Introduction

Fermented soybean products have traditionally been consumed in many Asian countries (Chung and Chung, 2008; Shin and Jeong, 2015), such as Miso in Japan, Tempeh in Indonesia, and Dubanjang in China. Doenjang is one of the most essential fermented soybean product in Korea (Kim and Lee, 2014; Kim et al., 2010) and it is frequently used in daily meals. Because Doenjang is very important in Korean cuisine, various studies on Deonjang have been conducted for processing methods (Jo et al., 2011; Ko et al., 1999; Lee et al., 2014; Park et al., 1994; 2000), flavor compounds (Jo et al. 2011; Kim et al., 1992; Kim and Rhee, 1990; Lee and Ahn, 2009; Park et al., 1994), sensory characteristics (Chung and Chung, 2008; Kim et al., 2010; Lee and Ahn, 2009; Roh et al., 2018), and consumer perception (Chung and Chung, 2008; Kim and Lee, 2014; Kim et al., 2010; Kwak et al., 2017; Roh et al., 2018).

Doenjang soup is frequently consumed in Korea and its soup base is prepared with beef or anchovy. Flavor enhancers are often added such as monosodium glutamate (MSG) to achieve a rich flavor in various dishes (Bellisle et al., 1991; Jinap and Hajeb, 2010; Sand, 2005). However, as consumers prefer naturally-made and healthy seasoning compounds, MSG substitutes containing high levels of glutamic acid obtained great attention from consumers, such as hydrolyzed vegetable proteins (HVPs) (Jo and Lee, 2008).

HVP is a seasoning agent that can impart savory flavor to food products. It has been widely used in industry all over the world (Manley, 1994). The conventional HVP is generally produced by acid hydrolysis, often using hydrochloric acid (HCl), of various protein sources mainly from soy, wheat gluten, rice, and maize. HVP contains free amino-acids and peptides which provide flavor enhancing effects (Aaslyng et al., 1998). An alternative to acid-hydrolyzed vegetable protein (aHVP), enzyme-hydrolyzed vegetable protein (eHVP) is produced using proteolytic enzymes under a more neutral pH and lower temperatures (Pommer, 1995). Because of the different processing conditions, the two types of HVP have different sensory profiles. Acid-hydrolyzed vegetable protein is usually dark-brown in color and has a strong savory flavor, whereas enzyme-hydrolyzed vegetable protein usually is lighter in color and has a mild savory flavor (Aaslyng et al., 1998; Weir, 1992). However, the contribution to sensory profile of these substances in actual food system has not been reported in previous studies.

It would be of great significance to investigate consumer preferences and perceived sensory attributes when various HVPs were applied to real food. The purpose of this study was to investigate the effects of aHVP/eHVP on consumers’ responses of Korean Doenjang soup and to compare the flavor enhancing effect with MSG-added samples. In addition, age-dependent differences have also been studied because consumers’ experience for traditional foods like Doenjang may vary with one’s age (Kwak et al., 2017; Roh et al., 2018).

Materials and methods

Subjects

One hundred and twenty female consumers who consume Doenjang soup on a regular basis were recruited over the internet, and by flyers at Ewha Womans University (Seoul, South Korea). Consumers were selected according to 2 different age groups in order to observe possible differences (N = 59, ages 18–29; N = 61, ages 40–62). Each consumer received monetary compensation for their participation in the study. Ethical approval to perform the study was obtained from Ewha Womans University Institute Review Board (IRB No.: 130-10).

Sample preparation and presentation

Two types of commercial aHVPs and eHVPs produced from wheat gluten were used respectively for this study. For the comparison, two different levels of MSG-added samples were included (Table 1). In consideration of the influence by salty taste on preference, a salt-added sample (NA) was also included in a manner similar to salty taste of 0.4% MSG-added sample. The levels of each ingredient were determined during preliminary test. The MSG content was determined at a level where flavor enhancing effects could be detected. The content of other HVPs was determined to be equal to the amount of MSG. The Doenjang used in this experiment was specially prepared to contain low sodium (8% salt) in order to use the effect of additives which give salty taste. The approximate salt content of the soup sample was 0.6%.

Table 1 Information of additional ingredients added to the Doenjang soup samples
Full size table

The Doenjang soup was prepared with the modified procedure of Roh et al. (2018). The main stock of Doenjang soup was prepared by boiling 1 kg of brisket beef in 10 L water for an hour. After removing the solid ingredients, the stock was cooled and kept in a refrigerator (3 °C) for 18 h and was then strained with filter paper to remove solidified fat. The prepared beef stock was divided into a portion of 4 L in an airtight container (295 × 230 × 188 mm, HPL883, Lock&lock Co.,Ltd., Seoul, Korea) and was kept in a freezer (− 24 °C) until usage (within 1 week).

The Doenjang soup samples were prepared 2 h before the evaluation. The 4 L of frozen beef stock was thawed in a microwave oven for 30 min and then heated for 1 h on a hot plate until boiling. To this, 400 g of Doenjang paste, 0.8 g of green onion powder, and 0.6 g of garlic powder (Garunara, Seoul, Korea) were added. After further heating for 5 min, each of flavor enhancers (aHVP, eHVP, or MSG) or NaCl was added and kept heating for another 3 min. Each of the prepared Doenjang soup sample was transferred to 3 L thermos (TAE2500P, Thermos®, Tokyo, Japan) prior to serving. Aliquot (70 ml) of samples were presented in plastic bowls (95 mm in diameter, 50 mm in height, Easepack Co., Gyeonggi, Korea) with plastic spoons right before the evaluation to maintain the temperature (55 ± 3 °C). Samples were monadically presented in a randomized order. Filtered tap water (22 ± 2 °C) was provided for mouth rinsing.

Evaluation procedure

On arrival, consumers received verbal and written instructions for the test. Then, they tasted each sample and rated overall liking, flavor liking, saltiness liking, umami-taste liking, saltiness intensity, umami-taste intensity, beef-flavor intensity, Doenjang-flavor intensity, familiarity (how familiar are you with the taste and flavor of this sample?), and congruency (how harmonious do you think the overall taste and flavor of this sample is?) of 8 samples using a 15-point scale (e.g.dislike extremely = 1, like extremely = 15; weak = 1, strong = 15). The participants were instructed to rinse their palate with provided water before tasting any sample. Between the samples, 3.5 min breaks were taken. After completion of the test, they answered to the general questionnaire for Deonjang soup including general preference, consumption frequency, and consumption occasion. The session took approximately 45 min.

Statistical analysis

ANOVA was performed to examine the differences between samples in each evaluated parameter. Duncan’s multiple range tests were conducted as post hoc comparisons (p < 0.05). ANOVA was also performed to analyze the effects of age group, sample, and their interaction for each parameter. Pearson’s correlation coefficients between overall liking and the rest of the other evaluated parameters were also calculated. Cluster analysis was conducted to segment consumers based on their overall liking of the 8 Doenjang soup samples. In order to investigate the relationship between parameters of sensory perception and liking, principal component analysis (PCA) was conducted on the mean sensory perception data of the samples, and the mean liking parameters were projected as the supplementary variables onto the sample space. PCA was performed with total consumer data and each clusters’ data, respectively. Internal preference mapping (IPM) was carried out on the hedonic ratings of young and old groups, respectively in order to examine the effects of age. All the analyses were performed using SPSS for Windows 18.0 (SPSS Inc. Chicago, IL, USA) and XLSTAT for Windows ver. 18.06 (Addinsoft Inc., Paris, France).

Results and discussion

Sensory perception and liking of total consumers for Doenjang soup samples

Applying analysis of variance (ANOVA) to total consumer data on Doenjang soup samples revealed significant sample effects on all of the evaluated parameters (p < 0.05). There were differences between the samples when comparing the average scores for each parameter, as indicated in Table 2.

Table 2 Mean rating scores for Doenjang soup samples (n = 120)
Full size table

The mean scores for overall liking decreased in the order of E1 > H2 > NA > E2 > ML, MH > CON > H1. Consumers significantly preferred certain types of HVP-added samples (E1 and H2), and exhibited lower preferences for H1, CON, ML, and MH.

The flavor liking scores showed a similar tendency to those of the overall liking scores. The main difference was that the score for MH was slightly higher than that for ML, while these scores being the same for overall liking. The correlation coefficient between these two parameters was high (r = 0.899). This result also seems reasonable given that the characteristic of the Doenjang soup used in this experiment had few affecting sensory characteristics other than flavor.

