Consumer perception and liking, and sensory characteristics of blended teas

Abstract

This study was conducted to identify the sensory characteristics of six blended teas containing different ingredients and analyze penalty factors for the products based on consumer acceptance, check-all-that-apply questions, and the just-about-right scale. Ten trained panelists created a descriptive set of 23 sensory attributes, and 93 consumers participated in the tests. The attributes were analyzed by classifying them as per appearance, odor/aroma, flavor/taste, texture/mouthfeel, and any aftertaste. Principal component analysis results showed that the blended teas were differentiated by artificial fruit flavor. According to the results of this study, the ideal products should be relatively sweet, mild, fruit flavored, and not too bitter, astringent, pungent, and strong or do not have fermented flavor; astringency is the most troublesome attribute. The consumers preferred teas that were less bitter and less astringent and did not leave the tongue coated with powder; therefore, these attributes were believed to act as drivers of “dislike.”

Introduction

Health is becoming an increasingly important personal and social value as the interest in wellness increases with the aging population. Consumers have begun to understand that their food choices can affect their health and are paying more attention to the health benefits of food in their quest to maintain a healthy lifestyle (Goetzke et al., 2014). This trend is affecting the mainstream global beverage market, especially coffee and sugary sodas, with a growing number of consumers now drinking healthy tea (Bolton, 2019).

Tea is one of the oldest drinks. It is made by processing the plant’s leaves and is classified into green tea, black tea, or oolong tea according to the manufacturing process. Tea can be further categorized into single teas and blended teas—a mix of two or more types of tea materials and also blended with some herbs (Lee et al., 2013). Blended teas have a richer taste and aroma than single teas, and the consumer can expect not only potency from the ingredients but also the synergistic effects of both the taste and the aroma. The millennial generation has been a driving force behind the growth of the world’s tea market, and blended teas are positioning themselves among teas enjoyed by the younger generations (FAO, 2018).

The definition and classification of blended teas varies according to researchers, as the field of blended teas is still in the early stages of market maturity and study in the domestic tea market. Blended teas can be categorized as follows: tea with other tea, tea with herbs, tea with fruits or petals, and tea with a variety of flavors (Ko and Park, 2016). Some researchers also consider blending based on the harvest period or manufacturing method leading to a wide range of blended teas (Lee, 2018).

Teas were originally consumed for their taste and aroma, but a recent awareness of their health benefits by consumers has attracted more attention (Byun and Han, 2004). Tea-related studies have been conducted for many years but mostly about the physiological activities or sensory characteristics of single teas (Alasalvar et al., 2012; de Godoy et al., 2013; Dos et al., 2005; Lee and Chambers, 2007; Lee et al., 2008a; 2008b; 2008c). Few sensory evaluation studies have been conducted on consumer preferences for blended teas (Kim et al., 2018a; 2018b; Oduro et al., 2013). It is crucial to focus on high-quality products when a new product development is undertaken. For consumers, it is important that new products have competitive sensory quality and that production is economically viable for the local community. Sensory quality is seen as one of the most important characteristics in the final acceptance in the market.

Penalty analysis combines the just-about-right (JAR) and overall liking tests to identify the optimum intensity of sensory attributes (Popper, 2005). Using a five- or seven-point bipolar scale, consumers are asked to indicate if the intensity of a sensory attribute is too strong, too weak, or JAR (Narayanan et al., 2014), thereby allowing the assessor to determine which attribute could affect the overall acceptability of the product. Also, Adams et al. (2007) proposed that check-all-that-apply (CATA) questions should be used to allow consumers to indicate their sensory perceptions of samples in a hedonic evaluation. CATA questions provide multivariate binary data that demonstrate the applicability of the descriptors applied to the samples. As this method is a very quick, simple, and easy task for the consumer, CATA questions are being increasingly used in consumer questionnaires. For a comprehensive understanding of consumer acceptability, consumer liking ratings are often interpreted using correlations with other sensory data, such as descriptive analysis data, CATA, and/or JAR. This study was conducted with the expectation of how the analysis of the results could be enriched using CATA analysis in conjunction with the JAR analysis. Therefore, the purpose of this study was to define the sensory characteristics of blended teas, to identify the drivers that affect positively and negatively affect consumer acceptability based on CATA test and JAR penalty analysis.

Materials and methods

This study was reviewed and approved by the Kookmin University Institutional Review Board, Seoul, Korea (approval number: KMU-201807-HR-181).

