Triangular : Um estudo de caso com refrigerante tipo guaraná Tetrad vs . triangle test : A case study with Brazilian guarana soft drink

Discrimination sensory tests aim to identify if a difference between two similar stimuli is detected. In this study we compared the efficacy of Tetrads and Triangle tests in the difference detection between two samples of guarana soft drink, by means of the calculation of proportion of discriminators and thurstonian distance. Evaluated samples were produced by different syrup clarification methods (activated carbon and ionic exchange column). For each test 99 testers were used; Triangle test evaluated three samples, while Tetrad four samples, in complete randomized blocks. Only Tetrad test was able to detect significant difference between the samples (p<0.05), with a low proportion of discriminators and thurstonian Research, Society and Development, v. 9, n. 4, e166943049, 2020 (CC BY 4.0) | ISSN 2525-3409 | DOI: http://dx.doi.org/10.33448/rsd-v9i4.3049 3 distance inferior to perception limit, demonstrating that Tetrad test is more powerful and sensible than Triangle test.


Introdução
Sensory analysis is a crucial stage in the food development cycle as well as in quality control. In this sense, difference testing methods can be effectively used to check for sensory differences between two or more similar stimuli, both in-company and in consumer testing outside the industry environment (Dutcosky, 2013). An important application of difference Development, v. 9, n. 4, e166943049, 2020 (CC BY 4.0) | ISSN 2525-3409 | DOI: http://dx.doi.org/10.33448/rsd-v9i4.3049 4 tests is in quality control, with the objective of guaranteeing the homogeneity of products using sensory methods capable of detecting problems that may not be detected through instrumental measures (Muñoz, 2002).
The ability to differentiate two stimuli is the fundamental process underlying sensorybased responses (Lawless & Heymann, 2010). For example, a common question involving production is if a difference between two food samples with the same formulation, but processed at different facilities, might be detected by consumers. In such scenario, trained and experience panels can be used to check for this difference using methods such as the Triangle test (O'Mahony, 1986) or the most recent Tetrad test, which presents four samples to the assessorinstead of three as in the Triangle -, and he must group the samples in pairs based on their similarity (Gridgeman, 1956;Masuoka, Hatjopoulos, & Mahony, 1995). Tetrad and Triangle can be either directionalfocusing a specified sensory dimension-or used for general, unspecified differences. Besides, both have the same guessing probability of 1/3, which is the likelihood of success by chance (Lawless & Heymann, 2010). Several studies have proved that Tetrad test is more powerful and sensitivity than Triangle test as it can detect differences with more confidence and fewer assessors (Masuoka, Hatjopoulos, & Mahony, 1995;Delwiche & O'Mahony, 1996;Ennis & Jesionka, 2011;Ennis, 2012;Garcia, Ennis, & Prinyawiwatkul, 2012;Bi & O'Mahony, 2013;Ishii et al., 2014;Ennis & Christensen, 2014b). However, due to the evaluation of four samples, sensory fatigue, adaptation and psychological may occur tetrads test, especially when alcoholic beverages, tobacco and fragrances are tested (Ennis & Christensen, 2014a;Ennis, 2012).
Few studies in the sensory science literature have compared the efficiency of both tests. Garcia et al. (2012) performed a high scale study with 404 children comparing the efficiency of Triangle and Tetrad tests for detecting differences between two apple juices. Research, Society and Development, v. 9, n. 4, e166943049, 2020 (CC BY 4.0) | ISSN 2525-3409 | DOI: http://dx.doi.org/10.33448/rsd-v9i4.3049 5 Authors concluded that Tetrad was more powerful and presented a higher proportion of discriminants than Triangle.
When investigating potential ingredients or process modifications, typical specifications are made: significance level (probability of Type I error -α) at 5%; power at 80% (Type II error at 20%) and level of acceptable difference measures as the proportion of discriminators Pd of 20%. Based on published tables (Schlich, 1993;Lawless & Heymann, 2010), the sample size can also be determinate.
A problem with Pd is that it is method-specific and thus the same sensory difference will correspond to different proportion of discriminator according to the selected difference methods. In this sense, Thurstonian distance δ, a standardized measure of sensory difference, is of interest ASTM E2262−03 (ASTM International, 2014) since it is method-independent and can be easily estimated using sensory data (Ennis, 1993).
In this study we applied Tetrad and Triangle tests to investigate whether a different production process of a carbonated beverage might induce a perceptible sensory difference, a typical case of use of sensory evaluation in quality control. The production of carbonated beverages follows two stages: preparation of simple syrup and preparation of complex syrup.
For the simple syrup, water previously treated and sugar are mixed, warmed, filtered and cooled. Afterwards, this simple syrup is mixed with the juice concentrate and further additives (complex syrup), followed by dilution, carbonation and bottling (Barnabé & Venturini Filho, 2010). Sugar and syrup used for beverage production may influence beverage color due to pigments produced during sugar refining and, in turn, color may influence other sensory dimensions such as flavor and taste by means of halo effect (Lawless & Heymann, 2010). In order to inhibit these products from altering the color of the final product, pigments removal Research, Society and Development, v. 9, n. 4, e166943049, 2020 (CC BY 4.0) | ISSN 2525-3409 | DOI: http://dx.doi.org/10.33448/rsd-v9i4.3049 6 is commonly done by means of activated coal or ionic-exchange resins (Rodrigues et al., 2000). For the exposed, we aimed to investigate the efficiency of Triangle and Tetrad tests in detecting sensory differences between two carbonated beverages elaborated using different simple syrup clarification processes, activated carbon and ionic exchange resins.