The saltiness liking scores showed a different tendency to those of the overall liking scores. Compared with overall liking, the salty-taste liking for H2 and E1 was the highest as similar to the overall liking but the salty-taste liking for CON was higher than that for ML and E2, and was lowest for E2, whereas E2 received a higher score than ML, MH, CON, and H1 for overall liking. However, there was still a significant correlation between overall liking and salty-taste liking (r = 0.636), which means that the saltiness liking influenced the overall liking to some extent.

The umami-taste liking decreased in the order of E1 > H2 > NA > MH > CON > E2 > ML > H1. The overall liking score was slightly higher for E2 than for MH, whereas the umami-taste liking score was higher for MH than for E2. However, since this difference was not statistically significant, the umami-taste liking appeared to show a similar tendency to that for overall liking, as confirmed by a strong correlation between the two parameters (r = 0.794).

The saltiness intensities of H1, H2, E1, and E2—which are samples to which HVP had been added—were higher than those of the CON and ML samples. HVP is produced using vegetable proteins. The manufacturing process involves hydrolyzing the protein with hydrochloric acid (HCl) for about 8 h, and it undergoes a neutralization process after cooling, and an alkali (e.g., sodium hydroxide) is added at that stage to react with HCl to produce NaCl. Enzymes are used to break down the proteins during the production of enzymatic HVP. To break down the protein into amino acids, proteases are added to the mixture of defatted protein and water. Since no salt is formed during the production process, salt is commonly added as an ingredient (Aaslyng et al., 1998). The present results indicated that E1 and E2 (eHVP-added samples) showed a higher salty taste intensity than NA (NaCl-added sample), while samples with added MSG or aHVP showed a lower salty taste intensity. Comparing the results for saltiness intensity with those for overall liking revealed that the moderately liked sample E2 exhibited the saltiness intensity with the highest score. In addition, the salty taste intensity of H2 (which received a high score for all of the liking parameters) had a relatively low value. Since the correlation between these two parameters was very weak (r = 0.170), it is thought that the intensity of salty taste of the samples used in this study did not greatly affect overall liking.

For the intensity of the umami taste, E1 and H2 (which contain added eHVP or aHVP) showed higher intensities than MH. Other HVP-added products (E2 and H1) showed lower intensities than the MH sample. It seems that the HVPs used in the Doenjang soup samples in this experiment (regardless of type) showed different intensity tendencies. CON and ML showed the lowest scores for umami taste. CON, which contained no added flavor enhancer, showed an intensity score for umami taste of at least 8 points. This is probably attributable to the beef stock used as the base of Doenjang soup and also to the flavor ingredients in Doenjang. Free amino acids and substances related to nucleic acid (IMP and GMP) that are produced during Doenjang fermentation are known to be main compounds of the umami taste (Kim and Rhee, 1990; 1998). Comparing the results for overall liking and the intensity of umami taste revealed that the umami taste of the E1 and H2 samples were preferred by the consumers. This is also interpreted as indicating that the intensity of the umami taste of the Doenjang soup sample plays an active role in consumer liking. This tendency is consistent with previous research findings that MSG improves flavor when added to a savory food system and increases liking among consumers (Bellisle et al., 1991; Prescott, 2001).

The beef flavor intensity decreased in the order of NA > H2, E2, and E1 > MH > H1 > CON > ML. As compared with CON, a beef-flavor-enhancing effect was shown in samples with added flavor enhancers (i.e., H2, E2, E1, and MH) except ML. This is in agreement with the previous research that HVPs promote a beef-like flavor (Aaslyng et al., 1998; Weir, 1992).

The intensity of the Doenjang flavor decreased in the order of E2 > MH, H1 > NA > H2 > E1 > CON > ML. The Doenjang flavor was enhanced in samples with added aHVP, eHVP, and MSG compared to that in CON. When compared with the overall liking results, the intensity of the Doenjang flavor of the E1 (which had the highest liking score) appeared relatively low, whereas H1 (with the lowest liking score) had a relatively strong Doenjang flavor. It was assumed that the Doenjang flavor intensity seemed to negatively affect overall liking.