Materials

In the preliminary study conducted in our lab, young consumers in their 20 s and 30 s (n = 112) usually drank a variety of teas, including green tea, rooibos tea, ginger tea, chamomile tea, dried petals tea, and oolong tea. In this study, various types of blended teas were selected to identify the sensory characteristics and penalty analysis of blended teas preferred by young consumers. Therefore, a total of six blended teas were selected as samples, which refer to the types of blended teas classified in the study of Ko and Park (2016) and the preliminary findings.

The samples are as follows: one sample of herbs blended with teas (CMint_Green), three samples blended among herbs (G_Breath, P_Mind_Chamomile, and W_Ginger), and two samples of flavored tea (H_Rooibos and Peach_Oolong). CMint_Green is a blended tea with green tea and two types of mint herb. G_Breath is a variety of herb blended tea with lemongrass and peppermint added to dried lavender flowers. P_Mind_Chamomile is an herb blended tea with orange peel and peppermint as the main ingredient. W_Ginger is a blended tea with three kinds of oriental herbs: ginger as the main ingredient, jujube, and dried balloon flower roots. H_Rooibos is a strawberry-flavored tea based on rooibos, which has been gaining preference among young consumers with non-caffeine tea. Peach_Oolong is a flavored tea based on oolong tea.

All six samples were blended teas with the LookOurTea brand manufactured by Resh Corp. (Gyeonggi-do, South Korea), a company that develops and produces a variety of blended teas in Korea. By selecting the same brand of samples, any differences in ingredients and manufacturing methods between brands are avoided. The blended teas differed by the main ingredients in the tea bags, which were made of bioplastic (polylactide) and contained 2 g of tea. The product abbreviation, main ingredients, and brewing conditions of the blended tea samples used in this study are presented in Table 1.

Table 1.

Characteristics of the six blended tea products used in this study

Sample	Main ingredients	Brewing conditions
		Water (mL/tea bag)	Temperature (°C)	Time (min)
CMint_Green	Green tea 80%, peppermint 5%, spearmint 15%	200	80 ± 2	2
G_Breath	Lemongrass 30%, lavender 20%, peppermint 20%, spearmint, thyme leaf		95 ± 2	3
H_Rooibos	Rooibos 98.6%, strawberry flavor, lemon oil
P_Mind_Chamomile	Chamomile 66.7%, orange peel 20%, peppermint
Peach_Oolong	Oolong tea 98%, peach flavor
W_Ginger	Ginger, jujube, dried balloon flower roots, dried burdock, orange peel, rosemary, cinnamon

Descriptive analysis

Sample preparation

Six blended tea samples were prepared for sensory evaluation. Freshly boiled mineral water (1.2 L) was poured into individual 1.5 L stainless steel thermos flasks (LHC6600B, Lock&Lock Co. Ltd., Seoul, Korea) pre-warmed with boiling water. After the water temperature was allowed to decrease to the appropriate steeping level (Table 1), six sample tea bags (excluding CMint_Green) were put into separate flasks, infused for three min, and removed. According to Richardson (2016), oolong tea should be extracted at 82 °C to 93 °C for 3–5 min. The herb tea was found to be ideal for extraction at 100 °C for 5–10 min. Based on this, an experiment was conducted to find the extraction temperature for each sample. As a result, the most effective temperature and time for each product were identified and applied. CMint_Green was infused at 80 ± 2 °C for 2 min. In green tea, caffeine and polyphenols with bitter and astringent tastes are released during brewing at high temperatures, but amino acids with rich flavors are released at relatively low temperatures (e.g., 60 °C); therefore, temperatures between 60 °C and 80 °C are found to be suitable (Labbé et al., 2006). In addition, although most of the ingredients in green tea are released within three min, the longer the extraction time (increased extraction of polyphenols) is, the stronger is the bitterness and the lesser is the fresh taste of the tea, making it worse overall. Therefore, to properly release the effective ingredients and ensure a green tea with appealing taste, it is best to steep for approximately 2–3 min at the temperature range noted above (Choi, 2002; Lee et al., 2008a; 2008b; 2008c). After steeping, the flask lid was replaced to minimize heat and aroma loss during sample evaluation.

Sample presentation

Each sample was provided to the panelists in 2.5 oz. (about 75 mL) white paper cups (EVERYPACK, SJ Co., Seoul, Korea) that were coded with a randomly selected three-digit number. About 50 mL of each infusion were served and were approximately 60 °C–70 °C at the time of tasting. The panelists received mineral water to cleanse their palates between samplings of tea.

Panelist selection and training

To determine the objective perceptions of the various blended teas, trained panelists were asked to conduct a descriptive analysis of the samples. Ten Korean subjects (5 males and 5 females, aged 20–26 years), who were interested in the sensory evaluation of the blended teas and had experience in the descriptive analysis of various food systems, were selected from a community church in Seongnam-si, Gyeonggi-do, Korea.