Samples
Guarana soft drink, a very popular and highly consumed carbonated beverage in Brazil, was used as test samples. We used samples of the same brand but processed in two plants under different processes: plant A, which uses activated carbon for sugar syrup clarification, and plant B, which uses the ionic exchange column clarification method.
Activated carbon is a very versatile material used in many industrial applications and particularly for removal of components that may confer undesirable color, flavor and odor to the food products (Qureshi et al., 2008). On the other hand, ionic exchange resins in syrup clarification consists in two stages: an acrylic-based column, more hydrophilic and that not form irreversible links with syrup color compounds, and a styrene-based column which removes more compounds due to a higher adsorption capacity. The advantage of the use of these columns compared to the use of activated carbon is the low cost of material regeneration, which enables reuse, and the lower contact time needed (15-30 minutes), compared to activated carbon (2-4 hours) (Konen & Wilson, 1992). Table 1 presents the physicochemical characteristics of the beverages. Samples were packed in 350 mL aluminum doses. Research, Society and Development, v. 9, n. 4, e166943049, 2020 (CC BY 4. Color EBC 7 ± 0.3 6.9 6.8

Sensory tests
Sensory tests were performed at the Sensory Analysis Laboratory in the School of Food Engineering at the University of Campinas, Brazil. The laboratory is equipped with individual booths with lighting and temperature control.

Differences tests
Before the application of the tests, the number of assessors was calculated using the V-Power software (Ennis & Jesionka, 2011), with a proportion of discriminators ( Development, v. 9, n. 4, e166943049, 2020 (CC BY 4.0) | ISSN 2525-3409 | DOI: http://dx.doi.org/10.33448/rsd-v9i4.3049 8 In both tests 25 mL of the samples were served at 15°C in plastic cups coded with a three-digit random number, on complete balanced blocks. Judges were asked to sip a little amount of water to rinse the palate between samples.
In the Triangle test, assessors received the task: "You have three samples of guarana beverage. Taste them and pick the different one out" (Meilgaard et al., 1999). Conversely, in Tetrad the task was: "You have four samples of guarana beverage. Taste them and form two pairs according to the similarity between them" (Dutcosky, 2013).

Statistical Analysis
The proportion of hits (correct answers) were computed and the real proportion of discriminators, the standard error, the upper and lower confidence limits, and the estimate of Thurstonian distance (d') between samples were calculated using the V-Power software for Microsoft Excel® (Ennis & Jesionka, 2011).
Forty-eight individuals answered the Tetrad test correctly, while thirty-one correct answers were obtained in the Triangle test.
Starting with the values c=48 and n=99 for Tetrad test, and c=31 and n=99 for Triangle test, we could calculate statistics related to each test, as shown in Table 3. Tetrad test proved to be more sensitive to subtle differences and more powerful than Triangle test, as stated in the literature (Masuoka, Hatjopoulos, & Mahony,1995;Delwiche & O'mahony, 1996;Garcia et al., 2012), with a power (likelihood of finding a true difference) of 91.1% against only 3.7% in Triangle.
The Pd for error percent correspond to 5%, with which zerror=-1.645.
For Tetrads test, Thurstonian distance was calculated with the following data: