Vous êtes sur la page 1sur 17



DOI: 10.1111/j.1745-459X.2010.00278.x

ILANA FELBERG1,6, ROSIRES DELIZA2,3, ADRIANA FARAH4, ERNICA CALADO5 and CARMEN MARINO DONANGELO4 Embrapa Agroindstria de Alimentos Av. das Amricas, 29501, Guaratiba Rio de Janeiro, RJ, CEP 23020-470, Brazil
2 Embrapa Labex Europe Avenue Agropolis, F-34394 Montpellier, Cedex 5, France 3 INRA, UMR CSGA 17 rue de Sully, F-21065 Dijon, France 4 Instituto de Qumica Universidade Federal do Rio de Janeiro Av. Athos da Silveira Ramos 149, Bloco A, Ilha do Fundo Rio de Janeiro, CEP 21941-909, Brazil 5 Escola de Qumica Universidade Federal do Rio de Janeiro Av. Horcio Macedo, 2030 Bloco E, Ilha do Fundo Rio de Janeiro, CEP 21941-909, Brazil 1

Accepted for Publication November 18, 2009

ABSTRACT Coffee consumers (n = 60) tasted and rated samples of a new soycoffee beverage made from instant coffee, soymilk powder and sugar. Ingredient concentrations (independent variables) varied according to a 23central composite design for overall degree of acceptance. Data were analyzed by analysis of variance (ANOVA), least square difference and response surface methodology, followed by internal preference mapping (IPM) with cluster analysis.

Corresponding author. TEL: 5521-3622-9600; FAX: 5521-2410-1090; EMAIL: ilana@ ctaa.embrapa.br Journal of Sensory Studies 25 (2010) 226242. 2010 Wiley Periodicals, Inc.




ANOVA from the consumers acceptance data revealed that samples differed signicantly (P 0.05). Although soymilk content did not inuence signicantly the consumers acceptance in the tested range, IPM with cluster analysis indicated that at least part of the acceptance differences was based on the soy beverage consumption habit. The nal beverage formulation was evaluated cold and hot for overall acceptability (9-point structured hedonic scale) by 112 coffee consumers and the cold beverage reached a good acceptability mean score (6.2) among the participants.

PRACTICAL APPLICATIONS The consumption of soy products has been reported to reduce the risk of several diseases and a number of recent studies have found benecial health properties attributed to coffee. Considering the current consumer trend for healthier alternatives in food products, we decided to combine the health benets of these two important Brazilian commodities in a functional beverage. In order to optimize the formulation and maximize sensory acceptance, we performed consumers tests using response surface methodology. Internal preference mapping and cluster analyses were also applied to provide information on the variability of the consumer individual opinions and segment them in groups of similar preference criteria.

INTRODUCTION The consumption of soy products has been reported to reduce the risk of several diseases, including breast and prostate cancers, osteoporosis and heart diseases (Barnes et al. 2006; Messina et al. 2006). Among soy derivatives, soymilk has been gaining an increased interest from food industries during the last decade (Behrens et al. 2007). Soymilk is an excellent source of protein and essential fatty acids, is cholesterol-free and is relatively cheap in comparison with other food protein sources (Lozano et al. 2007). Although plain soymilk is not well-accepted in Western countries because of its beany avor and astringency (Al Mahfuz et al. 2004), soy beverages blended with fruit juices are a new generation of soy products in these countries and a convenient and healthy way to include soy protein in the consumer regular diet. Soy beverage sales in the U.S.A. have more than doubled since 2000, and soyfruit beverages have become an important market. Even though Brazil is the second world soybean producer, soymilk is practically unknown to most Brazilian consumers, and most people who try the product tend to present a negative



