Food Quality and Preference 71 (2019) 279–285

Contents lists available at ScienceDirect

Food Quality and Preference

journal homepage: www.elsevier.com/locate/foodqual

Short Communication

Utilizing hedonic frame for projective mapping: A case study with Korean T
fermented soybean paste soup
⁎
Mi-Ran Kima, Kwang-Pyo Kimb, Seo-Jin Chunga,
a
Department of Nutritional Science & Food Management, Ewha Womans University, 52, Ewhayeodae-gil, Seodaemun-Gu, Seoul 120-750, Republic of Korea
b
Department of Food Science & Technology, Chonbuk National University, Republic of Korea

A R T I C LE I N FO A B S T R A C T

Keywords: Projective mapping (PM) based techniques are frequently used to develop consumer perception maps holistically
Hedonic-frame for identifying and characterizing samples with similar characteristics. In the present study, the criteria for
Projective mapping locating samples in projective mapping are narrowed from the original projective mapping methodology. This
Consumer acceptance study proposes the use of a hedonic frame (i.e. reasons for liking similarity), H_PM, and comparing it using a
Korean fermented soybean paste
sensory frame (i.e. sensory similarity), S_PM, with the aim of understanding how consumers perceive soups made
with various Korean fermented soybean pastes. The participants comprised a total of 69 consumers. Fifteen
fermented soybean paste products from different regions of Korea were selected. All consumers evaluated
samples using both S_PM and H_PM, which were conducted in separate sessions. The order of the two mapping
sessions was balanced between the subjects. In the S_PM method, subjects grouped samples with similar sensory
characteristics. In the H_PM method, subjects grouped samples which had similar reasons for liking or disliking
on a mapping sheet. Ultra flash profiling was conducted in both S_PM and H_PM after the mapping tasks.
Multiple factor analysis was used for statistical analysis. S_PM and H_PM resulted in different product positions.
Although some samples shared very similar sensory characteristics with each other in S_PM, distinct differences
appeared in the reasons for (dis)liking in H_PM. Critical attributes that affected sample positioning differed when
using different criteria for mapping the samples which resulted in discrete perceptual maps of S_PM and H_PM.
H_PM can identify important hedonic drivers of samples that may not be caught by a sensory based approach.

1. Introduction samples on a sheet.

Free sorting and projective mapping (a.k.a. napping) related tech-
niques are popular methods for product characterization due to their
simplicity as well as effectiveness in holistically discriminating samples
(Risvik, McEwan, & Rødbotten, 1997; Tuorila & Monteleone, 2009;
Cadoret & Lê, 2010). These methods basically identify and characterize
samples with (dis)similar characteristics (Faye et al., 2004; Abdi,
Valentin, Chollet, & Chrea, 2007; Lelièvre, Chollet, Abdi, & Valentin,
2008). In global projective mapping techniques, subjects use holistic
similarity criteria, which include a broad spectrum of similarity in
sensory, hedonic or other aspects of food, to position samples on a
sheet. More recently, researchers conducted projective mapping using a
narrower criteria (i.e. partial projective mapping), specific reference
sample (i.e. polarized projective mapping), or specific attribute (i.e.
freshness). Table 1 lists various projective mapping/napping techniques
which utilize different evaluation frame of similarity for positioning the
Projective mapping has often been proposed as an alternative quick
method for descriptive analysis (Valentin, Chollet, Lelièvre, & Abdi,
2012). Previous studies have investigated the reproducibility, validity,
and reliability of projective mapping techniques by comparing methods
with descriptive analysis (Kennedy & Heymann, 2009; Heymann,
Hopfer & Bershaw, 2014). The effectiveness of the projective mapping/
napping methodology as a sensory analysis tool has also been compared
with rapid descriptive analysis methods such as free multiple sorting,
flash profiling, ultraflash profiling, conventional profiling, and the
CATA method (Perrin et al., 2008; Nestrud & Lawless, 2010; Dehlholm,
Brockhoff, Meinert, Aaslyng, & Bredie, 2012; Santos et al., 2013; Liu,
Grønbeck, Di Monaco, Giacalone, & Bredie, 2016). The usage of pro-
jective mapping as an alternative for descriptive analysis implies that
the sensory based similarity between samples are one of the main im-
plicit frames adapted to determine the sample position on a map, al-
though subjects evaluate the samples holistically.

⁎
Corresponding author at: Department of Nutritional Science & Food Management, College of Science & Industry Convergence, Ewha Womans University, 52,
Ewhayeodae-gil, Seodaemun-Gu, Seoul 03760, Republic of Korea.
E-mail address: sc79d@ewha.ac.kr (S.-J. Chung).

https://doi.org/10.1016/j.foodqual.2018.07.014
Received 12 January 2018; Received in revised form 31 July 2018; Accepted 31 July 2018
Available online 01 August 2018
0950-3293/ © 2018 Elsevier Ltd. All rights reserved.
M.-R. Kim et al. Food Quality and Preference 71 (2019) 279–285

Table 1
List of projective mapping/napping techniques utilizing diverse evaluation frame for positioning the samples.
Type of method Evaluation Frame Publications

Global Projective Mapping/Napping Holistic_ Similarity/dissimilarity Risvik et al. (1994)

