Code Country manufacture Ingredients

VN Vietnam Oyster extract, starch, salt, caramel, sugar
TL Thailand Oyster extract (30%), soy sauce, sugar, salt, starch.
CN1 China Oyster mixture (oyster extract, water, salt), sugar, water,
MSG, starch, flour, caramel
CN2 China Oyster mixture (oyster extract, water, salt), sugar, water,
MSG, starch, flour, caramel

2.2 HS-SPME
A manual fibre-holder for SPME and one type of fibre coating Carboxen /
Polydimethylsiloxane CAR/PDMS 75μm was applied. The volatiles were desorbed in the GC
injection port for 7 min at 2700C. The optimised condition of extraction process was the
reference used in previous studies using SPME. Three grams of sauce were placed into 20ml
glassy vial and tightly capped with PTTE septum. The septum covering each vial was then
pierced with SMPMEneedle and the fibre coating was exposed into headspace for 35 min at
600C. A magnetic stirrer and temperature-control water bath was using during the extraction
process.
2.3 GC-MS
Chromatography was performed using an Agilent technology 6890N gas chromatograph
interfaced to an Agilent 5975B inner MSD mass spectrometer. A HP-5MS 5% Phenyl Methyl
Siloxane capillary non-polar column (30m  0.25mm i.d.  0.25μm film thickness) was
applied. The injection was made in splitless mode at a temperature of 2700C. The following
oven temperature program was used: initial temperature 300C hold for 1 min, then an increase
of 20C /min to 1300C; then followed by an increase of 150C/min to 2700C and hold at that
temperature for 5 min. Helium was used as carrier gas with flow of 1.0 ml/min. The mass
spectra were obtained in electron-impact mode (EI) at 70eV using full scan with a scan range
of 30-300m/z at a rate of 2.5s. scan-s. Data acquisition and integration were loaded out with
the ChemStation chromatography software. The compounds present in the volatile profile of
the oyster sauce samples were identified by matching of their mass spectra against the
NIST05 library and RI values. In addition, identifying the unknown peak we used previous
literature reviews and extra-references such as retention indices Adams.

2.4 Taste compounds

Free amino acids (FAAs). FAAs were determined according to the recommended procedure
from the manufacture AAA (L-8500, Hitachi, Tokyo, Japan)
Determination Na, K, and phosphorus. Sodium and potassium were evaluated by atom
absorption spectroscopy (ISO 8070-1987). Phosphorus were converted into molybdenum blue
and then determined by spectrometer (ISO 3946-1982)
SPISE2009 126

2.5 Sensory evaluation

Many traits related to sensory characteristics of products were evaluated by Quantitative
Descriptive Analysis (QDA®) (Stone & Sidel, 2007). Eleven assessors were selected from 25
students of Food Science and Technology College, Shanghai Ocean University. They had
acceptable ability to detect basic taste and common odours (Piper & Scharf, 2006).
About one gram of the oyster sauces was served on a small plate to each of the panellists. Not
all the samples were analysed simultaneously due to the short time rest between two samples.
For each sensory trial, the samples and replicates were served, in a random order, on coded
plates. The temperature of the experimental room was set up at 200C ± 20C and the room was
lighted with fluoresces lights. The panellists evaluated the samples using the QDA® procedure
on 9-point rating scales. Water and a light biscuits were consumed before the first sample to
minimise the effect of first sample as well as between samples as a refresher.
2.6 Consumer test
A total of 125 regular consumers (aged 21 to 53) living in Shanghai China, were recruited
according to their responses to a brief screening questionnaire about sex, age, and product
usage. The term “regular” indicated that the consumer uses oyster sauces (origin, cook, etc) at
least three times a week. In the first part of the questionnaire, consumers indicated their level
of acceptance on a nine-point hedonic scale (“dislike extremely” being 1 and “like extremely”
being 9) of four oyster sauces about colour, odour, taste, after-taste, mouth-feel, and overall
liking. They were asked to “rest” for at least 30 seconds between evaluating samples by
taking a sip of water. Then, the consumers answered the questions related to their product
routine such as frequency, time, place serving, etc. All these questions were translated into
Chinese and explained well to each participant. Oyster samples were coded randomly.

2.7. Statistical analysis

SAS software (version Release 9.1, SAS institute, Cary, NC, USA) was used to analyse
statistically data. Mixed univariate analysis of variance (ANOVA) with panellist and panellist 
sample treated as random factors were applied to the sensory attributes data of four samples.
In addition, the influence and the range differences between assessors, samples, and assessor
 sample on the scores were also used in the model. Principal component analysis (PCA) was
used to study the sensory profiling of four samples with 18 sensory attributes. Analyses were
performed using R software version 2.7. To understand consumers’ responses further, the
hedonic ratings were also analysed using agglomerative hierarchical clustering (AHC).
External preference mapping was performed by combining PCA and the consumer dataset.
SPISE2009 127

Table 3 Codes and mean peak area of volatile compounds identified in four kinds of oyster sauce and
p value of each volatile compound. Note: all volatile compounds were significantly different (p < .01)
between four samples (using ANOVA combined Kruskal Wallis test).

Relative peak area ( % Mean±SD)

Code Name of compounds
CN1 CN2 TL VN

Alcohols
A1 Ethyl alcohol 0.25 ±0.05 0.15 ± 0.01 1.75 ± 0.02 10.83 ± 3.20
A2 1-Butanol 2.98 ± 0.13 1.16 ± 00 N.D. N.D.
A3 1-Pentanol 0.36 ± 0.03 0.64 ± 0.08 N.D. N.D.
A4 1-Penten-3-ol N.D. N.D. N.D. 0.1 ± 0.02
A5 1-Butanol, 2-methyl N.D. N.D. 0.83 ± 0.08 N.D.
A6 2,3-Butanediol N.D. N.D. 0.12 ± 0.18 N.D.
Aldehydes
A7 Acetaldehyde N.D. N.D. 0.16 ± 0.02 N.D.
A8 Hexanal N.D. N.D. 0.08 ± 0.00
A9 2-Butenal N.D. N.D. 0.64 ± 0.09 N.D.
A10 Butanal, 3-methyl N.D. N.D. 23.37 ± 1.63 0.3 ± 0.12
A11 Butanal, 2-methyl 0.94 ± 0.06 0.57 ± 0.03 14.91 ± 1.18 0.07 ± 0.02
A12 Propanal, 2-methyl 1.48 ± 0.08 0.52 ± 0.02 4.59 ± 0.20 N.D.
Propanal, 3-
A13 N.D. N.D. 0.55 ± 0.10 N.D.
(methylthio)
Furans
A14 Furan 2.37 ± 0.39 0.84 ± 0.00 1.16 ± 0.03 0.47 ± 0.13
A15 Furan, 2-methyl 0.52 ± 0.02 0.72 ± 0.04 1.63 ± 0.13 N.D.
A16 Furan, 2-pentyl N.D. 0.09 ± 0.04 N.D. N.D.
A17 Furan, 2-ethyl N.D. 0.07 ± 0.00 N.D. N.D.
A18 Furfural N.D. N.D. 2.37 ± 0.09 0.58 ± 0.24
A19 3-Furanmethanol 16.65 ± 1.62 16.92 ± 061 N.D. N.D.
A20 Furan, 2,5-dimethyl N.D. N.D. 0.17 ± 0.02 N.D.
3(2H)-Furanone,
A21 1.2 ± 0.00 N.D. 0.28 ± 0.03 0.08 ± 0.01
dihydro-2-methyl
2-
A22 Furancarboxaldehyde N.D. N.D. 0.2 ± 0.04 0.05 ± 0.03
, 5-methyl
Pyrazines and pyridines
Pyrazine, 2,5-
A23 6.01 ± 0.62 15.28 ± 0.77 0.13 ± 0.02 0.1 ± 0.04
dimethyl
Pyrimidine, 4,6-
A24 3.19 ± 0.37 3.93 ± 0.38 0.97 ± 0.19 N.D.
dimethyl
Pyrazine, 2,3-
A25 0.86 ± 0.09 1.11 ± 0.16 0.15 ± 0.05 N.D.
dimethyl
A26 Pyrazine, methyl 11.91 ± 0.85 11.67 ± 0.68 3.92 ± 0.66 0.06 ± 0.00
A27 Pyrazine 6.15 ± 0.06 5.63 ± 0.04 N.D. N.D.
A28 Pyridine 0.39 ± 0.01 0.67 ± 0.20 N.D. N.D.
A29 Pyridine, 2-methyl N.D. 0.07 ± 0.04 N.D. N.D.
SPISE2009 128