attitude toward soy milk, mainly for its undesirable sensory properties (Tashima and Cardello 2003; Behrens and Silva 2004). On the other hand, coffee is very much appreciated among Brazilian people (ABIC 2009). Brazil is the rst coffee bean producer and the second largest coffee consumer market in the world. Although coffee is usually consumed mainly for pleasure because of its taste and aroma (Cristovam et al. 2000; Moreira et al. 2000), a number of recent studies have found benecial health properties related to coffees antioxidant activity and the capacity to normalize blood pressure and glicemia among others (Farah 2009). The consumption of small amounts of caffeine is also known to produce a well-being and alertness sensation (Farah and Donangelo 2006; Higdon and Frei 2006; Farah 2009). The popularity and use of instant coffee has increased signicantly all over the world because of the easy and fast preparation and long shelf life (Prakash et al. 2000; Nogueira and Trugo 2003). Although instant coffee represents only 5.8% of the internal Brazilian coffee market, this percentage is growing at an annual rate of 8%, thus reecting the increasing interest in this coffee product (ABICS 2009). Understanding the consumer attitude toward food is critical to achieve a product success (Deshpande et al. 2008; Hein et al. 2008). Therefore, the optimization of a formulation is very important to maximize sensory acceptability. Response surface methodology (RSM) has been used to investigate sensory characteristics when some ingredients are tested simultaneously (Damsio et al. 1999; Deshpande et al. 2008). Usually, the acceptability data of samples obtained from hedonic scaling are expressed as mean scores. In this case, it is assumed that consumers acceptability criteria are homogeneous (MacFie 2007). However, consumer preference responses are often quite heterogeneous, and mean scores may not be representative of individual opinions. Internal preference mapping (IPM) is a statistical tool that allows examination of individual rating by consumers (Greenhoff and MacFie 1994; Yackinous et al. 1999). This technique is based on principal component analysis, cluster analysis and polynomial regression (Yackinous et al. 1999; Wajrock et al. 2008). In addition to providing information on the variability of the consumer individual opinions, IPM provides the possibility of segmenting consumers in groups of similar preference criteria. When combining preference mapping with samples formulated according to a factorial design, IPM shows which formulas are liked by consumers and also which characteristics are preferred as well (Yackinous et al. 1999). Considering the current consumer trend for healthier alternatives in food products, the possibility of combining the health benets of these two important Brazilian commodities in a functional beverage and the fact that there are no reports of a similar product in the literature, the objective of the present



study is to formulate and optimize a coffeesoy-based beverage using RSM and IPM statistical tools applied to consumer acceptability.

MATERIALS AND METHODS SoyCoffee Formula Ingredients Robusta instant coffee (Coffea canephora cv. Conillon) was provided by Cocam Cia de Caf Solvel e Derivados, Brazil; soymilk powder was provided by Olvebra Industrial S.A., Brazil and rened sugar (from sugar cane) was purchased from the local market. Consumer Tests Three consumers tests were carried out. The rst test (1) was performed with coffee consumers aiming at determining the optimal beverage formulation. The second and third tests (2 and 3) were also performed with coffee consumers with the objective of verifying the acceptance of the previously identied optimal formulation when served either cold (2) or hot (3). Consumers Acceptance Test 1 A 23 central composite design with six axial points and four central points, totaling 18 formulas, was used to determine the optimum concentration of instant coffee, soymilk powder and sugar (independent variables) in the soycoffee beverage. The ingredients concentration ranges were selected following manufacturers recommendations and previous experiments (Felberg et al. 2008) and varied from 1.32 to 4.68 g for instant coffee, from 5.98 to 11.02 g for soymilk and from 6.64 to 13.36 g for sugar, then added to 100 mL of spring water at room temperature to prepare the beverage, as shown in Table 1. The prepared beverages were kept refrigerated (8 2C) until being tested by consumers on the next day. The test was carried out in standardized sensory booths with a group of 60 consumers (29 females and 31 males) aged between 18 and 66 years, habitually consuming one or more cups of coffee per day, recruited among staff, students and visitors at Embrapa Food Technology Research Center, Brazil. Each consumer evaluated the acceptability of all the 18 formulations using a 9-point structured hedonic scale where, 1 = dislike extremely and 9 = like extremely. The sample beverages were presented at 8 2C in 50 mL plastic cups coded with three-digit numbers. The samples were presented in a monadic sequential order. A complete randomized design was used, and each assessor evaluated six samples per session throughout three sessions. The order of sample presentation