Projective Mapping
Projective Mapping/Napping with Sorting
Partial Mapping/napping
Partial Mapping/napping with Sorting
Polarized projective mapping
Projective Mapping on Choice/Preference
Structured/directed mapping
Risvik et al. (1997)
Pagès (2005)
Perrin et al. (2008)
Dehlholm et al. (2012)
Kim et al. (2013)
Vidal, Cadena, Antunez, Gimenez, and Varela, Ares
(2014)
Marcano et al. (2015)
Varela et al. (2017)
Moelich et al. (2017)
Esmerino et al. (2017)
Holistic_Similarity/dissimilarity Pagès, Cadoret, and Le (2010)
Hopfer and Heymann (2013)
- Single attribute (Usage, Freshness)_similarity King et al. (1998)
- Single modality (appearance, aroma, taste, or texture)_Similarity Zhang, Lusk, Mirosa, and Oey (2016)
- Multi modalities (aroma & taste, palate or flavor & texture) _Similarity/ Dehlholm et al. (2012)
dissimilarity Reinbach, Giacalone, Ribeiro, Bredie, and Frøst
(2014)
Marcano et al. (2015)
Louw et al. (2015)
Moelich et al. (2017)
- Single modality (appearance, aroma, taste, or texture)_Similarity Blancher, Clavier, Egoroff, Duineveld, and Parcon
- Multi modalities (aroma & taste, palate: flavor & texture) _Similarity/ (2012)
dissimilarity
3 Reference samples_ Similarity/dissimilarity Ares et al. (2013)
Horita et al. (2017)
Preference_Similarity/dissimilarity Lezaeta et al. (2017)
Varela et al. (2017)
Usage & liking King et al. (1998)
Healthy & Liking Varela and Salvador (2014)

When evaluating samples for projective mapping, subjects will use
their own holistic criteria, and as mentioned earlier, the types of criteria
involved and the weight of these criteria (if there are more than 2) will
vary among subjects but will probably be consistent within a subject.
For example, foods that taste similar to each other should elicit similar
hedonic responses. However, since the components that build similarity
perception of food are multifaceted, similar sample properties may not
necessarily mean similar sensory property nor similar hedonic proper-
ties of food. That is, the importance (or weight) of sample attributes
affecting the perceived similarity of samples from sensory aspects and
hedonic aspects may well be different. Thus, similar sensory char-
acteristics of samples will not always correspond to similar acceptance
ratings. And partly due to these reasons, researchers attempt to narrow
the frame for evaluating similarity among samples.
A few researchers have attempted to incorporate hedonic aspects of
food evaluation into mapping/napping techniques. King, Cliff, and Hall
(1998) proposed a structured projective mapping method that defined
the x and y axes as liking and usage, respectively, and asked the subjects
to position the samples in this two-dimensional space based on their
individual liking and usage. Varela and Salvador (2014) applied a si-
milar approach to children. Very recently, projective mapping based on
preference was introduced and compared with classical projective
mapping resulting in a different perceptual map of the product (Varela
et al., 2017). The present study proposes a projective mapping method
which utilizes hedonic frame (H_PM), and compared this with a pro-
jective mapping utilizing a sensory based similarity strategy (S_PM)
using soups made with fermented soybean paste samples as products of
interest. S_PM was conducted using previously described protocols
(Risvik, McEwan, Colwill, Rogers, & Lyon, 1994; Pagès, 2005; Park, Lê,
Hong, & Kim, 2014). For the H_PM method, consumers were required to
sort the samples based on similarity of reasons for liking and disliking
the samples.
Fermented soybean paste soup was selected as a product category of
interest in this experiment since Korean fermented soybean paste
(called doenjang) is a very important culinary ingredient in Korean
foods and the flavors of fermented soybean pastes can vary widely
depending on the raw materials, fermentation methods, aging periods,
and other factors (Kim & Rhee, 1988; Lee, 2004; Ahn et al., 2012; Jeon,
Lee, Kim, & Kim, 2016).
2. Material and methods
2.1. Sample and sample preparation
Fifteen Korean fermented soybean paste (doenjang) products were
chosen as samples of interest (Table 2). A large set of samples was
obtained from major and local producers located in various regions of
Korea. Two products were national brands with the largest market
shares in Republic of Korea. Nine products were produced by various
small-scale local producers in five different provinces. Four products
were obtained from a single local producer and these were produced by
similar methods but used different starter cultures to ferment the soy-
bean paste.
Since fermented soybean paste is a seasoning ingredient that is
rarely consumed on its own, samples were evaluated as soup prepara-
tions (Kim, Hong, Song, Shin, & Kim, 2010; Kim & Lee, 2014). In the
first step, stock for fermented soybean paste soup was made by boiling
160 g of dried anchovy and dried kelp at a ratio of 3:7 (Seokha, Busan,
Korea) in 4 L of water (Jeju Sam Da Soo, Kwangdong Pharmaceutical,
Jeju, Korea) for 40 min. The salinity of the stock was approximately
0.09%. The fermented soybean paste sample was added to the stock at a
concentration of 10%. The soup was then boiled for 5 min, and soup
samples were kept in a heating cabinet (LH-1041G, Daeyeong E&B,
Changwon, Korea) at 70 °C until being evaluated. Fifty milliliters of
each sample was poured into a disposable cup (diameter 7 cm, height
4 cm; Samboopack, Incheon, Korea) just before the sample evaluation,
and each sample was labeled with a three-digit random code.

280
M.-R. Kim et al. Food Quality and Preference 71 (2019) 279–285

Table 2
Sample information, consumer acceptability scores and rank (between brackets) of 15 fermented soybean pastes.
Sample identification Province Ingredient Overall acceptability

JS Jeolla-do Soy bean, Sea salt 4.0bcd(11)

JN Soy bean, Sea salt, Purified water 5.0ghi(3)
KS Gyeongsang-do Meju/Soy bean, Sea salt 5.5i(1)
KN Meju, Sea salt, Purified water 3.1a(15)
CS Chungcheong-do Soy bean, Sea salt 4.6efgh(7)
CN Meju, Salt Purified water 3.8bc(12)
KW_1 Kangwon-do Meju, Sea salt 4.8fgh(5)
KW_2 Organic soy bean, Sea salt 4.1cde(10)
JJ Jeju-do Soy bean, Refined salt 3.2a(14)
MJ_1 Major company Soy bean, Purified water, Refined salt 4.7fgh(6)
MJ_2 Soy bean, Purified water, Sea salt, Soybean flour, Seed malt, ethanol 5.2hi(2)
R Jeolla-do Meju (Rice straw based starter culture1), Sea salt, Purified water, ethanol 3.4ab(13)
RA Meju (Rice straw based starter culture1 + A. oryzae), Sea salt, Purified water, ethanol 4.9fgh(4)
BA Meju (B. licheniformis + A. oryzae), Sea salt, Purified water, ethanol 4.4defg(8)
RBA Meju (Rice straw based starter culture1 + B. licheniformis + A. oryzae), Sea salt, Purified water, ethanol 4.3cdef(9)

Different superscripts within a column indicate significant difference according to Duncan’s multiple range test (p < 0.05). The samples were evaluated on a 9-point
hedonic scale (1: Dislike extremely, 9: Like extremely).

