Comparison of Rate-All-That-Apply (RATA) And: Descriptive Analysis (DA) For The Sensory Profiling of Wine

AJEV Papers in Press. Published online September 7, 2017.
American Journal of Enology and Viticulture (AJEV). doi: 10.5344/ajev.2017.17052

AJEV Papers in Press are peer-reviewed, accepted articles that have not yet been published in a print issue of the journal
or edited or formatted, but may be cited by DOI. The final version may contain substantive or nonsubstantive changes.
1 Research Article
2 Comparison of Rate-All-That-Apply (RATA) and
3 Descriptive Analysis (DA) for the Sensory Profiling of Wine
4 Lukas Danner,1 Anna M. Crump,1 Alexander Croker,1 Joanna M. Gambetta,1
5 Trent E. Johnson,1 and Susan E.P. Bastian1*
6 1
The University of Adelaide (UA), School of Agriculture, Food and Wine, Waite Research Institute, PMB
7 1, Glen Osmond, South Australia 5064 Australia.
8 *Corresponding author (sue.bastian@adelaide.edu.au; tel: +61 08 8313 7116)
9 Acknowledgments: The authors thank all the consumers and sensory panelists who participated in both
10 studies, Hélène Bossan for her help and assistance during the descriptive analysis and consumer trial, Jun
11 Niimi for helpful discussions, and Anne Hasted of Qi Statistics for valuable training. The authors are
12 grateful to members of the Australian wine industry for their support and donation of numerous wines.
13 This research was funded by Australian grapegrowers and winemakers through their investment body,
14 Wine Australia, with matching funds from the Australian Government (UA1203) and the South
15 Australian River Murray Sustainability Program (SARMS).
16 Manuscript submitted Jun 13, 2017, revised Aug 6, 2017, accepted Aug 11, 2017
17 Copyright © 2017 by the American Society for Enology and Viticulture. All rights reserved.
18
19 Abstract: The aim of this work was to investigate how useful the Rate-All-That-Apply (RATA) method
20 with naïve consumers is to profile a wide range of wines and how the sensory profiles obtained compare
21 to those of classical descriptive analysis (DA). For this purpose, two studies were conducted. Study 1
22 presents preliminary work comparing the discrimination ability of RATA, undertaken by 84 naïve
23 consumers, with a traditional DA using 11 trained panelists, in which the vocabulary lists remained the
24 same across methods and assessment was based on a set of 6 red table wines (6 different varieties). Study
25 2 aimed to further elucidate the discrimination ability of RATA using 71 naïve consumers compared to
26 traditional DA. It expands on Study 1 by increasing the number of samples assessed (12 white table
27 wines, 6 varieties) and varying the vocabulary between methods (RATA used a generic white wine
28 attribute list vs panel generated attribute list for the DA). In addition, similarity of sample configuration in
29 the sensory space between RATA and DA was assessed by means of Multiple Factor Analysis (MFA)
30 and RV coefficients.
1
31 The results of both studies showed high similarities in sample discrimination ability (in terms of number
32 of attributes significantly discriminating amongst samples) between RATA and DA. Furthermore, the
33 MFA indicated high agreement in sample configuration between RATA and DA, reinforced by highly
34 significant RV coefficients of 0.97 for Study 1 and 0.92 for Study 2, respectively. Overall, this supports
35 the trend towards more consumer centric approaches for sensory profiling and suggests that RATA could
36 be a valid, accurate and rapid addition to existing profiling methods used for wine.
37 Key words: consumer, descriptive analysis, multiple factor analysis (MFA), rapid sensory profiling,
38 Rate-All-That-Apply (RATA), sensory characterization
39 Introduction
40 Describing and quantifying the complex sensory properties of foods and beverages can be an elaborate,
41 time-consuming and expensive task. Traditionally, sensory evaluation of food products has been divided
42 into two method groups: analytical tests, where trained individuals or panels are used to objectively
43 evaluate sensory properties, and affective tests, where consumers are used to determine product
44 acceptance and preference (Stone and Sidel 1993, Lawless and Heymann 2010). Developed in the 1970’s,
45 Quantitative Descriptive Analysis (QDA®) (Stone et al. 1974) and its subsequent variations (for the
46 purpose of this study referred to as DA), have remained the most sophisticated sensory methodology to
47 quantitatively profile products (Lawless and Heymann 2010). However, due to time, economic and
48 logistical constraints, the methodology has limited applications for industry and product development.