Pyrazine, 2-ethyl-6-
A30 0.7 ± 0.16 0.7 ± 0.15 N.D. N.D.
methyl
Pyrazine, 2-ethyl-5-
A31 0.55 ± 0.29 2.18 ± 0.26 N.D. N.D.
methyl
Pyrazine, 2-ethyl-3-
A32 0.47 ± 0.23 0.39 ± 0.07 N.D. N.D.
methyl
A33 Pyrazine, trimethyl 0.63 ± 0.12 2.08 ± 0.25 0.05 ± 0.01 0
Pyrazine, 3-ethyl-2,5-
A34 0.38 ± 0.4 2.21 ± 2.5 N.D. N.D.
dimethyl
Ethanone, 1-(1H-
A35 N.D. 0.15 ± 0.06 N.D. N.D.
pyrrol-2-yl)
Pyrazine, 2-ethyl-3,5-
A36 N.D. 1.08 ± 1.91 N.D. N.D.
dimethyl
A37 Pyrazine, 2,6-diethyl N.D. 0.14 ± 0.02 N.D. N.D.
Pyrazine, 2,3-diethyl-
A38 N.D. 0.12 ± 0.03 N.D. N.D.
5-methyl
Pyrazine, 3,5-diethyl-
A39 N.D. 0.2 ± 0.05 N.D. N.D.
2-methyl
Pyrazine, 3,5-
A40 N.D. 0.08 ± 0.02 N.D. N.D.
dimethyl-2-propyl
2,3,5-Trimethyl-6-
A41 N.D. 0.12 ± 0.03 N.D. N.D.
ethylpyrazine
Pyrazine, 2,5-
A42 dimethyl-3-(2- N.D. 0.05 ± 0.02 N.D. N.D.
methylpropyl)
Esters
A43 Ethyl Acetate N.D. N.D. 2.09 ± 0.13 23.23 ± 2.52
1,2-Propanediol, 2-
A44 N.D. N.D. N.D. 0.6 ± 0.06
acetate
1,2-Propanediol,
A45 N.D. N.D. N.D. 0.23 ± 0.12
diacetate
Ketones
A46 Acetone 16.45 ± 0.92 8.69 ± 2.11 N.D. N.D.
A47 2-Butanone 8.51 ± 0.4 8.6 ± 0.93 0.92 ± 0.11 N.D.
2-Propanone, 1-
A48 2.75 ± 0.17 3.44 ± 0.52 0.62 ± 0.10 N.D.
hydroxy
2-Propanone, 1-
A49 N.D. 1.39 ± 0.408 N.D. N.D.
methoxy-
2-Butanone, 3-
A50 0.41 ± 0.00 0.6 ± 0.02 N.D. N.D.
hydroxy
A51 2,3-Butanedione N.D. N.D. 5.81 ± 0.21 N.D.
Ethanone, 1-(2-
A52 0.77 ± 0.05 N.D. 0.72 ± 0.16 0.15 ± 0.10
furanyl)
A53 3-Pentanone N.D. N.D. 0.18 ± 0.03 N.D.
Sulfur compounds
A54 Dimethyl sulfide N.D. N.D. 10.34 ± 0.12 4.41 ± 1.61
A55 Disulfide, dimethyl N.D. N.D. 0.63 ± 0.080 0.05 ± 0.01
A56 Methanethiol N.D. N.D. 0.17 ± 0.04 N.D.
Amines
A57 Trimethylamine N.D. 0.54 ± 0.19 N.D. N.D.
Chlorinated compounds
A58 Trichloromethane 1.78 ± 0.31 N.D. 0.94 ± 0.06 N.D.
Aromatics
A59 Benzaldehyde 0.34 ± 0.37 0.04 ± 0.00 0.21 ± 0.05 N.D.
SPISE2009 129

A60 Thiophene 0.08 ± 0.01 0.12 ± 0.00 N.D. N.D.

A61 Limonene N.D. 0.17 ± 0.13 N.D. 0.18 ± 0.20
Acids
A62 Acetic acid 5.07 ± 0.53 4.71 ± 0.2 8.27 ± 0.25 N.D.
A63 Propanoic acid 0.4 ± 0.36 0.27 ± 0.17 N.D. N.D.
A64 Butanoic acid N.D. 0.09 ± 0.01 N.D. N.D.
Benzenecarboxylic
A65 N.D. N.D. 1.17 ± 0.06 0.41 ± 0.90
acid
A66 Sorbic Acid N.D. N.D. 0.21 ± 0.31 N.D.
Unknown
A67 Triacetin N.D. N.D. N.D. 48.89 ± 8.41
Oxazole, 4,5-
A68 N.D. 1.845 ± 0.57 N.D. N.D.
dimethyl-
1,3-Dioxane, 2-
A69 N.D. N.D. 0.34 ± 0.01 N.D.
methyl
1,3-Dioxolane, 2-
A70 N.D. N.D. 0.25 ± 0.13 N.D.
heptyl-4-methyl
A71 4-Methylthiazole N.D. 0.37 ± 0.1 N.D. N.D.
A72 Thiazole, 2-methyl N.D. 0.15 ± 0.01 N.D. N.D.
A73 Methacrolein N.D. N.D. 0.07 ± 0.01 N.D.
A74 Propylene Glycol N.D. N.D. 4.73 ± 2.07 5.08 ± 0.90
A75 Ethyl ether N.D. N.D. 0.11 ± 0.005 N.D.

N.D. means non-detected, and the same as below.

3. RESULTS AND DISCUSSIONS

3.1 Analysing the volatile and semi-volatile compounds dataset
Table 3 shows the chemical volatiles well identified (% Mean ± SD) in oyster sauces
manufactured from different processing method and countries. A total of 75 compounds were
figured out, including esters, alcohols, aldehydes, sulphur compounds, pyrazines, furans,
ketones, aromatics, acids, and chlorinated compounds. Amongst of volatiles, the table
indicates that pyrazines, aldehydes, and ketones were dominant. The same results were
reported in agreement with previous studies (Shahadi, 1998).
To figure out the overall profile of volatiles in four samples, we applied PCA (see Figure 2).
The first principal component (PC-1) explained 55.82% of the variance across four samples
while PC-2 explained 25.74% of the variance. We decided to keep only two factors for further
analysis and to analyse the circle of correlation showing the correlations (i.e., loadings)
between the PCs and the variables used in the analysis (see Figure 2). We divided the plane
made by the first two components into four quadrants (called Q1 to Q4). In the first quadrant
Q1, the CN2 could be represented by most compounds of pryrazines such as pyrazine 3-ethyl-
2,5-dimethyl, pryazine 2,6-diethyl, pyrazine 2,3-diethyl-5-methyl, ketons such as acetone, 2-
butanone, and furans such as furan, furan 2-pentyl, 3-furanmethanol. The opposite side Q3
contains few volatiles but most of the compounds are the alcohols such as 1-pental-1-ol,
propandiol and triacetin. These compounds were the dominant volatiles of sample VN. Two
SPISE2009 130

samples CN1 and TL located in Q2 and Q4, respectively. The volatile contributions of the
quadrant Q4 are aldehydes such as acetaldehyde, butanal 3-methyl, butanal 2-methyl,
propananl 2-methyl, and propanal 3-methylthio-, with the exception hexanal. Aromatic and
chlorinated compounds constitute the volatile representative of Quadrant Q2. The location of
volatile compounds in the PCA plot and the quantitative values of each sample are explored
further in the next part of this study.