was balanced to prevent carryover effects (MacFie et al. 1989). Spring water and unsalted cracker were provided for mouth rinsing between samples. Subjects were asked to complete a questionnaire aiming to provide general information regarding their age, gender, occupation, monthly income and habitual consumption of coffee and soy beverages. Consumers Acceptance Tests 2 and 3 The consumer acceptance tests 2 and 3 were carried out with a total of 224 coffee consumers (112 for each test) randomly recruited in a supermarket in Rio de Janeiro, Brazil. The formula tested was the one with the highest acceptance in the previously described consumers test. For test 2, the coffeesoy beverage formulated was served cold (8 2C) and for test 3, hot (60 2C). The samples were evaluated for overall acceptability using a 9-point structured hedonic scale (1 = dislike extremely and 9 = like extremely) and intention to purchase using a 7-point structured hedonic scale (1 = denitely would not buy and 7 = denitely would buy). A dummy sample was used in both tests. The samples were presented in 50 mL plastic cups coded with three-digit numbers, and the order of presentation was balanced. Water and crackers were available for consumers to drink and clean the palate as they pleased. Statistical Analysis The software STATISTICA, version 8.0 for Windows (StatSoft, Inc., Tulsa, OK), was used for the experimental design and for RSM analysis. Analysis of variance (ANOVA) and Fishers test (LSD) were used to verify signicant differences among means (P < 0.05). The preference mapping and cluster analyses were performed using XLSTAT-MX (2005, Addinsoft, New York, NY) for Windows. For cluster analysis, we used Wards method and Euclidean distances.
TABLE 1. CODED LEVELS AND INGREDIENTS CONCENTRATIONS FOR CENTRAL COMPOSITE DESIGN OF THE SOYCOFFEE BEVERAGE (18 SAMPLES) Independent variable Coded levels -1.68 Concentration (g)* Instant coffee Soymilk powder Sugar -1.0 0 1 +1.68

1.32 5.98 6.64

2.00 7.00 8.00

3.00 8.50 10.00

4.00 10.00 12.00

4.68 11.02 13.36

* Added to 100 mL of spring water to prepare the beverage.



Sugar(L) Coffee(L) Coffee(Q) Sugar(Q) Soymilk (L) Soymilk (Q) Coffee (L)xSugar (L) Coffee (L)xSoymilk (L) Soymilk (L)xSugar (L)
-2.48404 -2.25668 1.695916 -1.05492 -.292237 -.255707 0. -6.6293


P = 0.05

FIG. 1. PARETO CHART SHOWING THE STATISTICAL SIGNIFICANCE OF THE INDEPENDENT VARIABLES (FORMULA INGREDIENTS) (L) = linear term and (Q) = quadratic term numbers near the boxes mean Students t values.

RESULTS AND DISCUSSION Consumer Tests to Establish Optimum Concentration of Ingredients for the SoyCoffee Beverage The acceptability scores assigned to the samples by 60 consumers were related to the concentration of ingredients. Fig. 1 shows a Pareto chart where boxes represent the Students t-values. The linear (L) and quadratic (Q) terms of the variables sugar and instant coffee signicantly inuenced the acceptance of the beverage as their boxes were located on the right side of the dashed line. The linear term of sugar was the one that most contributed to the model. Its positive value indicates that an increase in sugar concentration contributed to a higher acceptance of the beverage. This is probably because of the cultural inheritance of the Brazilian population with regard to the Portuguese traditional appeal for sweet products (Nestl 1990). In the case of instant coffee, its linear and quadratic terms had a signicant negative effect, indicating that a decrease in the coffee content within the limits tested in this study contributed to a higher acceptance of the beverage. On the other hand, neither the soymilk powder content nor the interaction between the variables inuenced signicantly the consumers acceptance in the studied range. Figure 2 shows the inuence of the concentration of the studied variables on the acceptance scores. As observed in Fig. 1, the highest consumer acceptance values were obtained for the highest sugar concentration and the lowest



> < < < <

5 5 4 3 2

B 14
13 12

Sugar (g)

11 10 9 8 7 6 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0

>5 <5 <4 <3 <2

Coffee (g)

FIG. 2. RESPONSE SURFACE (A) AND CONTOUR CURVE (B) FOR CONSUMER ACCEPTANCE (n = 60) AS A FUNCTION OF SUGAR AND INSTANT COFFEE CONCENTRATIONS AT 10% SOYMILK CONCENTRATION (Acceptance evaluated using a 9-point hedonic scale: 1 = disliked extremely; 9 = liked extremely).

instant coffee concentration, independently of soymilk concentration (1). The contour curve (2) indicates that there was a good distribution of the design experimental points. The dark region indicates that any point within this area represents a combination of instant coffee, soymilk and sugar that would result in a higher consumer acceptance. Table 2 presents the design matrix with the acceptability scores, approval indexes (consumer ratings >5.0) and rejection indexes (consumer ratings