49 Highly trained DA panelists cannot provide preference ratings indicative of regular consumers and the
50 wider market, leading to the requirement of further testing and greater expense to collect additional
51 product information (Lawless and Heymann 2010). Current practices within the wine industry often rely
52 upon experts (winemakers and marketers) to evaluate and communicate a products’ attributes to the
53 consumer. However, with consumer expectations and global competition increasing, it is essential for
2
54 wine producers to adapt consumer orientated approaches to understand drivers of liking and better engage
55 consumers.
56 Contradictory to the traditional view, current research suggests that naïve consumers can be utilized to
57 assess and quantify the sensory attributes of food and beverage products when given a suitable
58 methodology (Moussaoui and Varela, 2010). A wide range of methods have been proposed as suitable for
59 naïve consumers to profile a variety of products, including Napping®, Free Choice Profiling, Flash
60 Profiling, Check-All-That-Apply (CATA), and Rate-All-That-Apply (RATA) (Moussaoui and Varela
61 2010, Valentin et al. 2012, Delarue 2015). These methods tend to be rapid, easily understood by
62 consumers, flexible and potentially economical in their application. In addition, previous studies have
63 shown that rapid methods (e.g. CATA and RATA) are not likely to bias hedonic ratings (Jaeger et al.
64 2016, Schouteten et al. 2016) and can lead to increased consumer engagement and greater hedonic
65 discrimination of products (Jaeger and Ares 2015).
66 While several studies have shown the potential of rapid methods to profile food products (e.g., Ares,
67 Bruzzone, et al. 2014, Giacalone and Hedelund 2016, Oppermann et al. 2016), few have investigated their
68 applicability to complex commercial products such as wine (Sáenz-Navajas et al. 2017). Wines comprise
69 hundreds of volatile aroma compounds and a diverse range of chemical constituents that potentially
70 contribute to the wine aroma and flavor profiles (Thorngate 1997). As well as being a natural product,
71 grape variety, variations in geographical growing region, production and processing techniques further
72 add to the complexity of individual products and the perceived differences in aroma profiles observed
73 between styles.
74 CATA has previously been used to profile dairy, beer, peanuts, crackers and fruit products (Ares et al.
75 2010, Reinbach et al. 2014, Jaeger et al. 2013), however, the methodology is less suited for characterizing
76 products with subtle differences, such as wine (Ares and Jaeger 2015). The binary response of the CATA
3
77 format does not allow a direct measure of intensity. Therefore, CATA seems most suitable for
78 discriminating between products that differ widely in their sensory attributes, rather than those which
79 share many similarities in their sensory profile and possess more subtle differences in intensities for
80 attributes present across products.
81 The RATA method is a logical extension of CATA, where assessors are not only required to select
82 attributes applicable to a product, but additionally indicate the perceived intensity of these sensory
83 attributes using a rating scale, which may help overcome the limitations identified in the CATA method.
84 Previous studies have shown that compared to CATA, RATA can offer greater statistical power, requires
85 fewer samples for configurational stability and even if RATA is analyzed as binary data, it leads to an
86 increased use of terms by the assessor and also a higher percentage of discriminating attributes. (Ares,
87 Bruzzone, et al. 2014).
88 In terms of the most suitable data analysis for RATA, contradictory results were found. The studies by
89 Ares et al. (2014) and Meyners et al. (2016) found similar results in regards to discrimination ability and
90 configurational similarity independent of whether data were analyzed as binary or intensity data, whereas
91 the results of Oppermann et al. (2016) indicate that the incorporation of intensity ratings can result in
92 superior discrimination ability. Although within-assessor reproducibility when using RATA with semi-
93 trained individuals is only moderate, at a panel level RATA can show very good reproducibility
94 (Giacalone and Hedelund 2016). Oppermann et al. (2016) found that RATA with naïve consumers
95 provided similar results compared with those obtained by DA when describing unfamiliar emulsions.