3.2 Analysing the taste compounds dataset

The taste-active constituents are water-soluble low-molecular weight components. The most
important compounds are free amino acids (FAA) and inorganic salts. Table 5 illustrates the
content of the FAAs and minerals in four commercial brand oyster sauces.
Table 5. Free amino acids (FAAs) composition of four kinds of oyster sauce (p < 0.01) a
Schlichtherle-Cerny Grosch (1998),.

No. Amino acids (%) VN TL CN1 CN2 Tastea

1 Aspartic acid 0.06 2.82 0.01 0.31 Umami
2 Threonine N.D. 1.44 N.D. 0.05 Sweet
3 Serine N.D. 1.92 N.D. 0.05 Sweet
4 Glutamic acid 96.49 65.73 10.51 90.61 Umami
5 Glycine N.D. 6.35 0.03 1.9 Sweet
6 Alanine 0.03 0.5 0.04 2.1 Sweet
7 Cystine 1.44 0.99 0.06 0.51 N.D.
8 Valine N.D. 2.25 N.D. 0.12 Bitter
9 Methionine N.D. 0.72 N.D. N.D. N.D.
10 Isoleucine N.D. 1.76 N.D. 0.05 Bitter
11 Leucine 0.09 3.17 0.01 0.15 Bitter
12 Phenylalamine 0.95 1.54 N.D. 0.36 Bitter
13 Lysine N.D. 2.38 89.23 0.56 Bitter
14 Histidine 0.92 7.19 0.11 3.21 Bitter
15 Arginine N.D. 0.99 N.D. N.D. Bitter
Total (%)
Umami FAAs 96.55 68.55 10.52 90.92
Sweet FAAs 0.03 10.21 0.07 4.1
Bitter FAAs 1.96 20 89.35 4.45
Unkown FAAs 1.47 1.25 0.06 0.53
Minerals (mg/100g)
Phosphate 2437.58 ± 1551.222 ± 3495.264 ± 5286.685 Sour- Salt
111.98 196.35 0.00 ± 112.07
Sodium 3616.24 ± 2697.26 ± 4170.614 ± 4039.677 Sour- Salt
0.00 0.00 0.00 ± 0.00
Potassium 18.40017 ± 104.6486 ± 44.53474 ± 137.8532 Sour- Salt
0.00 0.00 0.00 ± 0.00
SPISE2009 131

All these FAAs were observed in previous reports related to oyster sauce and oyster raw
materials (Shadihi, 1998). Glutamate (Glu) was the highest among all kinds of FAAs. Its
amount in relative proportions was remarkable; the percentage of these FAAs in each product
was 96.5, 65.7, 10.5, and 90.6%, respectively. The result was similar to fermented oyster
sauce (Young & Park, 2005) with the highest percentage of Glu in free amino acid
composition. Proportions of the content between FAAs and the tasty one could contribute to
oyster sauce taste. Free amino acids, such as Glu, Glycine (Gly), Lysine (Lys) and
Phenylalamine (Ala) were abundant and are recognised as being important in the tastes of fish
and shellfish sauces (Young & Park, 2005).
The most interesting finding in mineral composition is that the concentrations of mineral are
much higher that threshold, excepting for potassium in CN1 and VN sample. (Schlichtherle-
Cerny, 1998). The highest concentration of sodium is found in CN1, followed by CN2. We
found the same pattern of results for phosphate content with the CN1 and CN2 samples. The
taste of oyster sauce was described as savoury and desirable, which corresponds to descriptors
of saltiness, umami, sweetness, and other tastes. The relationships between the taste sensory
attributes and non-volatile compounds will be determined in the next part of this study.

3.3 Analysing the sensory dataset

Table 6 shows the definitions of the sensory attributes assessed by the panellists and Figure 7
their mean intensities when evaluating the four kinds of oyster sauces. From the ANOVA
results conducted on the descriptive data of the four oyster sauce samples, we find that most
of the attributes were significantly different across samples (p < .001 for oyster odour, fishy
odour, salty, and fried pork) and some colour and appearance attributes (p < .01)
distinguished the oyster sauce. A significant assessor  sample interactions were found for
attributes. The replications and replication  assessor were not a significant source of
variation for the sensory attributes.
Mean intensity and SD of the four samples are given in Table 8. Samples CN1 and CN2,
which were manufactured under the same method, received similar rating for almost all
sensory attributes such as appearance (colour, texture), fishy odour, fried pork odour and
some kinds of taste. The higher intensities in typical odour attributes of sample such as oyster,
soy sauce for samples VN and TL may have been due to different ingredients and processing.
In contrast, samples CN1 and CN2 showed different sensory profiles than the two other
samples, with higher intensities in brow, stickiness, and fried pork odour, and with lower
intensity in oyster and soy sauce odours. In addition, sample CN2 was rated as very intensive
in saltiness and umani taste. This can be explained by the traditional Chinese oyster sauce
manufacture process which uses fresh oyster broth and salt.
SPISE2009 132

Table 6: List of sensory attributes of oyster sauce with descriptions.

Sensory attributes Definition

Appearance
Brown The colour of chocolate
Stickiness The quality of being sticky. Sticky is covered with a substance that
stays fixed to any surface it touches
Smoothness The quality of being smooth. Smooth is having a surface or substance
that is perfectly regular and has no holes, lumps or areas that rise or
fall suddenly. Your hand move cross it and feel it flat.
Thickness The quality of being thick. Thick is the solution with high
concentration of substance
Melting-mouth feel The quality of easiness of being melted. To turn something becomes
soft.

Odour
Oyster The odour associated with oyster hydrolysed solution1
Fishy The odour associated with fish sauce
Soya sauce The odour associated with soy sauce
Fried pork The odour associated with Chinese “rou song” fried pork
Caramel The odour associated with caramel

Taste
Sweet The taste on the tongue associated with sucrose solution
Sour The taste on the tongue associated with citric acid
Bitter The taste on the tongue associated with caramel solution
Salty The taste on the tongue associated with salt solution
Umami The taste on the tongue associated with Monosodium Glutamate

Aftertaste
Sweet The taste left the tongue after swallowing and associated with
sucrose solution
Umami The taste left the tongue after swallowing and associated with
monosodium glutamate
Bitter The taste left the tongue after swallowing and associated with
caramel solution
1
Oyster hydrolysed solution made from oyster and papaya enzyme

Table 7. Significant levels for the sensory attributes of four kinds of oyster sauce (***, **, * indicate
significant differences using F test at p < .001, p < .01, p < 0.5, respectively, using a mixed ANOVA model
with sample as random effect. NS: non significant.)

Sensory attributes Assessor Sample Assessor* Sample Replication

Brown ***a *** *** NS
Stickiness *** ** *** NS
Smoothness NS *** *** NS
Thickness *** *** *** NS
Melting-mouth feel *** *** *** NS
Oyster *** *** *** NS
Fishy *** *** *** NS
Soya sauce *** *** *** NS
Fried pork NS *** *** NS
SPISE2009 133

Carmel odour * * *** NS

Sweet *** NS *** NS
Sour *** NS *** NS
Caramel Bitter *** NS *** NS
Salty *** * *** NS
Umami *** NS *** NS
After-sweet *** NS *** NS
After-umami *** NS *** NS
After-bitter *** NS *** NS