2.2. Subjects
Seventy-eight female consumer subjects consisting of young female
college students (n = 40, aged 20–29) as well as housewives (n = 38,
aged 40–59) actively involved in cooking were recruited from local
community. Subjects who were interested in fermented soybean paste
soup tasting participated in the study. A total of sixty-nine subjects
completed both the S_PM and H_PM experiments.
2.3. Procedure
Fifteen fermented soybean paste samples were evaluated using both
the S_PM and H_PM methods. Half of the subjects evaluated the samples
using S_PM initially and then re-evaluated the samples using H_PM
one week later. The other half of the subjects evaluated the samples first
using H_PM followed by S_PM one week later. A detailed description of
the tasting and evaluation procedure for each mapping method along
with a handout describing the procedure were provided to the subjects
at the start of each session. Preliminary experiments were carried out to
find an optimal protocol for maintaining the correct temperature
(70 °C) and to minimize variations between the panelists in sample
handling. Unlike most PM studies, in which samples are served all at the
same time, samples were served in sequential monadic manner to
control the temperature of the samples during evaluation since tem-
perature drastically affects the flavors of soups. Additionally, partici-
pants commented on the need for writing down distinctive notes of
each sample since there were too many samples to remember prior to
positioning them on the sheet. Thus, in the main experiment, subjects
were allowed to freely write memos of each sample in both projective
mapping conditions (Fleming, Ziegler, & Hayes, 2015). The serving
order of the 15 samples was determined by a Williams Latin Square
design (Williams, 1949). Panelists were allowed to retaste the samples
during the projective mapping process. Upon request from the subjects,
warm samples were served again for re-evaluation when the samples
became cold. Warm water and unsalted water crackers (Carr’s Original
Table Water, Carr’s of Carlisle, Carlisle, UK) were provided to cleanse
the palate between samples.
2.3.1. Sensory based projective mapping (S_PM)
The following procedure was implemented for the S_PM evaluations
of 15 fermented soybean paste soup samples. Consumers tasted each
sample and wrote descriptions of the perceived sensory characteristics.
Upon request, a list of lexicons often used to describe the sensory
characteristics of fermented soybean paste soup were provided to help
the subjects to describe samples using appropriate attributes. The sub-
ject positioned each sample on a rectangular sheet of size 30 cm by
20 cm based on its description, such that samples that elicited more
similar sensory properties were located closer to each other, while those
showing different sensory profiles were farther away from each other.
The subjects were allowed to retaste the samples to confirm or change
their positions. Once the subjects had finalized the positions of all of the
samples, the three-digit code of each sample was marked on the sheet.
The subjects were then asked to group the samples into as many groups
as they wished based on the similarity of their sensory profiles. The
number of groups ranged from 2 to 14, and each sample could only
belong to a single group. After grouping the samples, the subjects
conducted ultra-flash profiling (Perrin & Pagès, 2009) freely describing
the sensory characteristics of the groups they had classified.
2.3.2. Hedonic based projective mapping (H_PM)
For the H_PM method, subjects were first asked to rate the overall
liking (OL) of each sample on a 9-point hedonic scale and the reasons
for liking and disliking its taste. Similar to the S_PM method, a list of
sensory and non-sensory terms were given to the subjects during sample
evaluation upon request to help them describe the reasons for liking
and disliking accurately. The subjects were then asked to position the
samples on a rectangular sheet of the same size as that used in S_PM
based on the similarity and dissimilarity of the liking level(score) and
the reasons for liking and disliking the samples; that is, samples rated
with similar liking scores due to similar reasons for (dis)liking were
positioned close to each other, while samples with similar acceptance
ratings were separated when the reasons for (dis)liking differed.
Additionally, samples were separated when their acceptance ratings
were different but the reasons for (dis)liking them were similar.
Subjects were asked to retaste each sample before finalizing its position
on the sheet. Subjects were allowed to retaste the samples as often as
they wanted to. Again, the three-digit code of each sample was marked
on the sheet when its position had been finalized, and the subjects then
categorized the samples into 2–14 groups based on the similarity of the
hedonic level and the reasons for (dis)liking them, with each belonging
to a single group. After grouping the samples, the subjects freely de-
scribed the reasons for (dis)liking the groups they had classified.
2.4. Statistical analysis
For both the S_PM and H_PM methods, multiple factor analysis
(MFA) was applied to the projective mapping data that each panelist
had generated to extract data for and produce the common perception
map of the products and the corresponding distinctive attributes per-
ceived by the subjects (Pagès, 2005; Kim, Jombart, Valentin, & Kim,
2013). In the case of descriptors obtained from the UFP, similar words
were first merged together (e.g. sour and tart were merged to sour). The