96 However, to date, only a very limited number of studies investigated RATA for the assessment of aroma
97 and flavor nuances in complex commercial products, such as wine. Therefore, opportunity exists to
98 explore the discriminative ability of the RATA method compared to other quantitative sensory profiling
99 methodologies, such as DA with trained panelists, for the evaluation of a complex product such as wine.
4
100 The purpose of this paper is to describe the profiling ability and suitability of the RATA method for wine
101 assessment in comparison to conventional DA with highly trained panelists, for industry and research
102 applications. The project focused on two specific research questions:
103 RQ1: Investigate whether RATA responses analyzed as intensity data will result in better discrimination
104 of samples compared to RATA responses analyzed as binary data (disregarding the intensity ratings and
105 only considering whether an attribute was selected or not), and compare the outcomes to those of previous
106 studies (Ares, Bruzzone, et al. 2014; Meyners et al. 2016; Oppermann et al. 2016) where contradictory
107 results were found.
108 RQ2: Investigate the discrimination ability and configuration similarity of the RATA method compared to
109 DA, when a) the same attribute list is used for RATA and DA, and b) a generic attribute list based on the
110 Wine Aroma Wheel© is used for RATA.
111 The research was conducted in two parts: Study 1 refers to preliminary work comparing the
112 discrimination ability of RATA undertaken by consumers compared to traditional DA methodology, in
113 which the vocabulary lists remain the same for both methods, and assessment is based on a small set of 6
114 red table wines (6 varieties). Study 2 aims to further elucidate the discrimination ability of RATA using
115 naïve subjects compared to traditional DA methodology. It expands on Study 1 by increasing the number
116 of samples assessed (12 white table wines, 6 varieties) and varying the attribute lists between
117 methodologies using a generic attribute list for RATA while the DA panel generated their own list.
118 Methods and Materials
119 All studies were performed in accordance with the ethical guidelines for scientific research at the
120 University of Adelaide and approved by the human ethics committee (approval numbers: H-2013-048 and
121 H-2016-037). Participants gave written informed consent prior to the tasting.
5
122 Study one.
123 Wine Samples.

124 Commercially available, South Australian red and rosé table wines were sourced from the Riverland wine
125 region and assessed by a panel of 12 industry experts (Parr et al. 2002) for quality and suitability for the
126 overarching consumer project i.e., samples which displayed a wide range of sensory attributes and were
127 sound. A subset of 17 wines were chosen for sensory evaluation by a trained DA panel, with a further
128 subset of 6 wines displaying the most diverse sensory attributes and greatest potential for consumer tests
129 as rated by the trained panel, included in the RATA assessment (Table 1). Wines were stored at 15°C
130 prior to all sensory experiments, and equilibrated at room temperature (22-23°C) for one hour prior to
131 serving.
132 Descriptive Analysis.

133 Eleven panelists (8 female, 3 male; average age 42) were recruited on the basis of their availability and
134 previous participation on wine DA panels, to assess the sensory attributes of 17 red and rosé wines
135 (including the 6 wines used in the RATA study) using a consensus training method. Prior to formal
136 assessment, panelists underwent 12 hours of training (comprising 6 x 2 h sessions held over 3 weeks)
137 involving detection, identification, evaluation, and intensity rating of red and rosé wine aroma and flavor
138 attributes. The training sessions also included two practice evaluation sessions conducted in sensory
139 booths, under the conditions used during formal assessment. Data were evaluated after each practice
140 evaluation to assess panelist-by-sample interactions, to determine when formal assessment should
141 commence. The panel agreed upon 20 aroma, 19 flavor and 4 mouthfeel attributes, (with the complete list
142 of attributes provided in Supplementary Table 1) to profile and differentiate the wines. Three formal, 2
143 hour assessment sessions were held over a 2 week period, with all 17 wines presented in each session,
144 such that all wines were assessed in triplicate. Wine aliquots (30 mL) were presented in covered, black
145 215 mL stemmed ISO wine tasting glasses, using a randomized presentation order. Panelists assessed
146 each wine in an isolated tasting booth at 22-23 °C, and rated the intensity of each sensory descriptor using
6
147 a 15 cm line scale with anchor points of “low” and “high” placed at 1.5 cm and 13.5 cm on the scale,
148 respectively. Sensory reference standards were developed during training sessions and provided to
149 panelists at each formal assessment session in covered, opaque black glasses (Supplementary Table 1).
150 Panelists were required to reacquaint themselves with the sensory reference standards and definitions of
151 attributes several times during each evaluation. Distilled water and crackers were provided as palate
152 cleansers, and panelists were required to have a 1 min break between samples and a 5 minute break every
153 four samples. Sample data were collected using Red Jade software (2016, Redwood City, USA).
154 Rate-All-That-Apply.
155 RATA assessments of the subset of 6 wines was conducted with 84 wine consumers (aged 18-54 years,
156 68% female and 88% of whom consumed wine once per fortnight), recruited through email and social
157 media advertising based on the following inclusion criteria: being of legal drinking age (i.e., over 18 years
158 of age), consume red wine at least once per month and have not completed any formal wine education
159 courses. The 43 attributes developed by the DA panel were used for the RATA assessment
160 (Supplementary Table 1), with a 7-point intensity scale, anchored from “extremely low” to “extremely
161 high” used to discriminate between the perceived intensity of sample attributes. There is no published
162 consensus regarding the optimum number of points for a RATA intensity scale, previous studies have
163 used 3, 5, 7, 9 and 15- point scales (Meyners et al. 2013, Reinbach et al. 2014, Ares et al. 2014, Franco-
164 Luesma et al. 2016, Giacalone and Hedelund, 2016, Oppermann et al. 2016, Reinbach et al. 2014).
165 However, it has been suggested that the smallest scale possible should be used to maintain rapidity and
166 simplicity for assessors (Ares, Bruzzone, et al. 2014). It was decided that a seven-point intensity rating
167 scale would allow sufficient discrimination between the samples while at the same time being simple and
168 easy for participants to use. Terms were separated by sensory modality, and randomized within modality
169 for each participant and each sample, in a balanced presentation order. All assessors were provided with
170 an attribute list providing a short explanation of the attribute (Supplementary Table 1). Samples (30 mL)
171 were served in a sequential, monadic order, in clear 215 mL ISO tasting glasses. Distilled water and
7
172 crackers were provided as palate cleansers, and panelists were required to have a 1 min break between
173 each of the 6 samples. Data were collected using Red Jade software (2016, Redwood City, USA).
174 Study two.
175 Wine Samples.

176 Twelve commercially available white wines, comprising 6 varieties (Chardonnay, Gewürztraminer,
177 Riesling, Sauvignon Blanc, Semillon and Viognier) and encompassing a range of flavor profiles and price
178 points, (Table 1) were selected for Study 2. Wine samples were stored under the same conditions as
179 described for Study 1.
180 Descriptive Analysis.

181 DA was conducted using the same method outlined in Study 1. The tasting panel, comprising six females
182 and five males (average age 40), underwent 16 hours of training over eight two hour sessions. The wines
183 were presented together with 9 other white wines of the same varieties not being part of the RATA study,
184 under the same conditions as Study 1. Formal assessment of the wines occurred in duplicate over four
185 sessions (twelve samples per session). Panelists rated 12 aroma, 16 flavor, and 2 mouthfeel attributes
186 (Supplementary Table 2) while holding the wine in mouth and after expectoration. A one minute break
187 between samples and 5 minute break between each group of six samples was enforced.
188 Rate-All-That-Apply.
189 Terminology and Survey Design.
190 RATA survey terminology was adapted from the Wine Aroma Wheel© developed at U.C. Davis by Ann
191 C. Noble (1987) which was used as a base. Throughout 3 focus group discussions, each comprising 12
192 regular wine consumers , and the feedback of 3 University of Adelaide wine experts (Parr et al. 2002) the
193 terms were further reduced making it suitable for consumers and at the same time covering the expected
194 sensory space, resulting in a final list of 52 terms. (Supplementary Table 3).
8
195 Assessment Sessions.
196 A total of 71 assessors (aged 21-65 years old, 56% female) were recruited based on the following
197 inclusion criteria: being of legal drinking age (i.e., over 18 years of age), consume white wine at least
198 once per month and have not completed any formal wine education courses.
199 Upon arrival, a 10 minute induction was conducted to introduce consumers to the procedure and explain
200 the assessment requirements. All assessors were provided with an attribute list (Supplementary Table 3)
201 to familiarize with the terms and their definitions. Wines were presented under the same testing
202 conditions as Study 1. A 5 minute break was enforced after 6 samples had been assessed.