Table 8. Mean value for sensory attributes of four oyster sauces

Sample
No. Distributes
VN TL CN1 CN2 P
1 Brown 3.45 ± 0.82 4.18 ± 1.05 6.77 ± 1.14 7.6 ± 0.83 <0.0001
2 Stickiness 3.9±0.80 4.04 ± 1.44 3.13 ±1.36 3.13 ± 1.39 0.017
3 Smoothness 2.13 ± 0.23 2.18 ± 0.33 3.27 ± 1.05 3.45 ± 1.40 0.0001
4 Thickness 5.86 ± 1.70 5.2 ± 1.73 6.59 ± 1.20 6.7 ±1. 06 0.0001
5 Melting-mouth feel 2.5 ± 1.20 1.9 ± 0.43 3.09 ± 1.28 3.77 ± 1.77 0.0012
6 Oyster 5.5 ± 0.74 1.86 ± 0.63 1.82 ± 0.9 2.0 ± 0.86 <0.0001
7 Fishy 3.27±2.30 2.45 ± 1.44 3.27 ± 1.99 3.54 ± 2.27 <0.0001
8 Soy sauce 2.90 ± 1.67 6.68 ± 1.47 3.13 ± 1.84 3.72 ± 1.97 <0.0001
9 Fried pork 2.09 ± 1.42 2.27 ± 1.47 5.0 ± 1.78 5.86 ± 1.87 <0.0001
10 Caramel 3.49 ± 2.39 3.77 ± 2.00 2.95 ± 1.61 3.3 ± 1.12 0.68
11 Sweet 4.95 ± 1.60 4.04 ± 1.70 4.13 ± 1.68 4.27 ± 1.70 0.33
12 Sour 3.8 ± 2.19 3.4 ± 2.02 4.0 ± 1.68 3.45 ± 2.20 0.58
13 Bitter 3.13 ± 1.79 3.36 ± 2.23 3.86 ± 1.98 3.72 ± 1.76 0.53
14 Salty 5.73 ± 1.72 6.18 ± 1.43 6.41 ± 0.86 7.0 ± 1.12 0.02
15 Umami 5.54 ± 1.94 5.22 ± 1.91 5.72 ± 1.87 6.0 ± 1.70 0.32
16 After-sweet 3.81 ± 2.08 3.81 ± 1.79 3.41 ± 1.26 3.4 ± 1.49 0.77
17 After-umami 4.72 ± 2.02 4.77 ± 2.28 5.31 ± 2.45 5/63 ± 2.46 0.13
18 After-bitter 2.4 ± 2.13 2.45 ± 1.43 3.18 ± 2.04 3.27 ± 2.10 0.26

In order to determine the overall sensory evaluation, PCA was used on all sensory attributes.
The results of the PCA are shown in Figure 2 where we plotted the “circle of correlation” along
with the correlations (i.e., loadings) of the sensory attributes with the PCs. For ease of
interpretation, we have number the four quadrants from 1 to 4. In the first quadrant, saltiness,
fried pork odour, and aftertaste perception characterise sample CN2. The second quadrant
shows soy sauce odour and stickiness which characterise the TL sample. In the fourth
quadrant melting and fishy odour characterise sample CN1. The third quadrant corresponds to
the oyster odour and sweet taste. The first two PCS explained 91.15% of the total variance and
therefore we can trust the positions of the variables displayed in Figure 2.
SPISE2009 134

1.0
Q2 S Q1
Likeness Salty
Stickiness Caramel bitter
0.5 Brown

Dimension 2 (25.62%)
Caramel Fried pork
Smoothne
Bitter
U i
0.0
Sweet aftertaste
Melting
Thickness
Umami
-0.5
Sour Fishy

Q3 Q4
-1.0 Oyster
Sweet

-1.0 -0.5 0.0 0.5 1.0

Dimension 1 (65.53%)

Figure 2. Plot of principal component analysis of the sensory attributes of four samples.

3.4 The relationship between sensory attributes, taste compounds and volatile
compounds
To determine the relationships between taste compounds and sensory data in the four oyster
sauce samples, we used canonical correlation. The correlation coefficients with each taste
sensory attributes are presented in Table 9 where we can see that the sweet FAAs show a high
correlation (r = .99) with sweet taste (including sweet aftertaste). Sample CN1, which had the
highest amount of bitter FAAs (89.36%), also showed the highest intensity in bitter taste.
However, this relation between the concentration of bitter FAA and intensity of perceived taste
does not hold for the other samples maybe because the bitter taste is also connected to the
hydrophobic peptides (Schlichtherle-Cerny, 1998). Components related to umami such as
FAAs, inosine monophosphate (IMP) are considered important contributors to the sensory
quality (Dang et al., 2008). However there was no significant difference of umami intensity
between our four samples (e.g., the sample VN had the highest concentration of umami FAAs
but yet received nearly the lowest rating in sensory profile). Therefore, it seems that the
correlation between FAAs and taste intensity is likely to depend on many other factors such as
the presence of peptides and some other flavour compounds, as well as their synergistic
effects (Schlichtherle-Cerny, 1998).
There has been little work on the flavour of minerals in food. Fuke and Konosu (1991)
believed that inorganic salts such ad Na+, K+, Cl- and PO43- contributed to fish flavour
(Shahidi, 1998). Case in point, CN1 and CN2 had the highest amount of minerals, and CN1
received the highest for fishy odour. This result is corroborated by canonical correlation, with
phosphate (r = .84) and sodium (r = .92). In addition, salty is highly correlated to phosphate
SPISE2009 135

and potassium with values of r equal to (respectively) .77 and .76. However salt was poorly
correlated to free amino acids.

PCA was applied to analyse the relationships between volatile compounds and sensory
attributes. The peak area values of volatile compounds were projected onto the plot built with
the sensory information evaluated by trained panellists. The information content is indicated
by the position of the compounds and sensory characteristic as well as the distance between
them. As a rule of thumb, the vicinity of a volatile to an attribute illustrates a good
relationship between them, but this interpretation may depend on other surrounding sensory
attributes. In addition, the distance of a volatile to the centre of the circle reflects how much
this volatile contributes to the aroma (Garcia-Gonzalez et al., 2008). This means that the
volatiles close to the centre of the circle will take up less contribution than the volatiles close
to the perimeter of the circle. From Figure 3 we can see that the most important volatile
compounds contributing to oyster sauce odour include furan 2-methyl, aldehydes, pyrazine
methyl, ethanone 1,2-furanyl, acid acetic. In Figure 3, we also see that most of the volatiles
are located in Quadrants Q1 and Q4. As mentioned above, Q1 was characterised by the
presence of most of the pyrazines, pyridines (excepting pyridine 3-methyl), furan 2-pentyl, 3-
furanmethol, furfural, ketons (excepting 2,3 butanedione, 3-pentanone). In addition, we found
in Q1 the sample CN2 and the variables from the sensory dataset, fried pork, salty taste, and
caramel bitter taste. We cannot, however, explain the presence of these pyrazines and ketons,
particularly the pyrazines near the fried pork flavour. The soy sauce flavour location is close
to all kinds of aldehydes excepting hexanal, alcohols such as 2,3-butanediol, methaneniol, and
1-3 dioxane 2-methyl, 1,3-dioxane 2-heptyl-4-methyl. As the result, Q2 could be defined by
the aldehydes, 2,3-butanediol group and the soy sauce odour as well. This location seemed to
agree with previous studies about soy sauce odour (Yan et al., 2008). One surprise is, in fact,
to find the sulphur compounds near the caramel odour and opposite to fishy odour which is
believed to be associated to these compounds. This surprising position could be explained by
the low concentration of the sulphur compounds in the samples which may be too low to
affect flavour.
In the third quadrant, we found compounds including 1-penten-3-ol, hexanal, 1,2-
propanediol-2-acetate, 1,2-propandiol diacetate and oyster odour. Therefore, the relationship
between oyster and those compounds has been observed. In previous studies, 1-penten-3-ol
and hexanal have been believed to contribute the odour of oyster (Shahidi, 1998).
 SPISE2009 136

A63A54 15
10 A17
Q1 TL
Q4 A60
Soy sauce
A14 A16
A22
A11
A13
A15
A53
A55
A67
A71
A70
A74
A3
A2
A9
A8
A7
A12
10
A57 A59
A20
A24 Salty
05 Caramel A28
A66 05
A26
CN2
Dim 2 (29 74%)

A56 Caramel
A50 bitter
A27
A49
Fried pork
A23 A35
A19
A18
A31
A44
A43
A42
A41
A40
A39
A38
A37
A51
A58
A65
A69
A73
A36
A33
A25
A5
A30
A52
A61
A72
A34
A32
A21
A29
A48 Bitter 00
A64Umami aftertaste
00 Sweet aftertaste A4 CN1
A75
Umami
05

-0.5
Fishy 10
A62
Sour VN
Oyster 15
Q3
A1
A45
A10
A47
A46
A68
A6
Sweet
Q2
-1.0

10 05 00 05 10 -1 0 1 2
Di 1 (55 82 %)
Dim 1 (55 82 %)
A4, A5, A18, A19, A21, A25, A26, A27,