TABLE 2. MEAN ACCEPTABILITY SCORES OF THE SOYCOFFEE BEVERAGES Sample number S18 S3 S11 S13 S16 S8 S2 S4 S9 S5 S10 S12 S15 S7 S14 S1 S17 S6 Coffee (g)* 3.00 2.00 2.00 1.32 3.00 3.00 3.00 3.00 4.00 4.00 3.00 2.00 3.00 2.00 4.68 4.00 3.00 4.00 Soymilk (g)* 8.50 7.00 10.00 8.50 11.02 8.50 8.50 8.50 10.00 7.00 8.50 10.00 5.98 7.00 8.50 7.00 8.50 10.00 Sugar (g)* 13.36 12.00 12.00 10.00 10.00 10.00 10.00 10.00 12.00 12.00 10.00 8.00 10.00 8.00 10.00 8.00 6.64 8.00 Acceptability score 5.8a 5.7ab 5.5abc 5.5abc 5.3abcd 5.2abcde 5.0bcde 4.9cdef 4.9cdef 4.7def 4.7def 4.7def 4.5efgh 4.1fgh 3.8gh 3.7h 3.7h 3.5h Approval index (%) 65 65 65 58 57 57 53 43 48 47 45 38 38 32 25 22 20 25 Rejection Index (%) 30 32 28 33 32 33 42 37 45 42 45 50 48 55 70 68 68 68

Values with different superscript letters within the same row are signicantly different by Fishers test (P 0.05). * Added to 100 mL of spring water to prepare the beverage. Evaluated using a 9-point hedonic scale varying from 1 = dislike extremely, 5 = neither like nor dislike, to 9 = like extremely.

<5.0). Hedonic ratings for the samples differed drastically. The most preferred ones scored (on average) from 5.0 (neither like nor dislike) to 6.0 (like slightly) on a 9-point hedonic scale. In general, soy products do not present high hedonic scores in the literature, being our results similar to those of other authors (Rosenthal et al. 2002; Felberg et al. 2004; Da Silva et al. 2007; Potter et al. 2007). Deshpande et al. (2008) considered 5 to be the lowest consumer rating limit of acceptability for optimizing formulations of a chocolateavored peanutsoy beverage, evaluated using a 9-point hedonic scale. When utilizing the same criteria, 39% of the formulated beverages presented consumer ratings 5.0 (Table 2). ANOVA results for acceptance data from all the consumers revealed that the samples differed signicantly in terms of acceptability, with a range of mean scores from 3.5 to 5.8 (Table 2). No signicant difference was observed among the most preferred samples (18, 3, 11, 13, 16 and 8). The approval index from those samples varied from 65% to 57%. The least accepted samples were 6, 17, 1, 14, 7 and 15. Both samples 1 and 6 (68% rejection index) had high



coffee concentration (4 g) and low levels of sugar (8 g) with 7 g and 10 g of soymilk content, respectively. Sample 17 (68% rejection index) had the lowest sugar content studied (6.6 g) and sample 14 with the highest rejection index of 70% presented the highest level of instant coffee (4.7 g). Based on the experimental design, the critical values of each one of the ingredients for the best predicted value of acceptability (5.9) were calculated and found to be 1.72, 9.66 and 14.07 g/100 mL of spring water for instant coffee, soymilk and sugar content, respectively. Taking into account that consumer preference responses are usually heterogeneous and averaged consumer data may not be representative of any individual opinion, preference mapping was applied to visualize the diverging liking directions of all the consumers on a map. Preference mapping showed that 48% of the variance could be explained by the rst three principal components, of which 30% was accounted by the rst preference dimension, 10% by the second and 8% by the third. Figure 3 presents the internal preference maps derived from the acceptance scores of the 60 consumers for the 18 soycoffee beverage samples dened by the rst two principal dimensions. Sample positions (S), consumers positions and the identied clusters can be seen in Fig. 3. Principal component 1 separated the beverages into three main groups. The rst group (samples 3, 11, 13, 18 and 16) was located at the top right quadrant and contained samples with low coffee and/or high sugar content. These results were similar to those reported by Geel et al. (2005) that related consumer preferences to sensory attributes of instant coffee. The authors called the consumers who preferred the coffee avor with lowest intensity and the highest sweetness as coffee blend lovers, Samples 15, 7, 12, 4, 2, 10, 9, 5 and 8 were included in a second group. Samples 2, 4, 8 and 10 were the center points of the design samples. Samples 17, 14, 6 and 1 made up the third group and were the least accepted ones. Sample 17 had the lowest level of sugar studied, sample 14 had the highest level of instant coffee, and samples 6 and 1 had high level of instant coffee and low level of sugar content. Although these results indicated that the majority of the participants did prefer the sweetest and the least intense coffee samples (Table 2, Fig. 3), we can see a small cluster that preferred samples with the opposite characteristics (Table 2, Fig. 3). Probably, they could be comparable to the coffee lovers described by Geel et al. (2005), who liked bitterer and more coffee avor character samples. Principal components 2 and 3 seemed not to separate consumers as well as principal component 1. Figure 3B revealed the existence of signicant differences in preference among consumers. Each point represents the end point of one consumer acceptance vector, and each vector can be visualized by drawing a line from