281
M.-R. Kim et al.
Fig. 1. MFA plot of 15 fermented soybean paste samples (A) and the corresponding sensory descriptors used as supplementary variables (B) evaluated by S_PM
method.
merged attributes were reviewed by two other sensory researchers for
validation. Then the frequencies of attributes mentioned per sample
were counted. In order to understand the common descriptors used to
characterize each sample group, an attribute was included as a sup-
plementary variable in the MFA if that attribute was elicited by more
than 10% of the subjects for at least one sample.
Analysis of variance (ANOVA) was conducted to analyze the effects
of fermented soybean paste type and age group on the overall accep-
tance of the samples. Subject was considered as a random factor.
Additionally, two types of cluster analysis, hierarchical cluster analysis
using Ward’s method and K-means cluster analysis, were conducted on
the overall liking scores of the 15samples to investigate the presence of
heterogeneity in terms of consumers’ preference patterns for the sam-
ples. MFA was conducted using the FactoMineR package (Husson,
Josse, Lê, & Mazet, 2011) of the R statistical system, ANOVA was im-
plemented using SPSS (version 23, IBM, Chicago, IL, USA), and Cluster
analysis was carried out by XLSTAT (Paris, France). When the effect of
age group on the preference and the perceptual configuration (S_PM
and H_PM) of samples were examined by ANOVA and MFA, the young
and old subject groups showed little differences in sample perception.
Thus the data from the two groups were merged for analysis and results
from group as a whole are presented in the results and discussion sec-
tion.
3. Results
3.1. Descriptors generated by S_PM and H_PM to profile fermented soybean
paste samples
Applying S_PM to the 15 fermented soybean paste samples resulted
in ninety-three attributes to characterize the samples. Sensory terms
with similar meanings were merged into 47 attributes, and ultimately
16 attributes were used as supplementary variables for MFA. Hedonic
or emotional terms were not mentioned by the subjects during the
profiling. Saltiness (25.4%), acetic acid odor/flavor (18.4%), brown
(18.9%), meju (fermented soy) odor/flavor (14.4%), meaty odor/flavor
(13.7%), and sweetness (10.8%) were the most frequently mentioned
attributes.
Concerning H_PM method, the subjects originally used 58 and 87
attributes to describe the reasons for liking and disliking the samples,
respectively. Combining the attributes with similar meanings resulted
in 24 liking and 28 disliking attributes. At final analysis, 16 attributes
(7 liking and 9 disliking) were used as supplementary variables to
282
Food Quality and Preference 71 (2019) 279–285
understand the perceptual map of the samples. The most frequently
mentioned reasons of liking attributes were meaty odor/flavor (15.2%)
and mild odor/flavor (11.8%) while the most frequently mentioned
disliking reasons were saltiness (27.2%), nauseating (14.6%), lack of
flavor (13.5%), and sourness (13.2%) detected from the samples. The
subjects used similar number of words, 27.7 and 32.3 on the average, to
describe all the samples using ultra flash profile in S_PM and H_PM,
respectively.
3.2. Overall liking of fermented soybean paste samples
The mean overall liking (OL) ratings and preference rank order of
the samples are presented in Table 2. The OL rating varied significantly
with the fermented soybean paste sample type (F = 13.708,
p < 0.001). Samples KS and MJ_2 were commonly preferred, while
samples KN, JJ, and R were on average the least preferred samples.
Scale usage between the two age groups significantly differed
(F = 15.602, p < 0.001) while the overall preference order of the
samples was similar in the young and older subjects (F = 1.493,
p = 0.107). When cluster analysis was conducted on the liking scores of
the samples, both hierarchical and K-means commonly resulted in 2
clusters consisting of approximately 70% of subjects in cluster 1 and
30% in cluster 2. This observation implies that the majority of the
subjects shared similar preference patterns for the samples.
3.3. Sensory profile of fermented soybean paste by S_PM
MFA maps of the 15 samples profiled by the S_PM method and the
corresponding sensory descriptors used as supplementary variables are
shown in Fig. 1A and B, respectively. Factor 1 (F1) and factor 2 (F2)
explained 17.1% and 11.5% of the total variance, respectively. When
perceived sensory characteristics were analyzed as supplementary
variables in the MFA, the darkness of color (brown vs. yellow) was the
main driver characterizing the samples in F1. A brown color itself
showed strong positive correlations with saltiness, meju odor/flavor
and soy sauce odor/flavor attributes, while a yellow color showed a
strong correlation with a roasted soybean odor. The negative axis of F2
showed high loadings with sour-related attributes (sourness, acetic acid
odor), while the positive F2 axis showed high loadings with sweet,
umami taste, and meaty odor/flavor, which are reported to be distinct
characteristics of commercially manufactured fermented soybean paste
(Chung & Chung, 2007).
Concerning the sample map (Fig. 1A), samples were roughly
M.-R. Kim et al.
clustered into 4 distinct groups. Samples KN, KW_2, CS, and JJ were
located on the positive F1 axis and negative F2 axis. These samples
were distinctly characterized as salty [average of KN, KW_2, CS, & JJ
39.6% vs. average of all samples 25.4%] and brown [24.0% vs. 18.9%],
and marginally as soy sauce [8.7% vs. 5.6%], and sour [12.2% vs.
9.9%]. Samples CN, JN, KS, and KW_1 were perceived similarly to each
other and these were loaded on the positive F1 and F2 axis largely due
to their brown color [average of CN, JN, KS, & KW_1 26.0% vs. average
of all samples 18.9%] and marginally due to saltiness [28.1% vs.
25.4%]. Samples MJ_1 and MJ_2 (national-brand products) were posi-
tioned at positive F2. These samples were described as eliciting sweet
[average of MJ_1 & MJ_2 22.9% vs. average of all samples 10.8%],
umami tastes [16.0% vs. 9.3%], and meaty odor/flavor [27.1% vs.
13.7%] and were perceived as being very similar. Samples RA, RBA,
BA, and R were located on the negative F1 and F2 axis and were
characterized noticeably by yellowness [average of RA, RBA, BA, & R
19.1% vs. average of all samples 8.1%], and weakly by soybean odor
[5.9% vs. 3.0%], and sourness [11.8% vs. 9.9%].
3.4. Hedonic based sensory profile of fermented soybean paste by H_PM
In the MFA, 28.3% of the total variance was explained by F1
(15.4%) and F2 (12.9%) when the samples were profiled by H_PM
method (Fig. 2A & B). Overall, the positive F1 axis was defined by the
reasons of liking attributes while the negative F1 axis was defined by
the reasons of disliking attributes. Liking familiar, rich, meaty, ripened
flavor and umami taste of samples were strongly loaded on the positive
F1 axis whereas disliking fermentation flavor, nauseating, strange
flavor were strongly loaded on the negative F1 axis. F2 axis was
strongly related to attribute strength. That is, liking mild flavor and
disliking lack of flavor were positively loaded on F2, while disliking
saltiness, bitterness, and color were negatively loaded.
Unlike the distinct clustering of samples observed in S_PM, the po-
sition of the samples were scattered more in the diagonal direction of
negative F1 and F2 to positive F1 and F2 in H_PM. The samples were
aligned based on their acceptance levels of the samples. The lowest-
rated sample was located on the negative F1 and F2 axes. Starting from
these axes, the acceptance rating increased in the diagonal direction
toward the positive F1 and F2 axes, where the highest-rated sample was
located. Additionally, 2 clusters of samples were located on the right
and left side of the diagonal. The samples on the right side of diagonal
were characterized with reasons of liking attributes while samples on
the left were delineated using reasons of disliking terms.
Fig. 2. MFA plot of 15 fermented soybean paste samples (A) and the corresponding sensory descriptors used as supplementary variables (B) evaluated by H_PM
method. Average consumer acceptability scores of 15 types of fermented soybean paste samples are shown in brackets.
283
Food Quality and Preference 71 (2019) 279–285
Samples KS (OL = 5.5) and MJ_2 (OL = 5.2) were ranked as the
most acceptable samples, and they were commonly liked because of
their mild [average of KS & MJ_2 13.3% vs. average of all samples
7.4%], salty [13.3% vs. 7.4%], meaty flavor [16.7% vs. 11.7%], and
umami taste [8.7% vs. 4.9%]. Rich [24.0% vs. 9.4%], ripened [17.3%
vs. 6.3%], and familiar flavors [14.7% vs. 8.1%] were additional liking
reasons of KS. Thus KS came closer with MJ_2 in H_PM than in S_PM
which showed only marginal ellipses overlap between the two samples.
Regarding samples MJ_1 and MJ_2, which were perceptually very close
in S_PM, the distance of the two samples increased although the ellipses
were slightly overlapping when evaluated in the hedonic contextual
framework. For lower rating samples, different attributes drove the
disliking of different samples. Samples KN (OL = 3.1), CN (OL = 3.8),
and KW_2 (OL = 4.1) which were positioned on the negative F2 axis,
were clearly disliked due to their strong saltiness [average of KN, CN, &
KW_2 40.9% vs. average of all samples 27.2%], and tendencies of eli-
citing slightly more sourness [14.7% vs. 13.2%], and strange flavor
[11.6% vs. 9.3%] than other samples. Bitterness [KN 10.7% vs. average
of all samples 3.6%] and dark color [18.7% vs. 6.1%] detected in
sample KN were additional reasons for subjects disliking this sample.
Samples JJ (OL = 3.2), R (OL = 3.4) and JS (OL = 4.0) were disliked
rather due to lack of flavor [17.8% vs. 13.5%], fermented odor/flavor
[11.6% vs. 7.5%], and strange-flavor characteristics [13.3% vs. 9.3%],
which were also associated with emotional terms such as “nauseating
[29.8% vs. 14.6%]”. Close to these samples, samples BA (OL = 4.4), RA
(OL = 4.9), and RBA (OL = 4.3) were commonly positioned together
because they were all disliked for lacking flavor [average of BA, RA, &
RBA 21.8% vs. average of all samples 13.5%] but were liked due to
having a mild flavor [15.6% vs. 11.7%].
4. Discussion
This study sorted and positioned 15 fermented soybean paste sam-
ples on mapping sheets by applying two types of criteria (sensory vs.
hedonic), with these criteria used by consumers to profile their char-
acteristics. In most of the previous projective mapping studies, subjects
were instructed to locate samples based on the samples’ holistic simi-
larities. While in most cases the subjects did sort the samples based on
the similarities of sensory properties, in some instances hedonic or
quality dimensions were identified in the results of sorted napping or
projective mapping (Ares, Varela, Rado, & Giménez, 2011; Torri et al.,
2013). Kim et al. (2013) observed that consumers were able to sort
samples based on sensory attributes related to the processing method
M.-R. Kim et al.
when they were familiar with the samples. However, when the products
were unfamiliar, consumers had difficulty detecting subtle differences
between the samples, and the hedonic aspects of the samples influenced
their napping positions.
The present study has proposed using similar liking levels(scores)
and reasons of liking as criteria for positioning samples in the projective
mapping procedure. Unlike global projective mapping, the present
H_PM method asks the subjects to locate the samples based on specific
criteria and implicitly urges subjects to become somewhat analytical
rather than holistic. Applying two different strategies to locate samples
on mapping sheet resulted in discrete product maps. There were several
occasions where the relative distances between the samples were in-
consistent between the S_PM and H_PM methods. Samples MJ_1 and
MJ_2, which were perceptually very close in S_PM, moved apart and
ellipses were only slightly overlapping when evaluated in the hedonic
contextual framework. Subjects gave higher acceptance ratings for
sample MJ_2 because it had fewer off-odors/off-flavors (MJ_2 4.0% vs.
MJ_1 12.0%), as well as an artificial flavor (MJ_2 0.0% vs. MJ_1 6.7%)
than MJ_1. These two samples were difficult to differentiate in the
sensory based mapping, but based on liking levels and reasons of liking,
they were perceived as being different when these traits were captured
using the H_PM method. A similar example is samples CN and JN: in the
S_PM map, sample JN (which was liked by consumers) was situated
very close to sample CN (which was relatively disliked). The presence
or absence of a meaty odor/flavor was an important factor determining
the hedonic levels of samples in the case of samples JN (presence) and
CN (absence). Similarly, samples KN, KW_2, CS were characterized as
salty, traditional soy sauce, and brown color, and perceived rather si-
milarly in terms of their sensory properties. However, the liking of these
samples varied quite widely. Varela et al. (2017) also observed similar
discrepancies between sample mapping by global similarity and global
choice/preference. These disagreements in the product configurations
between S_PM and H_PM are probably related to differences in the re-
lative importance given to certain sensory attribute when the task is
based on sensory vs. hedonic criteria. For example, samples may be
liked/disliked for having more/less intensity of certain sensory attri-
butes, which may not be responsible for the main differences/simila-
rities in terms of sensory attributes.
When profiling samples utilizing UFP method, different frameworks
applied during the evaluation had led to different descriptions for the
samples. That is, when subjects were inquired to describe sample
characteristics in sensory based framework they were not able to suf-
ficiently express their perceptions using sensory terms—emotional he-
donic terms were somewhat more useful. In the case of samples, R, JS,
BA, RBA, which were disliked for their lack of flavor, very few de-
scriptors were used to characterize these samples other than yellowness
and chemical odor in S_PM. When they were evaluated from the per-
spective of reasons for (dis)liking, the samples were described as having
an unripened and fermented flavor and off-flavors, and were frequently
associated with emotional terms such as “nauseating” and “disgusting.”
The main advantage of H_PM is that the samples can be roughly
grouped based on similar hedonic properties that are important drivers
influencing the acceptance of the samples. H_PM will be especially
useful when researchers try to screen and select samples from a large
pool of candidates to conduct further consumer studies within a target
product category. Projective mapping techniques have been proposed
as screening tools to select samples from a large pool of candidates
(Louw et al., 2013; Moelich, Muller, Joubert, Næs, & Kidd, 2017). We
suggest H_PM as a potential screening tool to select samples for a
consumer acceptance test.
The basis of projective mapping is to have subjects evaluate samples
under holistic mindset and intuitively position samples in such a way
that cannot be verbally or rationally explained. However in many
global projective mapping, researchers step further and ask subjects to
characterize the samples often in the form of ultraflash profiling
(Marcano, Ares, & Fiszman, 2015; Varela et al., 2017; Moelich et al.,
284
Food Quality and Preference 71 (2019) 279–285
2017; Esmerino et al., 2017). Thus subjects first evaluate samples hol-
istically but eventually have to characterize the samples verbally. Ad-
ditionally holistic nature of PM becomes somewhat analytical under the
context of partial PM since specific criteria are given for positioning the
samples (refer to Table 1). In the present study, subjects were rather
urged to become analytical mind set in both S_PM and H_PM. Particu-
larly in H_PM, subjects were instructed to position the samples on very
specific criteria, reasons of liking and disliking. In addition, the num-
bers of samples were relatively large and from a practical point of view,
it was difficult to locate 15 samples holistically by remembering all
their characteristics. It has been reported that individual’s cognitive
style, wholistic vs. analytic, significantly affected sample positioning in
PM (Varela et al., 2017). The performance of projective mapping, when
subjects explicitly utilize holistic or analytical frame as a psychological
strategy for sample positioning, especially in the case of large sample
number should be investigated further in future research.
For both S_PM and H_PM methods, the total variance explained by
factor1 and 2 in the MFA was relatively low, being approximately
28–29%. Factor 3 and 4 did not provide any additional insightful in-
formation to understand the sample configuration thus these dimen-
sions were not mentioned further. Total explained variance of dimen-
sion 1 and 2 for multivariate analysis such as MFA and GPA are often
critically dependent on the number of samples, latent structure of
sample set, and subjects’ level of training. Usually the larger the number
of samples, the lower the % of cumulative explained variance. The total
variance of dimension 1 and 2 in previous projective mapping or sorting
studies were approximately 38–56% when 7–9 samples were evaluated
(Lelièvre et al., 2008; Lezaeta, Bordeu, Næs, & Varela, 2017). The total
variance, however, dropped drastically to 21–31% when the sample
number increased to 12–18 (Hopfer & Heymann, 2013; Torri et al.,
2013; Park et al., 2014). These are similar % level with the present
study which evaluated 15 samples. Moreover, as reported previously for
other projective mapping (or napping) studies, some subjects find the
task difficult and may require training in order to get used to the pro-