The familiarity scores of the samples showed a similar tendency to the overall liking scores. The most preferred samples, E1 and H2 displayed higher familiarity. Consumers seemed to be familiar with these ingredients since they have been already used in commercial products (e.g., compound seasoning foodstuffs, soy sauce, and ramen soup). The familiarity showed a strong correlation with overall liking (r = 0.785). This result is supported by the previous study (Chrea et al., 2008).

The score for congruency decreased in the order of E1 > H2 > NA > E2 > MH > ML > CON > H1. The congruency appeared to be the highest for E1 and H2, which were the most-preferred samples, confirming that aHVP and eHVP were well matched with Doenjang soup and increased consumer preference. The congruency of the samples also had a strong correlation with overall liking (r = 0.846) and this result is consistent with recent study (Amsellem and Ohla, 2016).

As indicated above, scores of overall liking and flavor liking showed lower values for CON (a basic sample) than for the other samples. Salty taste and the umami-taste liking for CON also showed relatively low. Consistent with other intensity characteristics, familiarity and congruency of CON also showed low values, and consumers seemed to feel that CON (with a low flavor intensity) was less satisfactory than other samples containing added salt and flavor enhancers. In the case of the NA sample, to which only salt was added, the liking score was higher than that for CON, and the flavor intensities were also increased. Also, the familiarity scores were significantly higher than that for CON, which means that the salt content of Doenjang soup used in this study had a positive influence on consumer liking. The ML and MH samples contained different concentrations of added MSG, but their overall liking score appeared the same, while flavor intensity and familiarity were lower for ML than for MH. All aspects of liking, intensity, familiarity, and congruency, exhibited high scores for the most-preferred eHVP-added sample, E1.

In PCA, PC1, and PC2 explained 73.49% and 21.21% of the total variance, respectively (Fig. 1). It was found that all of the liking parameters (overall liking, flavor liking, salty-taste liking, and umami-taste liking) were located on the positive dimension of PC1 close to familiarity and congruency. Familiarity and congruency were located closest to E1, the most preferred sample, and attributes of umami taste, beef flavor, and salty taste were nearby. It seems that these attributes had positive influences on the liking score. H1 was positioned on the negative PC1 dimension, far from the liking parameters, familiarity, and congruency, but this sample was located closer to the sensory characteristics above mentioned compared to ML and CON. This indicates that the H1 might have other flavor characteristics (not evaluated in this study) which seemed to negatively affected consumer liking. Further study is recommended in this aspect.

Fig. 1
figure 1

PC loadings of the sensory perception data and Doenjang soup samples for total consumers with liking data as supplementary variables (n = 120). OVL, Overall liking; FL, Flaver liking; SL, Saltiness liking; ML, Umami taste liking; Salty_Int, Saltiness intensity; MSG_Int, Umami taste intensity; Beef_Int, Beef flavor intensity; Doenjang_Int, Doenjang flavor intensity

Full size image

The main flavor characteristics enhanced when HVP was added were salty taste, umami taste, beef flavor, and Doenjang flavor. Consumers reported that they felt more familiar with the samples containing aHVP and eHVP (especially E1 and H2) than samples with MSG. The role of HVP as a natural flavor enhancer has been reported previously (Manley, 1994), and the results from this study indicate that aHVP/eHVP can be used as a substitute season ingredient for MSG in Doenjang products.

Sensory perception and liking for Doenjang soup samples according to the different clusters

In the cluster analyses, the consumers were divided into three clusters. The mean liking scores of each cluster were projected as supplementary variables onto the sample space derived from PCA for each cluster (Fig. 2). The PCA result of cluster 1 (n = 46) showed that all liking parameters, familiarity, and congruency were located closely to the positive dimension of PC1 and PC2. Umami taste and beef flavor were relatively close to the liking parameters, which seems to have a positive effect on liking of cluster 1. This trend was similar to the result of total consumers, while the salty taste was located on negative PC2 far from the liking parameters. It seems that a salty taste negatively affected the liking score for cluster 1. E1 was the most-preferred sample for the total consumers, while the consumers in cluster 1 preferred ML the most. The intensity of salty taste was lowest for ML, which had the highest degrees of congruency and familiarity. Also, the tendency to dislike H1 appeared stronger for cluster 1 than for the total consumers. From the results of general questionnaire for the Doenjang soup, it was found that consumers in this cluster were more likely to prefer Doenjang soup than those in the other clusters (95.6%, 86.7%, and 93.1% for clusters 1, 2, and 3, respectively). They also showed a higher frequency of consumption for Doenjang soup, with more than 80.4% of the consumers consuming Doenjang soup at least once a week (57.8% for cluster 2 and 55.2% for cluster 3). Consumers in cluster 1 preferred E1, and H2 over MH. These results suggest that HVP was more beneficial to the consumers who consumed Doenjang soup more often.