6 4 2 S17

Observations (axes F1 and F2: 40.33 %)

S15 S7 S14 0 S1 -2 -4 -6 -8 -10 -8 -6 -4 -2 0 2 S6 S4 S10 S5 S2 S9 S8 S12

F2 (10.40 %)

S16 S13

S18 S11 S3

F1 (29.93 %)

Variables (axes F1 and F2: 40.33 %)

50 41 48 147 55 21 301738 39 28

0,75 0,5


F2 (10.40 %)


0 -0,25


8 36 2 53 25 33 4


9 42 49

Cluster 1

-0,5 -0,75

20 Cluster 3


31 37 46 51 16 32 57 6 1 59 24 18 35 56 22 58 23 52 15 47 19 6043 27 44 3 11 10 54 26

Cluster 2

-1 -1 -0,75 -0,5 -0,25 0 0,25 0,5 0,75 1

F1 (29.93 %)



TABLE 3. MEAN ACCEPTABILITY* SCORES OF THE SOYCOFFEE BEVERAGES FOR EACH OF THE THREE PREFERENCE SUBGROUPS (CLUSTERS) Preference subgroups (clusters) cluster 1 Sample number S18 S13 S3 S11 S16 S8 S12 S15 S2 S4 S9 S10 S7 S5 S14 S17 S6 S1 (n = 31) 6.8a 6.8a 6.4a 6.3a 6.0ab 5.2bc 5.1bc 5.0c 4.9c 4.9c 4.5cd 4.5cd 4.5cd 4.4cde 3.5ef 3.5def 2.8f 2.8f cluster 2 (n = 20 ) 5.2ab 4.2bcde 5.2ab 5.0abc 5.0abc 5.4a 4.0bcdef 3.9cdef 4.8abcd 4.6abcde 5.6a 4.5abcde 3.9cdef 4.8abcd 3.5ef 2.9f 4.1bcdef 3.6def cluster 3 (n = 9) 3.8bc 4.2bc 4.1bc 4.1bc 3.7bc 5.0abc 4.7abc 4.4abc 5.6abc 5.6abc 4.6abc 5.8ab 3.3c 5.8ab 5.8ab 5.8ab 4.8abc 6.8a

Values with different superscript letters within the same row are signicantly different by Fishers test (P 0.05). * Evaluated using a 9-point hedonic scale varying from 1 = dislike extremely, 5 = neither like nor dislike, to 9 = like extremely.

the center of the circle. The length of the vectors indicates how much variance of the consumers acceptability scores is explained. The circle represents the t of the regression model (P < 0.05). All the 60 consumers were signicantly tted and included in each plot. The consumers were segmented and distributed around the various quadrants of the preference space. Cluster analysis was conducted and included later in the preference mapping in order to visualize the group of consumers who liked the samples in a similar way. Three clusters of consumers were found for overall acceptance. The number of consumers per cluster and the overall acceptance mean scores for each sample are shown in Table 3. ANOVA results for acceptance data from the consumers segmented in the three clusters revealed that the samples differed signicantly in terms of acceptability and a formulated soycoffee beverage that is agreeable to one group does not satisfy others (Table 3).



The consumers from cluster 1 (n = 31), corresponding to 52% of all the consumers, gave better scores and preferred the beverages with low coffee and/or high sugar content (3, 11, 13, 18 and 16), with acceptance mean scores 6.0 (like slightly) on a 9-point hedonic scale. The consumers in cluster 2 (n = 20) also liked some of the samples liked by the consumers in cluster 1. However, they gave lower acceptability scores. The preferred samples were 9, 8, 3, 18, 11, 16 and 5, with acceptance mean scores >5.0 (neither like nor dislike) on a 9-point hedonic scale. cluster 3 (n = 9), with only 15% of the consumers, liked products 1, 17, 14, 10, 5, 4 and 2, with acceptance mean scores >5.0. Those consumers had a very different pattern for liking and preferred the samples with low levels of sugar and/or high levels of instant coffee. It was not possible to characterize any of the three clusters based on the available demographic data recorded in the questionnaires such as income, educational level, gender, etc. However, we did nd differences in soy beverages and coffee consumption patterns of the three clusters. Although the consumers acceptability of the beverages obtained from hedonic scaling expressed as mean scores was not signicantly inuenced by soymilk concentration, the percentage of the participants who consumed soy beverages regularly varied considerably in a diminishing way in clusters 1, 2 and 3 (77%, 60% and 44% of the consumers, respectively). The opposite was observed with coffee consumption. About 52%, 69% and 77% of the consumers, respectively, consumed more than two cups of coffee a day, with 33% of cluster 3 consuming more than ve cups a day. Unfortunately, since we did not ask the consumers about their sweetness preferences, we were not able to investigate this issue when dividing them into subgroups. In terms of the methodological approach, it is noteworthy to comment that cluster analysis could have been thought of as a preliminary to RSM as the purpose of cluster analysis is to nd possible consumer segmentation. Once consumer segmentations are identied, the formulation and optimization of a new product in RSM may be based on liking means of a particular consumer cluster. However, such approach was not used in the present study because it would require a much higher number of participants in order to come out with a reasonable degree of freedom. Because our interest was to develop a beverage with a good sensory acceptance, but also with potential functional properties and therefore with a favorable concentration of functional ingredients, we concluded that the best formulation, combining both criteria, contained 2 g% of instant coffee, 10 g% of soymilk and 13 g% of sugar. The sugar content was established taking into account that 14 g% of sugar extrapolated the experimental design and that 13 g% of sugar was enough considering the total solid content of the beverage and health concerns related to sugar consumption.



COLD 25 20 Percentual (%) 15 10 5 0


4 5 6 Acceptance scores


Consumer Acceptance of the Selected SoyCoffee Beverage Formulation, Served either Cold or Hot The selected beverage formulations reached a good acceptability among the consumers when served cold, with a mean score of 6.2, i.e., between 6 (like slightly) and 7 (like moderately). Similar results were observed for fermented soy beverages, chocolate-avored peanutsoy and blueberrysoy beverages (Behrens et al., 2004; Deshpande et al. 2005 and Potter et al. 2007, respectively). Considering that coffee beverages are usually consumed hot by Brazilians, who are not familiar with cold coffee beverages, in addition to testing the cold beverage, we decided to evaluate the acceptance of the soycoffee beverage also when served hot. We chose the temperature of 60 2C, since according to Lee and OMahony (2002) and Brown and Diller (2007), this is usually the most appreciated temperature by consumers when drinking coffee and coffee beverages. However, as we can see in Fig. 4, the results for the hot beverage were not as good as for the cold beverage, with a mean acceptance score of 5.5 (between neither like nor dislike and like slightly on the 9-point hedonic scale). Despite the fact that a mixture of coffee and milk is a very popular beverage in Brazil, and it is usually consumed hot, the lower preference score for the hot product was expected because some soymilk volatile compounds are not well-accepted by Western populations (Al Mahfuz et al. 2004), and they can be more easily perceived in a hot beverage than in a cold version.