Before the assessments, the panelists completed two training sessions to select the relevant sensory attributes of the teas. The subjects were first exposed to various types of blended teas to become familiar with the general category of green tea. Once trained using the same samples intended in the main experiment, the participants developed a list of terms that describe the sensory attributes of the blended teas using those samples, generated the proper attributes for each sample, and defined and selected reference samples for each of the descriptors (Lee et al., 2009). Also, the reference intensity point for reference samples was established so that the criteria for scores could be set for repeated tests (Neal et al., 2010). The panelists’ results which were not within three points of the mean of ratings from all replicates of each panelist were calibrated by calculating a mean rating. Based on the method of Villarino et al. (2007) and the previous study of Yang et al. (2012), those whose ratings were not within three points of the mean were conducted to reevaluate the standard and adjust their rating until a consensus was reached (Villarino et al., 2007; Yang et al., 2012). Training sessions were held for 3 days a week for 2 weeks, and each training session lasted approximately 1 h. The final list of 23 descriptors used for the sensory profile of the blended tea samples described the appearance, odor/aroma, flavor/taste, texture/mouthfeel, and aftertaste attributes. The definitions of the attributes and the reference standards are listed in Table 2.

Table 2.

Definitions and reference standards for the appearance, odor/aroma, flavor/taste, texture/mouthfeel, and aftertaste attributes used in the descriptive analysis of blended tea products

Attribute	Descriptor	Abbreviation	Definition	Reference	Intensity score
Appearance	Brightness	BrightA	Intensity of brightness of corn silk tea	Corn silk tea Prep.: 30 mL Corn silk tea (Kwang Dong Pharmaceutical Co. Ltd., Seoul, Korea) poured into a white paper cup	6
	Transparent	TranspA	Intensity of transparency of lemonade	Lemonade (Sunkist Growers, Inc., Valencia, CA, USA)	7
	Greenness	GreenA	Intensity of greenness of sports drink	Gatorade with lemon flavor (Lotte Chilsung Beverage Co., Ltd., Seoul, Korea)	7
	Redness	RedA	Intensity of redness of Ceylon tea	Black tea Prep.: 30 mL canned black tea (Ceylon tea, Lotte Chilsung Beverage) poured into a white paper cup	10
	Yellowness	YellowA	Intensity of yellowness of apple jam diluted solution	Apple jam Prep.: 2 tbsp apple jam (Bokumjari Co., Ltd., Gyeonggi-do, Korea) diluted with 200 mL boiling water	7
Odor/aroma	Mint odor	MintO	Pleasant smells associated with mint	Mint flavor candy (Eclipse, Wm. Wrigley Jr. Co., Chicago, IL, USA)	8
	Flower odor	FlowerO	Aromatics associated with dried flowers	Dried floral leaves (Gypsophila, chamomile, lavender [Romantic November, online market, Busan, Korea])	10
	Jelly odor	JellyO	Sweet smells associated with fruit jelly	Fruit jelly (Haribo Goldbaren [Haribo GmbH & Co. KG, Bonn, Germany])	11
	Bitter odor	BitterO	Aromatics associated with herbal medicine	Herbal medicine Prep.: 50% herbal medicine (Ssanghwatang [Kwang Dong Pharmaceutical]) diluent	8
	Sweet odor	SweetO	Sweet smells associated with a honey solution	Honey Prep.: 2% honey from various flowers (Dongsuh Foods Co., Seoul, Korea) diluent	8
	Artificial sweeteners odor	ASweetO	Aromatics associated with artificial sweeteners, such as syrup-like medicine	Liquid medicine Prep.: 10% liquid medicine with strawberry flavor (Coldy S Syrup [Samil Co., Ltd., Seoul, Korea]) diluent	10
	Dried mugwort odor	DMugwortO	Aromatics associated with sauna bath, such as dried mugwort	Moxa (Ssuktteum Myeongga Co., Seoul, Korea)	7
Flavor/taste	Mild flavor	MildF	Mild flavor associated with corn silk tea	Corn silk tea (Kwang Dong Pharmaceutical)	10
	Artificial sweeteners taste	ASweetF	Flavor associated with artificial sweetener	Liquid medicine Prep.: 10% liquid medicine with strawberry flavor (Coldy S Syrup [Samil]) diluent	10
	Bitter taste	BitterF	Flavor associated with herbal medicine	Herbal medicine Prep.: 50% herbal medicine (Ssanghwatang [Kwang Dong Pharmaceutical]) diluent	9
	Sweet taste	SweetF	Flavor associated with honeyed water	Honey Prep.: 2% honey from various flowers (Dongsuh Foods) diluent	7
	Fruit taste	FruitF	Flavor associated with fruit tea	Apple jam Prep.: 2 tbsp Apple jam (Bokumjari) diluted with 200 mL boiling water	7
	Sour taste	SourF	Flavor associated with lemonade	Lemonade Prep.: 30% Lemonade (Sunkist Growers) diluent	9
	Astringency	AstrinF	Feeling of shriveled tongue associated with tannins	Burdock tea Prep.: 5 g dried burdock slices (Jirisangol, online market, Gyeonggi-do, Korea) infused with 200 mL boiling water for 2 min	7
Texture/mouthfeel	Powdery mouthfeel	PowderT	Mouthfeel of tongue coated with powder	Yakult Prep.: 100 mL Yakult (Maeil Dairies Co., Ltd., Seoul, Korea) diluted with 200 mL water	7
Aftertaste	Powdery mouthfeel	PowerAF	Mouthfeel of tongue coated with powder	Yakult Prep.: 100 mL Yakult (Maeil Dairies) diluted with 200 mL water	7
	Bitter flavor	BitterAF	Bitterness associated with herbal medicine in the mouth after swallowing tea	Herbal medicine Prep.: 50% herbal medicine (Ssanghwatang [Kwang Dong Pharmaceutical]) diluent	9
	Sweet flavor	SweetAF	Sweetness in the mouth after swallowing tea	Honey Prep.: 2% honey from various flowers (Dongsuh Foods) diluent	7

Evaluation procedure

The panelists evaluated the intensities of the sensory attributes of the six samples in individual booths, and the intensity of each attribute in the main experiment was rated on a 15-point categorical scale ranging from 1 = “weak” to 15 = “strong” (Lee et al., 2008a; 2008b; 2008c). The panelists were given a 10-min break after evaluating the first three samples to prevent sensory adaptations and rinsed their mouths with mineral water between samples. The panelists were not permitted to eat or drink anything other than water 1 h before the descriptive analysis session (Chung and Chung, 2008). The test was repeated three times on three consecutive days.

Consumer acceptability test

Sample presentation

Six blended tea samples were prepared for the descriptive analysis as well as for tasting to assess consumer acceptability. The teas were individually presented to each consumer in a white paper cup coded with a three-digit random number. The size of the serving cup and the serving temperature were the same as in the descriptive analysis. The samples were served monadically, and the serving order of the six samples followed a balanced Williams Latin squares design (Jaeger et al., 1998; Williams, 1949).

Evaluation procedure

Ninety-three consumers (38 males and 55 females, aged 25–38 years) were recruited from a community church in Seongnam-si, Gyeonggi-do, Korea. The consumers rated their overall liking and their liking of the appearance, odor, flavor, and texture on a nine-point hedonic scale that was developed by Peryam and Pilgrim (1957) to assess the acceptability of the food items. This nine-point scale has been the most commonly used scale to rate consumer preferences for food, which ranges from 1 = “dislike extremely” to 9 = “like extremely,” with 5 (midpoint) = “neither like nor dislike.” The study showed that scales with more choices, such as the nine-point scale, tend to be more discriminating than those with fewer choices, such as a seven-point scale (Jones et al., 1955).

The JAR scale provides a quick indication of the direction of attribute intensity, and the JAR scale data that are normally distributed around the center are indicative of an optimized level of a specific product attribute. However, consumers are not generally familiar with very specific attributes, so the use of the JAR scale should be limited to very simple attributes, such as sweetness or saltiness (Meullenet et al., 2007). In this study, only bitter taste and astringency were evaluated using the nine-point JAR scale, where 1 = “not enough,” 5 = “JAR,” and 9 = “too much.”

In addition, the subjects were presented with various possible sensory descriptions of each sample, and the reasons for their liking and disliking a sample were surveyed using CATA analysis. The sensory terms used in the CATA analysis were constructed based on the sensory characteristics defined in the description analysis, and in addition, the item “does not apply” could be added to the list. Finally, using the six tea samples, the subjects rated their levels of agreement associated with the degree of familiarity, willingness to try again, and willingness to recommend a sample to friends and family.

Mineral water was available to cleanse the palate before each test and between samples, and the consumers were given careful instructions on the test protocol, use of scale, tasting methods, and rinsing procedure.

Statistical analyses

Analysis of variance (ANOVA) using the generalized linear model was conducted to evaluate the samples and the effects of their interaction and was performed separately on the descriptive analysis data sets and ratings of consumer acceptability and attitude. Duncan’s multiple range test was used for post hoc comparisons between samples, with a significance level of p < 0.05. Principal component analysis (PCA) was conducted to determine the mean values for the appearance, aroma, flavor, and texture attributes to summarize the relationships between the blended tea samples and attributes.

Partial least-squares regression (PLSR) was conducted to determine the relationship between descriptive sensory data and consumer acceptability data. In PLSR, the mean values of the descriptive ratings were used as the X data set (dependent variables) and the mean values of the consumer acceptability ratings were used as the Y data set (explanatory variables; XLSTAT Tutorials, 2019b).

In the CATA questions, the frequency of use of each sensory descriptor was determined by counting the number of consumers who used that term to express the reasons for their liking or disliking each sample. To identify a difference between the samples based on the term used in the CATA questions, attributes that were checked by more than 20% of the respondents from each testing site are listed, and a correspondence analysis was conducted to obtain a bidimensional representation of the relationships between the samples and their descriptors (Dos Santos et al., 2015).

In the JAR scales, the consumers’ rating for overall liking and the JAR attributes are required. The penalties (or mean drops) are plotted against the percentage of consumers giving each response in a so-called “mean drop plot.” Attributes comprising a large percentage of consumers and penalties are found on the top right quadrant of the plot, which provides a quick summary of the most critical diagnostic points on a product (Pitts, 2009).

ANOVA was performed using SPSS 22.0 (IBM Corp., Armonk, NY, USA). PCA, PLSR, and penalty analyses were achieved using XLSTAT 2018 (Addinsoft, New York, NY, USA; https://www.xlstat.com/en/).

Results and discussion

Descriptive analysis

Sensory characteristics of blended teas

The application of a multivariate ANOVA to the descriptive data indicated that the attributes among the tea samples exhibited an overall significant difference (Wilks’ λ, p < 0.05). As shown in Table 3, for appearance characteristics, the brightness attribute was high for green tea (CMint_Green) and chamomile tea (P_Mind_Chamomile) as well as for W_Ginger, which comprised ginger, jujube, and root vegetables. The results showed that there is a significant similarity between green tea and yellow tea and their degree of transparency. Conversely, in the strongly red samples, such as Peach_Oolong and H_Rooibos, the brightness was evaluated as low. This behavior is in line with the degree of transparency in the sample appearance.

Table 3.

Mean intensities of the sensory attributes, consumer acceptability, and attitude scores of the six blended tea products

		CMint_Green	G_Breath	H_Rooibos	P_Mind_Chamomile	Peach_Oolong	W_Ginger	F ratio
Appearance	Attributes2
	BrightA	10.0a1	7.3b	5.3c	9.2a	7.1b	9.8a	11.793***
	TranspA	7.4ab	6.1bc	5.6c	7.6a	6.1bc	7.9a	4.016**
	GreenA	6.1a	6.0a	2.8c	5.7a	4.0b	3.5bc	13.820***
	RedA	2.4d	4.7c	11.4a	2.5d	6.5b	2.8d	87.034***
	YellowA	9.3a	5.1b	2.5c	8.3a	5.0b	5.9b	42.083***
	Odor/aroma
	MintO	6.0b	7.8a	4.2c	6.1b	4.7bc	3.4c	11.305***
	FlowerO	5.5a	7.1a	6.0a	5.7a	6.4a	3.9b	4.002**
	JellyO	2.2c	2.7c	8.8a	3.3c	5.5b	2.2c	29.991***
	BitterO	3.1bc	4.8a	2.0c	3.4b	2.5bc	3.7b	6.073***
	SweetO	2.2d	3.0 cd	9.2a	4.1c	6.7b	6.1b	33.218***
	ASweetO	1.8d	2.6 cd	9.1a	3.5bc	4.0b	4.0b	33.418***
	DMugwortO	5.4b	7.1a	2.9d	5.0bc	4.6bc	3.9 cd	8.189***
	Flavor/taste
	MildF	7.9b	7.9b	9.6a	8.1b	7.7b	8.3b	2.875**
	ASweetF	1.8b	2.4b	5.1a	2.6b	2.4b	2.5b	10.321***
	BitterF	4.2b	3.6bc	2.8c	2.4c	6.2a	3.1bc	8.901***
	SweetF	1.9b	2.4b	4.3a	2.7b	2.5bc	4.2a	12.571***
	FruitF	1.6c	1.7c	4.7a	2.3bc	2.9b	2.3bc	11.302***
	SourF	1.9b	1.9b	1.5b	1.9b	3.0a	1.7b	2.497**
	AstrinF	4.3b	2.9c	2.6c	2.8c	6.3a	2.8c	12.879***
	Texture/mouthfeel
	PowderT	4.3b	3.7b	3.2b	3.1b	5.7a	3.3b	5.932***
	Aftertaste
	PowerAF	4.0b	3.3b	3.0b	3.0b	5.8a	3.6b	6.938***
	BitterAF	4.4b	3.7bc	2.5c	2.4c	6.3a	3.6bc	7.319***
	SweetAF	1.7d	2.3 cd	3.9a	3.0bc	2.4 cd	3.8ab	8.289***
Acceptability3	Overall liking	4.8c	5.3bc	5.7ab	5.9a	5.6ab	4.2d	10.623***
	Appearance liking	5.5bc	5.3 cd	6.4a	5.6bc	5.9b	5.0d	9.185***
	Odor liking	5.7b	5.5b	6.0b	5.8b	6.8a	4.3c	18.333***
	Flavor liking	4.9c	5.3bc	5.7ab	5.9a	5.5abc	4.2d	9.458***
	Texture liking	4.9c	5.5b	6.1a	6.0a	5.3bc	4.9c	8.639***
Attitude	Familiarity	5.2bc	5.5b	5.8ab	6.2a	5.5b	4.9c	5.120***
	Try again	4.3bc	4.6ab	5.1a	5.2a	4.9a	3.7c	6.539***
	Recommend	4.2c	4.6bc	5.1ab	5.2a	4.9abc	3.6d	8.176***

¹Means within a row not sharing a superscript letter are significantly different (***p < 0.001, **p < 0.05, Duncan’s multiple range test)

²Descriptive analysis: 15-point categorical scale ranging from 1 = “weak” to 15 = “strong.”

³Consumer acceptability test: nine-point hedonic scale ranging from 1 = “dislike extremely” to 9 = “like extremely,” with 5 (midpoint) = “neither like nor dislike.”

For the aroma attribute, a dried flower odor was significantly higher (p < 0.05) in all samples, except for W_Ginger. In the other samples, except for G_Breath and P_Mind_Chamomile, those with added flowers, dried tea leaves, fruit flavors, and fruit peels appeared to mimic the characteristics of the dried flower aroma. When the sweetness of two samples—H_Rooibos and Peach_Oolong (which had fruit flavor added)—was compared among the six samples, H_Rooibos was evaluated to be significantly sweeter than Peach_Oolong (p < 0.05). This pattern was the same as that of the artificial sweetener, jelly, and mild flavor attributes. Peach_Oolong, which had added fruit flavor, presented more intense sweetness than the other samples; however, the intensity was significantly lower (p < 0.05) than in H_Rooibos. This result suggested that the sweetness intensity of Peach_Oolong was relatively weak because other characteristics, such as bitter and sour tastes, were significantly stronger in this tea than in the other samples. In addition, the bitterness and astringency intensities of Peach_Oolong were significantly higher than those in the other samples, and the mouthfeel of tongue coated with powder was strong. These characteristics are most likely a result of oolong tea, a semi-fermented green tea containing the most catechins, such as tannins (Choi 2002).

PCA indicated that principal components 1 (PC1) and 2 (PC2) explained 50.26% and 28.93% of the total variance, respectively (Fig. 1A). As shown in Fig. 1A, the PC1 dimension was defined mainly by the attributes of the artificial sweeteners, sweetness, fruitiness, mild flavor, and jelly-like (positive PC1 dimension), and the dried mugwort odor, green and yellow characteristics, brightness, bitterness odor, and mint odor (negative PC1 dimension). This dimension mostly explained the presence of artificial flavoring (CMint_Green, G_Breath, and H_Rooibos). Thus, H_Rooibos, which had significantly high scores (p < 0.05) for the artificial sweetener flavor, fruit flavor, jelly odor, and mild and sweet flavors, was loaded onto the positive PC1, and CMint_Green, G_Breath, and P_Mind_Chamomile, which had the dried mugwort aroma, bitter odor, mint odor, and appearance of brightness, transparency, and green and yellow characteristics, were loaded onto the negative PC1. The PC2 dimension was defined mainly by the mouthfeel of tongue coated with powder, a bitter and astringent flavor, and a sour flavor (positive PC2 dimension) and by the transparency and brightness attributes to a lesser extent (negative PC2 dimension). Thus, Peach_Oolong, which was rated significantly high (p < 0.05) for bitterness, astringency, mouthfeel of tongue coated with powder, and sour flavor, was loaded onto the positive PC2. According to the above results, the blended teas were largely differentiated by artificial fruit flavoring and, to a lesser extent, by the characteristics of their main ingredients.

Fig. 1.

PCA loadings and scores of the sensory attributes and the blended tea samples (A) and PLSR results indicating the relationship between the sensory characteristics of the six blended tea samples and consumer acceptability (B). OL, overall liking; APPL, appearance liking; ODOL, odor liking; FLL, flavor liking; and TXTL, texture liking

Consumer acceptability and sensory perception

The acceptability data from the consumers were analyzed using ANOVA, and significant differences were established using Duncan’s multiple range test (Table 3). P_Mind_Chamomile received the highest acceptability, with significant differences among the six samples (p < 0.05), although the acceptability scores for the teas with added fruit flavoring (H_Rooibos and Peach_Oolong) were also relatively high. In contrast, the acceptability for appearance was significantly higher in H_Rooibos, with its red and dark characteristics; for odor acceptability, Peach_Oolong, with its sweet and sour peach flavors, was rated significantly higher (p < 0.05). Similar patterns were observed in the evaluation of the degree of familiarity, willingness to try again, and willingness to recommend (Table 3). Chamomile is one of the oldest, most widely used, and well-documented medicinal plants in the world and has been recommended in a variety of healing applications (Astin et al., 2000; Srivastava et al., 2010). The results of the current study show that consumers prefer tea that is more familiar and has a nonirritating flavor.

Relationships between sensory descriptors and consumer acceptability

The PLSR result on the relationship between sensory descriptors and consumer acceptability (Fig. 1B) and the quality of the PLSR plot obtained by the first two PLS dimensions was 55.6%. From the PLSR results, we observed that all the acceptability items were strongly loaded in a positive direction (+) of t1, and among them, overall, flavor, odor, and appearance acceptability were strongly loaded together and were highly relevant. However, mouthfeel acceptability was loaded in a negative direction (−) of t2; therefore, it did not affect the overall acceptability.

The consumers preferred H_Rooibos, which has a lower intensity of bitter taste, astringency, and mouthfeel of tongue coated with powder compared to the other samples; thus, these attributes were believed to act as drivers of “dislike” for those consumers. In contrast, the characteristics closely loaded with acceptability items such as sweetness, artificial sweetener flavor, mild flavor, and fruit and jelly flavor, were believed to be the drivers of “like” for those consumers. In this regard, Peach_Oolong with strong bitterness and astringency intensities appears to have been highly evaluated for overall acceptability.

By the way, P_Mind_Chamomile with significant consumer preferences can be seen located on the opposite side of the acceptability items. It appears to have been influenced by its appearance and aroma acceptability. Nevertheless, the reason why the overall acceptability of this sample is significantly high seems to be related to “familiarity.” This is also related to the study of Tan et al. (2016) that shows that familiar preparations can effectively increase the consumers’ appetite for food.

CATA and JAR attributes

The reasons for liking and disliking each sample that were checked by more than 20% of the respondents are listed in Table 4. Focusing on Peach_Oolong, P_Mind_Chamomile, and H_Rooibos where the acceptability score was high, consumers generally picked the familiar flavor, fruit flavor, soft flavor, harmonious flavor, and sweet flavor for reasons of their liking for blended teas and otherwise mentioned the brightness of appearance, scent of flowers, and mint flavor. In contrast, they chose strong flavor, inharmonious flavor, bitter taste, astringency, and acrid flavor as the reasons for disliking the blended tea. In addition, samples with higher overall liking scores had a longer list of liking attributes and a shorter list of disliking attributes than those with lower overall liking scores. Additionally, subjects tended to check “none” for disliking reason of the preferred samples and “none” for liking reason of the unacceptable samples. These results correlate with some previous studies that have proven the usefulness of the CATA method, which delineates consumers’ perception of target food products, in sensory research (Ares et al., 2010; Dooley et al., 2010). The corresponding plot of the CATA questions enabled us to verify the quality of the analysis (Fig. 2), which was high (92.07% of the explained total inertia on the first two dimensions; Fig. 2A). According to the map of the analysis, the ideal products (P_Mind_Chamomile, H_Rooibos, and Peach_Oolong) should be relatively sweet, mild, and sour, fruit flavored, and harmonious flavored and exhibit appearance cues. Moreover, it should not be relatively too bitter, astringent, pungent, and strong or have a fermented flavor. W_Ginger, CMint_Green, and G_Breath are less than ideal because of their relative bitterness, sourness, and astringency. Conversely, based on the plot of samples that were less than ideal (W_Ginger and CMint_Green; Fig. 2B), we observed that they were associated with a bitter, astringency, strong, acrid, and pungent flavor. In addition, tea should not be relatively sweet, fruit flavored, and dark, for example; therefore, bitterness, astringency, pungency, and fermented flavor are product penalties for consumers.

Table 4.

List of attributes that the consumers liked and disliked about the six blended tea samples

	CMint_Green	G_Breath	Peach Oolong	P_Mind_Chamomile	H_Rooibos	W_Ginger
Liking	Brightnessa (22)b Familiar flavor (23) Mint flavor (20) Does not apply (22)	Flower flavor (20) Familiar flavor (20) Mint flavor (28) Mouthfeel of tongue coated with powder (26)	Flower flavor (23) Fruit flavor (40) Soft flavor (27) Familiar flavor (20) Harmonious flavor (22) Sweet odor (27)	Brightness (34) Transparency (22) Soft flavor (29) Familiar flavor (28) Harmonious flavor (25)	Appearance (25) Fruit flavor (37) Soft flavor (25) Harmonious flavor (21) Sweet odor (37)	Does not apply (37)
Disliking	Astringency (38) Inharmonious flavor (22) Bitter taste (21) Mouthfeel of tongue coated with powder (22) Pungency (20)	Strong flavor (20) Astringency (22) Does not apply (23)	Astringency (44) Bitter taste (22) Does not apply (27)	Does not apply (40)	Does not apply (34)	Strong flavor (31) Astringency (20) Inharmonious flavor (22) Pungency (21) Acrid flavor (20)

^aAttributes selected by more than 20% of the subjects in each testing site are listed

^bNumbers in parentheses are the percentage of respondents who checked the attribute

Fig. 2.

CATA plot; drivers of liking (A) and drivers of disliking (B) and the results of JAR analysis; mean drop plot of all the six blended tea products (C). ***p < 0.001

The descriptive statistics for the liking data and JAR variables are shown in Table 5, and the correlation matrix displays whether the JAR variables have either “low” or “high” impact on the overall liking and in which direction it would go (“too much” or “too little”). For blended tea products, JAR variables for bitterness and astringency have a low impact on the overall liking (p < 0.05), showing a significant difference from zero (Table 5); however, the correlation for bitterness and astringency was negative, meaning that the “too much” samples had a greater impact than the “too little” samples (Iserliyska et al., 2017). Table 5 also corresponds to the penalty analysis. Most of the consumers rated blended teas as being too weak in bitterness (54%) and astringency (48%). According to a previous study (Iserliyska et al., 2017), ≥70% of the responses should be within a JAR category to conclude that a specific attribute is at its optimal level; however, in the present study, none of the attributes were at the optimal levels (27.42% and 30.82%). The mean drops were calculated as the difference between the overall liking mean for the JAR levels and the “too much/too little” levels. This information is interesting because it shows how many liking points were lost from having a product with “too much” or “too little” for a consumer (XLSTAT Tutorials, 2019a). The penalty is a weighted difference between the means (i.e., the mean of liking for JAR minus the mean of liking for the two other levels taken together). For the bitterness dimension, the test results were not significant (p < 0.05). For the astringency dimension, the consumers penalized the product when they considered it to be too astringent. The mean drops plotted against the percentage of consumers giving responses on each sensory attribute of all six blended tea samples are presented in Fig. 2C. Mean drops are represented by plotting them on the y-axis against the percentage of respondents from a non-JAR group on the x-axis. Attributes located in the top right quadrant would be selected as those that need improvement (Meullenet et al., 2007). The perceived intensity of astringency should be decreased to increase overall liking (Fig. 2) because more than 20% of the respondents considered that the astringency was too strong.

Table 5.

Impact of the JAR variables on the overall liking (Spearman’s correlation coefficient) and penalty analysis

Variable	Correlation coefficienta	Level	%	Sum (overall)	Mean (overall)	Mean drop	Penaltyb
Bitterness	− 0.271	Not enough	54.48	1698.0	5.586	− 0.220
		JAR	27.42	821.0	5.366		0.144
		Too much	18.10	417.0	4.129	1.237
Astringency	− 0.254	Not enough	48.39	1503.0	5.567	− 0.032
		JAR	30.82	952.0	5.535		0.395**
		Too much	20.79	481.0	4.147	1.388***

^aImpact of the JAR variables on the overall liking (Spearman’s correlation coefficient). Values in bold are different from zero, with a significance level α = 0.05

^b***p < 0.001, **p < 0.05

According to these results, the blended tea products are highly astringent and thus highly penalized. It is expected that the results of this study will not only help to externally expand the tea products in the domestic market, where imports are high, but will also be used as a basis for the development of blended teas that Koreans prefer.