A28, A29, A30, A32, A33, A34, A35, A36,

Figure 3. Plot of principal component analysis of the sensory attributes and volatile compounds and
individual factors. The codes are listed in Table 3

The next step was to determine the variance of each of the sensory traits that could be
correlated to individual compounds of the volatiles dataset. Table 9 shows the correlation
detected between the sensory attributes and the volatiles identified and quantified in the four
experimental samples. Only the larger r-coefficients were illustrated in the table and, as
expected, the highest r (> .97) was found with sweet taste and odour. This strong correlation
maybe due to volatile compounds being mostly linked to odour perception (see, e.g., Park,
2005)
Table 9. The statistical correlation value between sensory attributes and volatile compounds identified
in oyster sauces
Sensory attribute Volatile compounds R-coefficient
Oyster A1, A4, A5, A6 > 0.97
A2, A13, A15 > 0.96
Fishy
A61, A62 > 0.99
A8, A9, A10, A11 > 0.96
Soya sauce
A51 > 0.99
A17, A21, A42, A45, A50, A51, A52, A53, A54, > 0.98
Fried pork
A55, A58, A64, A65, A68, A69, A72, A73, A74
Caramel A14, A26 > 0.95
Sour A31, A32, A33, A35 > 0.90
3.5. Consumer test
Significant differences (p < .05) occurred among demographics for consumer’s overall liking.
Sample VN was rated higher than any of its counterparts, while sample CN2 gave lowest
SPISE2009 137

scores. Sample TL took the second position of grades, ahead of sample CN1. However,
consumers’ rates on distributions (colour, odour, taste, after taste, mouth feel) did not show
any clear and significant clustering.

Analysis of the overall acceptance data by Agglomerative Hierarchical Clustering (AHC)

identified four consumer clusters including cluster 1 (26.4%, N = 33), cluster 2 (24%, N = 30),
cluster 3 (32%, N = 40) and cluster 4 (17.6%, N = 22). Cluster analysis of consumers’ overall
liking data showed that four clusters of consumers with similar preference patterns could be
formed.

Figure 3: Dendrogram of the four major consumer clusters idendified for all categories

Finally, the relationships between sensory attributes and consumer liking were analysed with
preference mapping (PREFMAP). PREFMAP was performed on the coordinates of the oyster
source in the two-dimensional factor space and on the ratings given by the consumers,
summarised by the standardised ratings for the four clusters. Results indicate that consumers
of Cluster 1 preferred VN and TL because of their light brown colour, and sweet aftertaste,
but that VN was preferred because of its sweet taste and oyster odour. In Cluster 2, the
participants liked TL’s strong soy sauce odour, sour taste and caramel bitter taste, stickiness
in texture but did not like VN’s oyster and fishy flavour and sour taste. By contrast,
participants of Cluster 3 liked CN1and CN2’s dark brown colour, smooth, thick, melt texture,
fried pork flavour, umami after-taste, flavour of oyster aroma, fish flavour, and sweet-sour
taste. The participants of Cluster 4 preferred VN because of its sweet taste and oyster odour,
(like the participants of Cluster 1) also because of its moderate brown colour, rough thin, melt
texture and sweet after-taste.
PREFMAP shows that the sample TL which is characterised by a strong caramel flavour, soy
sauce flavour, and salty taste, is not liked by all clusters. By contrast, sample VN is the most
preferred one for most clusters, especially Clusters 1 and 4.
SPISE2009 138

Figure 4. Preference map of combined consumer data with descriptive analysis results.

4 CONCLUSIONS
This study shows the differences in volatile compounds, taste compounds and sensory
attributes of four commercial brands of oyster sauces. Each sample had the typical odour of
the oyster sauce, but there were significant differences in taste perception between these four
samples. These differences in taste can be explained by differences in ingredients and
processing of oyster extracting (fermentation or boiling). Overall, the CN1 and CN2 were
very similar in volatile composition and sensory characteristic. This similarity could be
explained by the similarity of processing and ingredients of two these samples. We found also
a strong relationship between the volatile, taste compounds and sensory attributes. In
particular, sweet taste highly correlated with the concentration of sweet free amino acids. The
flavour of oyster, soy sauce, and fried pork odour were correlated with the concentration of 1-
penten-3-ol, aldehydes and pyrazines.

Consumer liking of the four samples was evaluated at the same time. Preference mapping
revealed the relationships between consumer clusters and sensory descriptions. We can draw
the conclusion that sample VN was accepted by most of consumer clusters because of its
sweet taste and oyster odour, but also because the presence of 1-penten-3-ol and sweet free
amino acids. Although the complexity of the problems means that we could not explain every
sensory characteristic from their bio-chemical composition, we feel that the correlation
between the concentration of aldehydes and the perception of soy sauce odour is large enough
to recommend that producers avoid incorporating these aldehydes in their formulation.
Finally, we need to keep in mind that consumers like a specific product not only because of its
sensory attributes, but also because of its price, nutritional value, etc (Park, 2005). Further
studies should explore more deeply complex relationships between different characteristics of
SPISE2009 139

products. This will help food companies to find the “multi-aspects” important for their
marketing and product development.
Acknowledgement
We gratefully thank the members of the Lab of Nutrition and Food Safety, Shanghai Ocean
University for their excellent help.

REFERENCES
Dang, Y.L., Wang, Z., & Xu, S.Y. (2008). Methods for extracting the taste compounds from
water soluble extract of Jinhua ham, Eur Food Res Technol 228 (1), 93–102
Fuke, S., Konosu, S., (1991). Tast active components in some foods: A review of Japanese
research, Physiology and Behaviour, 49, 863-868
Garcıa-Gonzalez, D.L., Tena, N., Aparicio-Ruiz, R., & Morales, T.M. (2008). Relationship
between sensory attributes and volatile compounds qualifying dry-cured hams, Meat
Science 80 (2), 315-325
ISO (1988). Determination of the moisture content and volatile matter contents of animal or
vegetable fats and oil, ISO 662, McGraw Hill
Jurado, J.M., Ballesteros, O., Alcázar, A., Pablos, F., Martín, M.J., Vílchez, J.L., & Navalón,
A. (2008). Differentiation of certified brands of origins of Spanish white wines by HS-
SPM GC and chemometrics, Anal Bioanal Chem 390, 961-970
Park, J.W. (2005). Surimi and Surimi Seafood. CRC Press, Boca Raton, FL. 923p
Piper, D. & Scharf, A. (2006) Descriptive analysis- state of the art and recent developments,
Series Sensory Analysis No. 1, Goettingen, 2nd ed., 28 pp.
Shahadi, F. (1998). Flavor of meat, meat products and seafoods. 2nd ed. Blackie academic &
Professional, New York, USA
Schlichtherle-Cerny, H. & Grosch, W. (1998). Evaluation of taste compounds of stewed beef
juice, Z Lebensm Unters Forsch A, 207 (5), 369–376
Stone, H. & Sidel, J.L. (2007). Sensory Evaluation Practices, Elsevier (Singapore) Pte Ltd,.
Yan, L.J., Zhang, Y.F., Tao, W.Y., Wang, L.P., & Wu, S.F. (2008) Rapid determination of
volatile flavor components in soy sauce using head space solid-phase microextraction and
gas chromatography-mass spectrometry, Chin J Chromatogr , 285–291.
SPISE2009 140

RELATIONSHIP BETWEEN SENSORY DESCRIPTIVE AND CHEMICAL

PROPERTY OF JIAOGULAN (Gynostemma pentaphyllum.) TEA WITH LIME JUICE
N. Utama-ang* and S. Jaisam
Department of Product Development Technology, Faculty of Agro-Industry,
Chiang Mai University, Thailand.
*Corresponding author: [email protected], Tel::+66-53-948230

Abstract
The objective of this research was to study the relationship between sensory and chemical
properties of Jiaogulan tea with lime juice (JTLJ) fortified with Jiaogula extract. We used
mixture design methodology to create eight formulas of JTLJ. The chemical and physical
factors were moisture content, total soluble solid, total acidity, total saponin, gypenoside,
ginsenoside Rb1 and colour (L*, a*, b*, E). The hybrid sensory descriptive analysis was
conducted by ten panellists. They generated 13 attributes to rate the beverage. The results
showed that a principal component analysis could explain 88.71% of the total variance with
two components. The first principal component explained 50.44% of the variance and
showed that, on one hand, total soluble solid was associated to sweet and sweet aftertaste and
that, on the other hand, the moisture, total saponin, ginsenoside Rb1, gypenoside were
associated with bitter and astringent aftertastes. The second principal component explained
38.28% of the variance and showed that the yellow colour, clearness, lime aroma, Jiaogulan
aroma, lime flavour, Jiaogulan flavour and total acidity were negatively correlated with pH
and b*.
Key Words: Jiaogulan, descriptive analysis, PCA, Gypenoside, Saponin

1. INTRODUCTION
Gynostemma pentaphyllum Makino (Cucurbitaceae), known as Jiaogulan in Chinese herbal
medicine, is a perennial vine endemic in southern China, Japan, India, and Korea (Blumert &
Liu, 1999). In oriental countries, G. pentaphyllum is a well known edible and medicinal plant
(Hu, Chen & Xie 1996). Recently, G. pentaphyllum has attracted great attention owing to its
anti-tumour activities (Zhou, Liang, & Hu, 2001), anti-gastric ulcer effect (Rujjanawate,
Kanjanapothi, & Amornlerdpison, 2004), immunomodulatory effect (Qian, Wang, & Tang
1998) and potential for treating hyperlipidaemia (Cour, Molgaard & Yi, 1995). Because of
this recent interest, the cultures of G. pentaphyllum and their extraction process in health care
have been put into large scale production. Phytochemical studies of this plant had identified
about 90 dammarane-type glycosides, mainly named gypenosides (Gy), and closely related to
the ginseng saponins. Indeed, gypenosides III, IV, VIII, XII and malonyl gypenosides III and
VIII are identical to ginsenosides Rb1, Rb3, Rd, F2, and malonyl ginsenosides Rb1 and Rd
(Takemoto et al., 1983; Kuwahara et al., 1989; Piacente & Pizza, 1995).
SPISE2009 141

Even though G. pentaphyllum is said to be one of the most commonly used medicinal plants
in many countries in Asia, especially China, there is no report of the sensory properties of this
plant. Therefore, the objective of this research was to investigate the sensory descriptive
analysis and to study the relationship between sensory and physical-chemical properties of
Jiaogulan tea product.
2. MATERIALS AND METHODS
2.1 Preparation of Jiaogulan powder and Jiaogulan extract
Jiaogulan leaves were purchased from the Royal Project, Chiang Mai, Thailand. The fresh
leaves were washed and dried by vacuum microwave drier until the moisture content was
below 7%, as specified by the Thai Industrial Standard Institute of green tea TIS 460-2526
(TISI, 1983). After that, it was milled and packed in tight closed container. About 5 g of
Jiaogulan powder was extracted with distilled water by the microwave extraction at full
power (800 watt) for 1.47 minutes (Sittikaipong, Apichartsrangkoon, & Utama-ang, 2008).
The extract was filtered and evaporated at 60C by vacuum rotary evaporator. The
concentrated solution was freeze- dried and kept at –20oC until used.
The Jiaogulan powders consisted of saponin 95.03 mg/g, gypenoside 29.70 mg/g and
ginsenoside Rb1 0.18 ppm. The Jiaogulan extract consisted of saponin 180.94 mg/g,
gypenoside 122.47 mg/g and ginsenoside Rb1 0.48 ppm.
2.2 Formulation of Jiaogulan tea with lime juice (JTLJ)
The Jiaogulan beverage consisted of 1% Jiaogulan tea, lime juice, sugar, 0.035% ascorbic
acid and 0.01% jiaogulan extract. The experiment used three variables; 1% Jiaogulan tea (50
to 80%), sugar (5 to 30%) and lime juice (5 to 20%). Eight treatments—shown in Table 1—
were generated using D-optimal mixture design (Gacula, 1993),.
Table 1: Eight formulas of JTL created by mixture design.
Formulation 1% Jiaogulan tea (%) Sugar (%) Lime juice (%)
1 85 5 9.95
2 68.32 18.32 13.32
3 85 9.95 5
4 64.95 30 5
5 50 29.95 20
6 74.95 5 20
7 59.16 24.14 16.66
8 64.95 30 5

2.3 Physical and chemical properties

The physical measurement of JTLJ were assessed as follows: colour L*, a*, b* and E
(Minolta, Chroma meter CR-310, Japan) using D65 illuminant, 10 observer. The Chemical
measurement included: total soluble solids (refractometer, PAL-1 Atago, Japan), pH (Hanna
HORIBA: F-22, Japan), total acidity (AOAC 2000), total saponin (Kwon et al., 2003), total
gypenoside (Department of Medical Sciences, 2005) and total gynsenoside Rb1 (Wu, Lin, &
Chau, 2001).
SPISE2009 142

2.4 Sensory descriptive analysis

We evaluated JTLJ using a hybrid descriptive analysis method (Einstein, 1991; Grosso &
Resurreccion, 2002) that was modified from both the Quantitative Descriptive Analysis (QDA,
Tragon Corp., Redwood City, CA, U.S.A.) and the SpectrumTM Analysis methods (Sensory
Spectrum, Inc., Chatham, NJ, U.S.A.). The sensory evaluation was conducted in the sensory
evaluation research unit, Department of product development technology, faculty of agro-
industry, ChiangMai University, and using Su sense program from Silpakorn University,
Thailand. Ten panellists were selected from the students who passed the sensory evaluation
course. The training program was completed over a period of four weeks. Six 2-hour training
sessions were conducted over that period of time. The training program included terminology
development, reference, line scale training (unstructured 15 cm.), using warm-up samples
(Plemmons & Resurreccion, 1998). The trained panellists generated 13 attributes along with
the consensusal references corresponding to each attribute (Table 2). Before evaluating
products in individual booth, the panellists were calibrated using four basic taste reference
standards (Meilgaard et al., 2007) and a warm-up sample for reliability of rating. Each
panellist received 30 ml of sample in 1 oz plastic cup with a cover slip. The samples were
served at room temperature.
Table 2. Attributes of Jiaogulan tea with lime juice created by trained panellists using descriptive
analysis
Attributes Definitions References
Appearance
Yellow colour Light yellow to intense yellow Tratarzine 0.01%
Clearness The degree of clarity of sample Distilled water,
corn starch 0.4%
Aroma
Lime aroma Aromatic associated with artificial lime aroma Artificial lime aroma
(Winner, Thailand)
Jiaogulan aroma Aromatic associated with Jiaogulan aroma Jiaogulan tea
Flavour
Sourness The taste on the tongue associated with aqueous Citric acid solutions
solution of citric acid ( 0.05, 0.08, 0.15%)
Sweetness The taste on the tongue associated with aqueous Sucrose solutions
solution of sugar (2.0, 5.0, 10.0, 16.0%)
Bitterness The taste on the tongue associated with aqueous Caffeine solutions
solution of caffeine (0.05, 0.08, 0.15, 0.20%)
Lime flavour Aromatic in mouth associated with artificial Concentrate lime juice
lime flavour (5%)
Jiaogulan flavour Aromatic in mouth associated with Jiaogulan Jiaogulan tea (1%)
flavour
Aftertaste
Sour aftertaste The intensity of sourness after swallowing the Citric acid solutions
sample ( 0.05, 0.08, 0.15%)
Sweet aftertaste The intensity of sweetness after swallowed the Sucrose solutions
sample (2.0, 5.0, 10.0, 16.0%)
Bitter aftertaste The intensity of bitterness after swallowed the Caffeine solutions
sample (0.05, 0.08, 0.15, 0.20%)
Astringent aftertaste The shrinking of feeling after swallowed the Alum (0.07%)
sample
SPISE2009 143

2.5 Statistical analysis

The mixture design was created by Design-Expert program version 6.0.1.0 (Stat-Ease Inc.,
MN). The response data obtained from the physical and chemical measurements, sensory
descriptive analysis and consumer acceptance test were analysed 1) by analysis of variance
(ANOVA) with a second order model followed by mean difference which was determined by
Tukey HDS, 2) multiple regression analysis, and 3) principal component analysis (PCA),
Statistical analyses were performed with the SPSS® software package (SPSS Inc., Chicago, IL).

3. RESULTS AND DISCUSSION

3.1 Physical and chemical properties
The physical properties of eight JTLJ: colour L*, a*, b*, E, are given in Table 3, along with
their significant differences. Because jiaogulan tea and lime juice were mixed with different
ratios all differences are significant. The chemical properties of the eight JTLJs are given in
Table 4, along with their significant differences. Again all treatments were significantly
different. The soluble solid, total acidity and pH were also significantly different in all
formulations. The active compounds; saponin and gypenosid varied significantly in some
formulations but gensenoside Rb1 showed no significant difference.
Table 3. Means and standard deviations of physical properties of JTLJ: Different letters in the same
column indicate significant differences at p < .05

Formulations L* a* b* E
1 53.72 ± 0.01g 6.87 ± 0.01b 71.50 ± 0.04c 51.10 ± 0.00b
2 60.68 ± 0.03f 4.68 ± 0.01d 68.52 ± 0.04e 47.04 ± 0.01c
3 62.12 ± 0.15e 4.91 ± 0.01c 73.57 ± 0.03a 51.10 ± 0.01c
4 69.81 ± 0.09a 1.88 ± 0.01h 72.54 ± 0.06b 44.10 ± 0.02g
5 64.63 ± 0.01c 3.07 ± 0.01f 63.05 ± 0.02h 44.28 ± 0.00e
6 49.01 ± 0.01h 7.17 ± 0.01a 67.73 ± 0.01f 53.61 ± 0.01a
7 63.03 ± 0.01d 3.95 ± 0.01e 66.91 ± 0.00g 45.69 ± 0.01d
8 68.46 ± 0.03b 2.51 ± 0.03g 70.97 ± 0.03d 44.22 ± 0.03f

Table 4. Chemical properties of JTLJ. Different letters in the same column indicate significant
differences at p < .05
Total Total
Total saponin Total soluble Total acidity
Formulations gypenoside gynsenoside pH
(% w/v) solid (%) (%)
(% w/v) Rb1 (% w/v)
1 12.71 ± 0.11g 2.81 ± 0.10f 6.3 ± 0.85ns 3.04 ± 0.01d 7.17 ± 0.06g 0.79± 0.04d
2 37.87 ± 0.09e 3.36 ± 0.14e 6.57 ± 1.88ns 2.82 ± 0.01f 17.10 ± 0.10e 1.07± 0.11c
3 21.78 ± 0.14f 3.78 ± 0.11d 6.47 ± 0.90ns 3.37 ± 0.01b 11.07 ± 0.06f 0.39± 0.04ef
4 78.20 ± 0.15a 8.03 ± 0.07a 5.53 ± 1.10ns 3.58 ± 0.01a 28.93 ± 0.06c 0.30± 0.04f
5 68.85 ± 0.06c 7.59 ± 0.12b 8.43 ± 1.00ns 2.75 ± 0.02g 29.67 ± 0.58b 1.52± 0.08a
6 12.16 ± 0.13h 2.47 ± 0.16g 6.73 ± 0.81ns 2.91 ± 0.02e 7.00 ± 0.00g 1.47± 0.07a
7 48.30 ± 0.16d 4.13 ± 0.12c 6.60 ± 0.82ns 2.90 ± 0.02e 25.00 ± 0.00d 1.28± 0.04b
8 79.97 ± 7.37 ± 0.08b 7.57± 1.26ns 3.29 ± 30.67 ± 0.58a 0.47± 0.04e
b c
0.11 0.02
SPISE2009 144

3.2 Sensory descriptive analysis

Ten trained panellists evaluated eight formulations of JTLJ with the thirteen attributes: yellow
colour, clearness, lime aroma, Jiaogulan aroma, sourness, sweetness, bitterness, lime flavour,
Jiaogulan flavour, sour aftertaste, sweet aftertaste, bitter aftertaste and astringent aftertaste
followed by perception. The results are shown in Figure 1. The ANOVA showed that
treatments affected all sensory attributes (Table 5).

Yellow
15
Astringent aftertaste Clearness
10
Bitter aftertaste Lime aroma
5

Sweet aftertaste Jiaogulan aroma

0

Sour aftertaste Sourness

Jiaogulan flavor Sweetness

Lime flavor Bitterness

Formula 1 Formula 2 Formala 3 Formula 4

Figure 1. Spider web of descriptive sensory data of JTLJ

Table 5. Sensory descriptive mean rating of eight JTLJformulations. Different letters in the same
column indicate significant differences at p < .05
Form Lime Jiaogulan Sweetnes
ulatio Yellow Clearness Sourness Bitterness
aroma aroma s
1 2.18 ± 0.29d 6.16 ± 6.77 ± 1.71 ± 9.62 ± 2.79 ± 3.58 ±
0.30e 0.44d 0.21e 0.39b 0.33f 0.42a
2 2.40 ±0.24c 6.96 ± 7.83 ± 2.09 ± 8.70 ± 6.88 ± 2.30 ±
0.27d 0.41bc 0.29c 0.45c 0.47d 0.38c
3 1.45 ± 0.25e 5.16 ± 6.21 ± 2.26 ± 6.40 ± 5.47 ± 1.96 ±
0.41f 0.33e 0.33b 0.40d 0.38e 0.28d
4 1.43 ± 0.24e 4.65 ± 5.84 ± 1.89 ± 3.60 ± 11.43 ± 0.89 ±
0.26g 0.37f 0.44d 0.37f 0.38a 0.22f
5 3.60 ± 0.25a 9.21 ± 10.65 ± 2.45 ± 9.58 ± 9.25 ± 3.62 ±
0.52a 0.39a 0.39a 0.43b 0.51b 0.42a
6 2.56 ± 0.29b 7.79 ± 7.60 ± 2.15 ± 12.21 ± 2.13 ± 3.57 ±
0.59b 0.53c 0.30bc 0.65a 0.27g 0.35a
7 2.08 ± 0.28d 7.31 ± 8.05 ± 2.13 ± 9.70 ± 7.37 ± 3.09 ±
0.37c 0.58b 0.30bc 0.35b 0.57c 0.29b
8 1.28 ± 0.25f 3.84 ± 5.93 ± 1.65 ± 4.07 ± 11.28 ± 1.51 ±
0.57h 0.38f 0.26e 0.39e 0.45a 0.32e
SPISE2009 145

Table 5 (continued)
For Aftertaste
Jiaogulan
mula Lime flavour
flavour Sourness Sweetness Bitterness Astringent
tion
1 7.33 ± 0.50d 1.99 ± 0.47d 6.74 ± 0.33c 2.64 ± 0.31g 3.26 ± 0.31b 3.53 ± 0.37a
2 8.24 ± 0.35c 2.33 ± 0.48 c 6.85 ± 0.35bc 4.25 ± 0.57f 1.83 ± 0.15f 1.83 ± 0.15d
3 6.30 ± 0.70e 2.58 ± 0.43b 3.95 ± 0.31e 4.95 ± 0.33e 2.33 ± 0.24e 2.33 ± 0.24c
4 5.10 ± 0.45f 1.93 ± 0.34d 2.22 ± 0.26 g 9.14 ± 0.50a 1.23 ± 0.27g 1.23 ± 0.27e
5 9.30 ± 0.67a 2.93 ± 0.50a 6.32 ± 0.33d 7.26 ± 0.30c 2.75 ± 0.27c 2.75 ± 0.27bc
6 9.00 ± 0.62b 1.76 ± 0.23e 9.80 ± 0.50a 1.79 ± 0.23h 3.65 ± 0.51a 3.65 ± 0.51ab
7 9.31 ± 0.66a 2.43 ± 0.50c 6.93 ± 0.42b 5.42 ± 0.46d 2.52 ± 0.22d 2.52 ± 0.22c
8 4.35 ± 0.74g 1.68 ± 0.22e 2.46 ± 0.29f 8.28 ± 0.30b 1.37 ± 0.18fg 1.37 ± 0.18e

3.3 The relationship between sensory descriptive and physical-chemical properties

PCA) was applied to investigate the relationship between sensory and physical-chemical data.
PCA transforms the original variables into new orthogonal variables called principal
components (PCs). Therefore, PCA expressed the total variation in the data set with only a few
PCs. In PCA, the components are ranked according to their explained variance (Moigne et al.,
2008).

1.0
A3 A1 TAA8
A2 A7
A4 A10
A9 A5 A12
0.5
gypenoside
A13
PC2 (38.28%)

TSS a
saponin
DE
0.0

TS A11 moisture
Rb1
A6
L
-0.5

pH
b
-1.0

-1.0 -0.5 0.0 0.5 1.0

PC1 (50.44%)
Figure 2: Principal component analysis generated by physical-chemical properties and sensory
descriptive attributes A1: Colour A2: Clearness A3: Lime aroma A4: Jiaogulan aroma A5: Sourness
A6: Sweetness A7: Bitterness A8: Lime flavour A9: Jiaogulan flavour A10: Sour aftertaste A11:
Sweet aftertaste A12: Bitter aftertaste A13: Astringent aftertaste TS:Total solid L: Lightness +a:
redness, -a: green +b: yellow, -b: blue DE : Colour difference TSS: Total soluble solid
SPISE2009 146

The results showed that PCA could explain 88.71% of the variance with two components
(Figure 2). PC1 explained 50.44% of the variance and showed that, on the one hand, total
soluble solid was associated to sweet and sweet aftertaste and that, on the other hand, the
moisture, total saponin, ginsenoside Rb1, gypenoside were associated with bitter and
astringent aftertastes. The bitter taste of Jiaogulan comes from its saponin glycoside content
(Cheeke, 2001) and the bitterness perdures as an aftertaste. Peleg et al. (1999) studied the
flavonoid compound in tea by a time-intensity test and found that the maximum bitter
intensity decreased whereas the astringency increased. But Utama-ang (2006) observed that,
when the saponin in Jiaogulan tea increased, so did the bitterness and astringency. PC2
explained 38.28% of the variance and showed that the yellow colour, clearness, lime aroma,
Jiaogulan aroma, lime flavour, Jiaogulan flavour and total acidity were negatively correlated
with pH and b*.

4. CONCLUSION
In this research, our objectives were to investigate the relationship between physical-chemical
properties and sensory descriptive property of JTLJ. The sensory descriptive analysis by
trained panellists defined 13 attributes of JTLJ.; yellow, clearness, lime aroma, Jiaogulan tea
aroma, sourness, sweetness, bitterness, lime flavour, Jiaogulan tea flavour, sourness aftertaste,
sweetness aftertaste, bitterness aftertaste and astringent aftertaste. PCA showed a clear
relationship between sensory attributes and physical-chemical properties with PC1 and PC2
explaining respectively 50.44% and 38.28% of the experimental variance. This shows that
the sensory descriptive method by trained panellists could relate to instrument measurements
for the evaluation of the properties of JTLJ

Acknowledgement. The authors would like to express their thanks to Sensory Evaluation
Research Unit and Department of Product Development Technology, Faculty of Agro-
Industry, ChiangMai University for sensory laboratory facilities.

REFERENCE
AOAC. (2000) Official methods of analysis of AOAC international. 17th ed. Arlington, VA:
Association of Official Analytical Chemists.
Blumert, M. & Liu, J.L. (1999). Jiaogulan China’s ‘‘immortality’’ herb. Badger, USA:
Torchlight Publishing Inc.
Cheeke, P.R. (2001). Glycosides: naturally occurring. Available Source: http://
ls.wiley,com/search/glycoside, November 25, 2005.
Cour, L.B., Molgaard, P., & Yi, Z. (1995). Traditional Chinese medicine in treatment of
hyperlipidaemia. Journal of Ethnopharmacology, 46, 125–129.
Department of Medical Sciences. (2005). Thai Herbal Pharmacopoeia. Nonthburi: Ministry
of Public Health.
SPISE2009 147

Einstein, MA. (1991). Descriptive techniques and their hybridization,. In: H.T Lawless, &
B.P. Klein (Eds). Sensory science theory and applications in foods. New York: Marcel
Dekker Inc, pp. 317-338.
Gacula, M.C. (1993). Design and analysis of sensory optimization. Connecticut: Food and
Nutrition Press, Inc.
Grosso, N.R. & Resurreccion, A.V.A. (2002). Prediction consumer acceptance rating of
cracker-coated and roasted peanuts from descriptive analysis and hexanal measurements.
J. Food Sci., 65, 173-180.
Hu, L., Chen, Z., & Xie, Y. (1996). New triterpenoid saponins from Gynosternma
pentaphyllum. Journal of Natural Products, 59, 1143–1145.
Kwon, J., Balenger, M.R.J., Pare, J.J.R..,& Yaylayan, A.V. (2003). Application of the
microwave-assisted process (MAPTM) to the fast extraction of ginseng saponin. Journal
of Food Research International, 36, 491-498.
Kuwahara, M., Kawanishi, F., Komiya, T., & Oshio, H. (1989). Dammarane saponins of
Gynostemma pentaphyllum Makino and isolation of malonylginsenosides- Rb1, -Rd, and
malonylgypenoside. V. Chem. Pharm. Bull., 37, 135–139.
Meilgaard, M., Civille, G.V., & Carr, B.T. (2007). Sensory Evaluation Techniques. Boca
Ration, Florida: CRC Press.
Moigne, M.L., Maury, C., Bertrand, D., & Jourjon, F. (2008). Sensory and instrumental
characterization of Cabernet Franc grapes according to ripening stages and growing
location. Food Quality and Preference, 19, 220-231.
Peleg, H., Gacon, K., Schlich, P., & Noble, A.C. (1999). Bitterness and astringency of flavon-
3-ol monomers, dimmers and trimers. J. Sci. Food Agri., 79, 1123-1128.
Piacente, S. & Pizza, C. (1995). New Dammarane-type glucoside Gynostemma pentaphyllum.
J. Nat. Prod., 58, 512–519.
Plemmons, L.E. & Resurreccion, A.V.A.. (1998). A warm-up sample improves reliability of
responses in descriptive analysis. J. Sensory Study, 13, 359-376.
Qian, X.H., Wang, Y.X., & Tang, X.L. (1998). Effects of Gynostemma pentaphllum Makino.
Polysaccharide(PGP) on immune function. Journal of China Pharmaceutical University,
30, 51–53.
Rujjanawate, C., Kanjanapothi, D., & Amornlerdpison, D. (2004). The anti-gastric ulcer
effect of Gynostemma pentaphyllum Makino. Phytomedicine, 11, 431–435.
Sittikaipong, N., Apichartsrangkoon, A., & Utama-ang, N. (2008). Saponin extraction
efficiency from Jiaogulan (Gynostemma pentaphyllum) using microwave technology.
KU-CMU Academic Symposium 2008, Takamatsu, Japan.
Takemoto, T., Arihara, S., Nakajiama, T., & Okuhira, M. (1983). Studies on the constituents
of Gynostemma pentaphyllum. I. Structures of gypenoside I–XIV. Yakugaku Zasshi, 103,
173–185.
SPISE2009 148

TISI. (1983). Thai industrial standards institute of green tea Tis 460-2526 (in Thai).
Bangkok: Ministry of Public Health.
Utama-ang, N. (2006). Development of Jiaogulan tea (Gynostemma pentaphyllum). Ph.D
Thesis, Graduate school, Kasetsart University, Thailand
Wu, J., Lin, L,.& Chau, F. (2001). Ultrasound-assisted extraction of ginseng saponin from
ginseng roots and cultured ginseng cells. Journal of Ultrasonics Sonochemistry, 8, 347-
352.
Zhou, H.H., Liang, J.X.& Hu, W.H. (2001). An experimental study of the antitumor effect of
Jiaogulan and Oolong tea. Journal of Xiamen University, 40, 990–993.






























BK
TP HCM