These compounds are especially aldehydes (Behrens et al. 2004). Additionally, all ready-to-drink soy beverages available in the Brazilian market are sold to be consumed cold. On the other hand, as it was mentioned before, considering that almost all coffee beverages are consumed hot in Brazil, this cold coffeesoy beverage may represent a new market niche for the Brazilian industry. CONCLUSIONS By using the RSM tool, we were able to dene a soycoffee formula. Acceptability of samples (obtained from hedonic data expressed as mean scores) was inuenced by sugar and instant coffee concentrations and not by the soymilk concentration. However, when IPM with cluster analysis was performed, we were able to distinguish three segments of consumers, and considering the demographic characteristic of the consumers for each cluster, part of the acceptance differences seems to be explained by the soy beverage consumption habit. The nal beverage formulation, served cold, reached a good acceptability among the participants and may represent a new market niche in Brazil. ACKNOWLEDGMENTS We thank the nancial support from Consrcio Brazileiro de Pesquisa e Desenvolvimento do Caf and Olvebra Industrial S.A (Eldorado do Sul, Brazil) and Cocam Cia de Caf Solvel e Derivados (SP, Brazil) for kindly offering the samples. REFERENCES ABIC. 2009. Estatsticas. Indicadores da Indstria. http://www.abic.com.br (accessed January 30). ABICS. 2009. Anlises Econmicas. http://www.abics.com.br (accessed January 30). AL MAHFUZ, A., TSUKAMOTO, C., KUDOU, S. and ONO, T. 2004. Chances of astringent sensation of soy milk during tofu curd formation. J. Agric. Food Chem. 52, 70707074. BARNES, S., PRASAIN, J., DALESSANDRO, T., WANG, C., ZHANG, H. and KIM, H. 2006. Soy isoavones. In: Nutritional Oncology, 2nd Ed. (D. Heber, G.L. Blackburn, V.L.W. Go and J. Milner, eds.) pp. 559571, Academic Press, San Diego, CA.



BEHRENS, J.H. and SILVA, M.A.A. 2004. Atitude do consumidor em relao soja e produtos derivados. Cin. e. Tec. Alimentos 24, 431439. BEHRENS, J.H., ROIG, S.M. and SILVA, M.A.A.P. 2004. Fermentation of soyamilk by commercial lactic cultures: Development of products with market potential. Acta Alimentaria 33, 101109. BEHRENS, J.H., VILLANUEVA, N.D.M. and SILVA, M.A.A.P. 2007. Effect of nutrition and health claims on the acceptability of soymilk beverages. Int. J. Food Sci. Technol. 42, 5056. BROWN, F. and DILLER, K. 2008. Calculating the optimum temperature for serving hot beverages. Burns 34(5), 648654. CRISTOVAM, E., RUSSELL, C., PATERSON, A. and REID, E. 2000. Gender preference in hedonic ratings for espresso and espressomilk coffees. Food Qual. Prefer. 11(6), 437444. DA SILVA, J.B., PRUDNCIO, S.H., FELBERG, I., DELIZA, R. and CARRO-PANIZZI, M.C. 2007. Acceptability of drinks prepared with different water-soluble soy extracts. Aceitabilidade de bebidas preparadas a partir de diferentes extratos hidrossolveis de soja. Pesqui. Agropecu. Bras. 42(12), 17791784. DAMSIO, M.H., COSTELL, E. and DURN, L. 1999. Optimising acceptability of low-sugar strawberry gels segmenting consumers by internal preference mapping. J. Sci. Food Agric. 79(4), 626632. DESHPANDE, R.P., CHINNAN, M.S. and MCWATTERS, K.H. 2005. Nutritional, physical and sensory characteristics of various chocolate-avored peanutsoy beverage formulations. J. Sens. Stud. 20(2), 130 146. DESHPANDE, R.P., CHINNAN, M.S. and MCWATTERS, K.H. 2008. Optimization of a chocolate-avored, peanutsoy beverage using response surface methodology (RSM) as applied to consumer acceptability data. Lebensm.-Wiss. Technol. Food Sci. Technol. 41(8), 1485 1492. FARAH, A. 2009. Coffee as a speciality and functional beverage. In Functional and Speciality Beverage Technology (P. Paquin, ed.) pp. 350375, Woodhead Publishing Limited, London. FARAH, A. and DONANGELO, C.M. 2006. Phenolic compounds in coffee. Braz. J. Plant Physiol. 18(1), 2336. FELBERG, I., DELIZA, R., GONALVES, E.B., ANTONIASSI, R., FREITAS, S.C. and CABRAL, L.C. 2004. Bebida mista de extrato de soja integral e castanha-do-Brasil: Caracterizao fsico-qumica, nutricional e aceitabilidade do consumidor. Alimentos E Nutrio 15(2), 163 174. FELBERG, I., DELIZA, R., FARAH, A., SUNDFELD, E., CALADO, V. and DONANGELO, C.M. 2008. Coffeesoy based beverage: Product formu-