Fig. 2
figure 2

PC loadings of the sensory perception data and Doenjang soup samples for each cluster group with liking data as supplementary variables. OVL, Overall liking; FL, Flaver liking; SL, Saltiness liking; ML, Umami taste liking; Salty_Int, Saltiness intensity; MSG_Int, Umami taste intensity; Beef_Int, Beef flavor intensity; Doenjang_Int, Doenjang flavor intensity

Full size image

The result of cluster 2 (n = 45) showed similar trend to that of cluster 1. All liking parameters were loaded close to familiarity and congruency on positive PC1. In addition, all sensory attributes (umami taste, beef flavor, Doenjang flavor, and saltiness) were located on positive PC1, closer to the overall liking when compared to the total consumer results indicating that cluster 2 was more influenced by these characteristics. Similar to the results of the total consumers, E1 was the most preferred, but NA, H2, and H1 samples also have relatively highly preferred compared to the total consumers. In contrast to cluster 1, where ML was the most preferred, cluster 2 least preferred ML. For cluster 2, ML had the lowest level of familiarity and congruency among the samples and also lower levels of flavor intensities. The cluster 3 (n = 29) was relatively minor group and showed different tendency from cluster 1 and 2. This group preferred CON the most and disliked NA which had the highest saltiness intensity. Flavor likings of Doenjang soup were different according to each cluster. This indicated that the preference for Doenjang soup samples was affected by the consumers’ preference for flavor.

Sensory perception and liking for Doenjang soup samples by young and old consumers

The results from the ANOVA of Doenjang soup samples according to age groups showed a significant main effect of sample in all of the evaluations (p < 0.05). Age did not influence the overall liking (F = 0.02, p = 0.880) and flavor-liking scores (F = 1.36, p = 0.244), and there was also no significant effect of interaction between sample and age group (F = 0.6, p = 0.755; F = 0.8, p = 0.584, respectively). However, young and old consumer groups showed differences in ratings for saltiness liking (age effect: F = 5.79, p = 0.016), intensity of salty taste (age effect: F = 15.93, p < 0.0001), umami-taste intensity (age effect: F = 4.14, p = 0.042), and Doenjang flavor intensity (age effect: F = 14.76, p < 0.0001) for the various Doenjang soup samples (Table 3). In general, older consumers responded with higher scores for saltiness liking and umami-taste intensity compared to the younger consumers, while younger consumers gave higher scores for saltiness and Doenjang flavor intensity. This indicates that the older consumers seemed to be more influenced by the umami taste than the younger consumers and that the younger consumers were more conservative about the saltiness and Doenjang flavor of the Doenjang soup samples.

Table 3 Mean rating scores for Doenjang soup samples by age groups
Full size table

Internal preference maps for the younger and older consumers are shown in Fig. 3. Most of the younger consumers were located on the negative F2 dimension, which is where E1 and H2 were located. Similar trends were found for the older consumers, with most of them located on the positive F1 dimension, which is also where E1 and H2 were located. Also, H1 was the least-preferred sample among both the younger and older consumers. However, the IPM revealed that the older consumers were more spread out than the younger consumers. It seemed that since they had more experience of consuming Doenjang soup compared to those in the younger group, their criteria for their favorite Doenjang soup could have been more clearly formed. This can be confirmed from the general questionnaire results, in that the average weekly intake frequency of Doenjang soup was 83.7% among older consumers and 47.5% among younger consumers. Also, most of the older consumers were housewives, with 92% of them eating cooked food at home, while this was the case for 78% of the younger consumers. Product familiarity is one of the most important determinants influencing the overall liking of consumers for traditional fermented foods such as Doenjang (Roh et al., 2018), although the present study found that familiarity with Doenjang samples used in this study did not differ significantly with age.

Fig. 3
figure 3

Internal preference mapping of hedonic ratings by age groups

Full size image