cedure prior to participating in the actual experiments (Ares, Deliza,
Barreiro, Giménez, & Gámbaro, 2010; Liu et al., 2016). Thus familiar-
izing subjects to the procedures of S_PM or H_PM may increase the %
variance explained by each dimension.
5. Conclusions
This study has proposed a projective mapping method that accounts
for the hedonic aspects of product evaluation. The results showed that
the subjects used different criteria to sort samples in the S_PM and
H_PM methods. The inconsistent results between S_PM and H_PM imply
that projective mapping utilizing sensory properties of samples as the
main criteria for locating samples may overlook the hedonic drivers
that significantly influence the liking of samples in the mapping task.
The H_PM method seems to be appropriate when it is used to screen
samples for further analysis in consumer taste testing. Nevertheless,
asking subjects to evaluate a large number of samples by either S_PM or
H_PM resulted in low % of total variance explained by the main two
factors using MFA. Since the projective mapping task by itself may be
more challenging for some subjects, initial familiarization to the eva-
luation procedure itself may be necessary in order to reduce potential
errors occurring due to the complicated nature of the task.
Acknowledgements
This research was supported by Sunchang Jangryu Co. We deeply
thank the anonymous reviewer for the valuable comments.
References
Abdi, H., Valentin, D., Chollet, S., & Chrea, C. (2007). Analyzing assessors and products in
sorting tasks: DISTATIS, theory and applications. Food Quality and Preference, 18,
M.-R. Kim et al.
627–640.
Ahn, J. B., Park, J. A., Jo, H. J., Woo, I. H., Lee, S. H., & Jang, K. I. (2012). Quality
characteristics and antioxidant activity of commercial fermented soybean paste and
traditional Fermented soybean paste in Korea. The Korean Journal of Food and
Nutrition, 25, 142–148.
Ares, G., Deliza, R., Barreiro, C., Giménez, A., & Gámbaro, A. (2010). Comparison of two
sensory profiling techniques based on consumer perception. Food Quality and
Preference, 21, 417–426.
Ares, G., de Saldamando, L., Vidal, L., Antunez, L., Gimenez, A., & Varela, P. (2013).
Polarized projective mapping: Comparison with polarized sensory positioning ap-
proaches. Food Quality and Preference, 28, 510–518.
Ares, G., Varela, P., Rado, G., & Giménez, A. (2011). Identifying ideal products using
three different consumer profiling methodologies. Comparison with external pre-
ference mapping. Food Quality and Preference, 22, 581–591.
Blancher, G., Clavier, B., Egoroff, C., Duineveld, K., & Parcon, J. (2012). A method to
investigate the stability of a sorting map. Food Quality and Preference, 23, 36–43.
Cadoret, M., & Lê, S. (2010). The sorted napping: A new holistic approach in sensory
evaluation. Journal of Sensory Studies, 25, 637–658.
Chung, L., & Chung, S. J. (2007). Cross-cultural comparisons among the sensory char-
acteristics of fermented soybean using Korean and Japanese descriptive analysis
panel. Journal of Food Science. 72, S676–S688.
Dehlholm, C., Brockhoff, P. B., Meinert, L., Aaslyng, M. D., & Bredie, W. L. (2012). Rapid
descriptive sensory methods–comparison of free multiple sorting, partial napping,
napping, flash profiling and conventional profiling. Food Quality and Preference, 26,
267–277.
Esmerino, E. A., Tavares Filho, E. R., Carr, T. B., Ferraz, J. P., Silva, H. L. A., Pinto, L. P. F.,
... Bolini, H. M. A. (2017). Consumer-based product characterization using Pivot
Profile, Projective Mapping and Check-all-that-apply (CATA): A comparative case
with Greek yogurt samples. Food Research International, 99, 375–384.
Faye, P., Brémaud, D., Daubin, M. D., Courcoux, P., Giboreau, A., & Nicod, H. (2004).
Perceptive free sorting and verbalization tasks with naive subjects: An alternative to
descriptive mappings. Food Quality and Preference, 15, 781–791.
Fleming, E. F., Ziegler, G. R., & Hayes, J. E. (2015). Check-all-that-apply (CATA), sorting,
and polarized sensory positioning (PSP) with astringent stimuli. Food Quality and
Preference, 45, 41–49.
Heymann, H., Hopfer, H., & Bershaw, D. (2014). An exploration of the perception of
minerality in white wines by projective mapping and descriptive analysis. Journal of
Sensory Studies, 29, 1–13.
Hopfer, H., & Heymann, H. (2013). A summary of projective mapping observations – The
effect of replicates and shape, and individual performance measurements. Food
Quality and Preference, 28, 164–181.
Horita, C. N., Esmerino, E. A., Vidal, V. A. S., Farah, J. S., Amaral, G. V., Bolini, H. M. A.,
... Pollonio, M. A. R. (2017). Sensory profiling of low sodium frankfurter containing
garlic products: Adequacy of polarized projective mapping compared with trained
panel. Meat Science, 131, 90–98.
Husson, F., Josse, J., Lê, S., & Mazet, J. (2011). FactoMineR: Multivariate Exploratory
Data Analysis and Data Mining with R. R package version 1.16. http://factominer.
free.fr,http://www.agrocampus-rennes.fr/math/.
Jeon, H., Lee, S., Kim, S., & Kim, Y. (2016). Quality characteristics of modified fermented
soybean paste and traditional fermented soybean paste. Journal of the Korean Society
of Food Science and Nutrition, 45, 1001–1009.
Kennedy, J., & Heymann, H. (2009). Projective mapping and descriptive analysis of milk
and dark chocolates. Journal of Sensory Studies, 24, 220–233.
Kim, H. G., Hong, J. H., Song, C. K., Shin, H. W., & Kim, K. O. (2010). Sensory char-
acteristics and consumer acceptability of fermented soybean paste (Fermented soy-
bean paste). Journal of Food Science, 75, S375–S383.
Kim, M. J., & Rhee, H. S. (1988). The contents of free amino acids nucleotides and their
related compounds in soypaste made from native and improved meju and soypaste
product. Journal of the Korean Society of Food Science and Nutrition, 17, 69–72.
Kim, M. K., & Lee, K. G. (2014). Correlating consumer perception and consumer ac-
ceptability of traditional fermented soybean paste in Korea. Journal of Food Science,
79, S2330–S2336.
Kim, Y. K., Jombart, L., Valentin, D., & Kim, K. O. (2013). A cross-cultural study using
Napping®: Do Korean and French consumers perceive various green tea products
differently? Food Research International, 53, 534–542.
King, M. C., Cliff, M. A., & Hall, J. W. (1998). Comparison of projective mapping and
sorting data collection and multivariate methodologies for identification of similarity
of use of snack bars. Journal of Sensory Studies, 13(3), 347–358.
Lee, S. B. (2004). Fermented foods. Seoul: HyoIl65–82.
Lelièvre, M., Chollet, S., Abdi, H., & Valentin, D. (2008). What is the validity of the sorting
task for describing beers? A study using trained and untrained assessors. Food Quality
and Preference, 19, 697–703.
285
Food Quality and Preference 71 (2019) 279–285
Lezaeta, A., Bordeu, E., Næs, T., & Varela, P. (2017). Exploration of consumer perception
of Sauvignon Blanc wines with enhanced aroma properties using two different de-
scriptive methods. Food Research International, 99, 186–197.
Liu, J., Grønbeck, M. S., Di Monaco, R., Giacalone, D., & Bredie, W. L. (2016).
Performance of Flash Profile and Napping with and without training for describing
small sensory differences in a model wine. Food Quality and Preference, 48, 41–49.
Louw, L., Malherbe, S., Naes, T., Lambrechts, M., van Rensburg, P., & Nieuwoudt, H.
(2013). Validation of two Napping® techniques as rapid sensory screening tools for
high alcohol products. Food Quality and Preference, 30, 192–201.
Louw, L., Oelofse, S., Naes, T., Lambrechts, M., van Rensburg, P., & Nieuwoudt, H.
(2015). Optimisation of the partial napping approach for the successful capturing of
mouthfeel differentiation between brandy products. Food Quality and Preference, 41,
245–253.
Marcano, J., Ares, G., & Fiszman, S. (2015). Comparison of partial and global projective
mapping with consumers: A case study with satiating cheese pies. Food Research
International, 67, 323–330.
Moelich, E. I., Muller, M., Joubert, E., Næs, T., & Kidd, M. (2017). Validation of projective
mapping as potential sensory screening tool for application by the honeybush herbal
tea industry. Food Research International, 99, 275–286.
Nestrud, M. A., & Lawless, H. T. (2010). Perceptual mapping of apples and cheeses using
projective mapping and sorting. Journal of Sensory Studies, 25, 390–405.
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.
Pagès, J., Cadoret, M., & Le, S. (2010). The sorted napping: A new holistic approach in
sensory evaluation. Journal of Sensory Studies, 25, 637–658.
Park, H. S., Lê, S., Hong, J. H., & Kim, K. O. (2014). Sensory perception of Yackwa
(Korean traditional fried cookie) by consumer groups of different age using the sorted
napping procedure. Journal of Sensory Studies, 29, 425–434.
Perrin, L., & Pagès, J. (2009). Construction of a product space from the ultra-flash pro-
filing method: Application to 10 red wines from the Loire Valley. Journal of Sensory
Studies, 24, 373–395.
Perrin, L., Symoneaux, R., Maître, I., Asselin, C., Jourjon, F., & Pagès, J. (2008).
Comparison of three sensory methods for use with the Napping® procedure: Case of
ten wines from Loire valley. Food Quality and Preference, 19(1), 1–11.
Reinbach, H. C., Giacalone, D., Ribeiro, L. M., Bredie, W. L. P., & Frøst, M. B. (2014).
Comparison of three sensory profiling methods based on consumer perception: CATA,
CATA with intensity and Napping®. Food Quality and Preference, 32, 160–166.
Risvik, E., McEwan, J. A., Colwill, J. S., Rogers, R., & Lyon, D. H. (1994). Projective
mapping: A tool for sensory analysis and consumer research. Food Quality and
Preference, 5, 263–269.
Risvik, E., McEwan, J. A., & Rødbotten, M. (1997). Evaluation of sensory profiling and
projective mapping data. Food Quality and Preference, 8, 63–71.
Santos, B. A., Pollonio, M. A. R., Cruz, A. G., Messias, V. C., Monteiro, R. A., Oliveira, T. L.
C., ... Bolini, H. M. A. (2013). Ultra-flash profile and projective mapping for de-
scribing sensory attributes of prebiotic mortadellas. Food Research International, 54,
1705–1711.
Torri, L., Dinnella, C., Recchia, A., Naes, T., Tuorila, H., & Monteleone, E. (2013).
Projective Mapping for interpreting wine aroma differences as perceived by naïve and
experienced assessors. Food Quality and Preference, 29, 6–15.
Tuorila, H., & Monteleone, E. (2009). Sensory food science in the changing society:
Opportunities, needs, and challenges. Trends in Food Science & Technology, 20, 54–62.
Valentin, D., Chollet, S., Lelièvre, M., & Abdi, H. (2012). Quick and dirty but still pretty
good: a review of new descriptive methods in food science. International Journal of
Food Science and Technology, 47, 1563–1578.
Varela, P., Antúnez, L., Berget, I., Oliveira, D., Christensen, K., Vidal, L., ... Ares, G.
(2017). Influence of consumers' cognitive style on results from projective mapping.
Food Research International, 99, 693–701.
Varela, P., Berget, I., Hersleth, M., Carlehög, M., Asioli, D., & Næs, T. (2017). Projective
mapping based on choice or preference: An affective approach to projective mapping.
Food Research International, 100, 241–251.
Varela, P., & Salvador, A. (2014). Structured sorting using pictures as a way to study
nutritional and hedonic perception in children. Food Quality and Preference, 37,
27–34.
Vidal, L., Cadena, R. S., Antunez, L., Gimenez, A., Varela, P., & Ares, G. (2014). Stability
of sample configurations from projective mapping: How many consumers are ne-
cessary? Food Quality and Preference, 34, 79–87.
Williams, E. J. (1949). Experimental designs balanced for the estimation of residual ef-
fects of treatments. Australian Journal of Chemistry, 2(2), 149–168.
Zhang, T., Lusk, K., Mirosa, M., & Oey, I. (2016). Understanding young immigrant
Chinese consumers’ freshness perceptions of orange juices: A study based on concept
evaluation. Food Quality and Preference, 48, 156–165.