lation and consumer acceptance. In 22nd International Conference on Coffee Science/ASIC in 1419 September, Campinas, SP, Brazil. GEEL, L., KINNEAR, M. and DE KOCK, H.L. 2005. Relating consumer preferences to sensory attributes of instant coffee. Food Qual. Prefer. 16, 237244. GREENHOFF, K. and MACFIE, H.J.H. 1994. Preference mapping in practice. In: Measurement of Food Preferences (H.J.H. MacFie and D.M.H. Thomson, eds.) pp. 137166, Blackie Academic and Professional, London. HEIN, K.A., JAEGER, S.R., TOM CARR, B. and DELAHUNTY, C.M. 2008. Comparison of ve common acceptance and preference methods. Food Qual. Prefer. 19(7), 651661. HIGDON, J.V. and FREI, B. 2006. Coffee and health: A review of recent human research. Crit. Rev. Food Sci. Nutr. 46(2), 101123. LEE, H.S. and OMAHONY, M. 2002. At what temperatures do consumers like to drink coffee? Mixing methods. Journal of Food Science, 67(7), 27742777. LOZANO, P.R., DRAKE, M., BENITEZ, D. and CADWALLADER, K.R. 2007. Instrumental and sensory characterization of heat-induced odorants in aseptically packaged soy milk. J. Agric. Food Chem. 55(8), 3018 3026. MACFIE, H.J.H. 2007. Consumer-Led Food Product Development (H.J.H. MacFie, ed.), Consultant in Sensory and Consumer Science, U.K. MACFIE, H.J.H., BRATCHELL, N., GREENHOFF, K. and VALLIS, L.V. 1989. Designs to balance the effect of order of presentation and rst-order carry-over effects in hall tests. J. Sens. Stud. 4, 129148. MESSINA, M., MCCASKILL-STEVENS, W. and LAMPE, J.W. 2006. Addressing the soy and breast cancer relationship: Review, commentary, and workshop proceedings. J. Natl. Cancer Inst. 98(18), 12751284. MOREIRA, R.F.A., TRUGO, L.C. and DE MARIA, C.A.B. 2000. Volatile components in roasted coffee. Part II. Aliphatic, alicyclic and aromatic compounds. Qum. Nova 23(2), 195203. NESTL. 1990. Uma doce herana. In O Doce Brasileiro. (Nestl Industrial e Comercial Ltda., ed.) p. 8, Nestl Brasil Ltda, So Paulo, Brazil. NOGUEIRA, M. and TRUGO, L.C. 2003. Chlorogenic acid isomers, caffeine and trigonelline contents in Brazilian instant coffee. Cin. e. Tec. Alimentos 23(2), 296299. POTTER, R.M., DOUGHERTY, M.P., HALTEMAN, W.A. and CAMIRE., M.E. 2007. Characteristics of wild blueberrysoy beverages. Lebensm.Wiss. Technol. 40, 807814.



PRAKASH, M., RAVI, R., SARVAMANGALA, G.K. and RAJALAKSHMI, D. 2000. Sensory proling and product positioning of roasted and ground (brew) coffee and soluble (instant) coffee with and without added avor. J. Sens. Stud. 15(1), 101117. ROSENTHAL, A., DELIZA, R., CABRAL, L.M.C., CABRAL, L.C., FARIAS, C.A.A. and DOMINGUES, A.M. 2002. Effect of enzymatic treatment and ltration on sensory characteristics and physical stability of soymilk. Food Control. 14(3), 187192. TASHIMA, E.H. and CARDELLO, H.M.A.B. 2003. Perl sensorial de extrato hidrossolvel de soja (Glicine max L. Merril) comercial adoado com sacarose e com sucralose. B. CEPPA. 21(2), 409428. WAJROCK, S., ANTILLE, N., RYTZ, A., PINEAU, N. and HAGER, C. 2008. Partitioning methods outperform hierarchical methods for clustering consumers in preference mapping. Food Qual. Prefer. 19(7), 662 669. XLSTAT-MX. 2005. Users Guide, Version 7.5.3., Addinsoft Inst. Inc., New York. YACKINOUS, C., WEE, C. and GUINARD, J.-X. 1999. Internal preference mapping of hedonic ratings for Ranch salad dressings varying in fat and garlic avor. Food Qual. Prefer. 10(45), 401409.