203 Statistical analysis.
204 Data obtained for both Study 1 and Study 2 were analyzed in the manner reported below.
205 Descriptive analysis.
206 Panel performance was checked using ANOVAs with regards to attribute ratings considering panelist
207 (random factor), samples and replicate (fixed factors) including all their second order interactions. Two-
208 way ANOVAs with samples as fixed factors, and panelist as random factors, including interactions were
209 performed on all rated attributes to identify attributes significantly discriminating between samples.
210 RATA analyses.
211 The discrimination ability between RATA data analyzed as frequency (RATA frequency, considering
212 only the frequency of attribute selection) and analyzed as intensity ratings (RATA intensity, considering
213 the intensity rating on the 7-point scale) was compared. For this purpose Cochran’s Q tests were
214 performed for each sensory attribute treating the data as binary data (0 – attribute not used, 1 – attribute
215 used irrespective of intensity rating) and two-way ANOVA (sample as fixed and panelist as random
9
216 factor) treating the data as continuous data (a non-selected attribute was treated equivalent to “not
217 perceived” and assigned as intensity = 0, (Meyners et al. 2016, Oppermann et al. 2016)). As the results for
218 both studies (see section: Comparison of discriminative ability between RATA responses analyzed as
219 frequency and intensity data) indicated that the discrimination ability of the RATA intensity approach is
220 superior over the RATA frequency approach, the RATA data was considered as intensity data for all
221 analyses reported herein.
222 To compare the discriminative ability between DA and RATA methods, only those attributes shared by
223 the two methods were compared within Study 1 and 2. This is particularly relevant for Study 2 in which a
224 generic attribute list was used for RATA assessment, while the DA panelists generated their own set of
225 terms.
226 To assess the configurational similarity of product sensory spaces between DA and RATA, Multiple
227 Factor Analysis (MFA) was performed, using only attributes significantly differentiating between
228 samples. Configurational similarity was visually evaluated through sample and attribute representation in
229 the sensory space. In addition regressor vector (RV) coefficients were calculated to statistically quantify
230 the configuration similarity. This approach has been used successfully in previous studies to address
231 comparable research questions (Giacalone and Hedelund, 2016, Oppermann et al. 2016).
232 To investigate the frequency of attribute usage over the course of one tasting session, an ANOVA with
233 serving position as fixed factor and participant as random factor was carried out on the number of terms
234 used per sample.
235 ANOVA and Cochran’s Q tests were performed with SPSS 24 (2013, IBM Corporation, Armonk, USA),
236 XLSTAT Version 2016.03.31333 (Addinsoft, New York, USA) was used for MFA and RV coefficients.
237 All statistical analyses were performed at 5% level of significance.
10
238 Results and Discussion
239 Comparison of discriminative ability between RATA responses analyzed as frequency and intensity
240 data.
241 When considering the RATA responses as binary data (frequencies), significant differences between
242 sample descriptors were observed for 34 out of 43 (79%) attributes used in Study 1 (Table 2), and 32 out
243 of 52 (62%) for Study 2 (Table 3). In comparison, a substantially higher number of discriminating
244 attributes were observed for both studies when intensity ratings were taken into account. Intensity data
245 analyzed by ANOVA resulted in 42 (98%) and 40 (77%) attributes significantly discriminating between
246 samples in Study 1 and 2, respectively. The increase in discriminating terms when intensity ratings are
247 considered is in agreement with Oppermann et al. (2016) who found better discrimination considering
248 intensity ratings. In contrast, earlier studies (Ares, Bruzzone, et al. 2014, Meyners et al. 2016) found very
249 few differences between the different analysis methods. A possible reason for this variation, as discussed
250 by Oppermann et al. (2016), could be the perceptual differences between the samples. When samples are
251 considerably different and a diverse range of attributes are used to describe the individual sensory
252 profliles, small differences are to be expected between RATA frequency and intensity analyses. However,
253 if multiple samples share the same sensory characteristics but differ in intensities of the shared attributes,
254 RATA intensity ratings are expected to be more discriminative.
255 Comparison of discriminative ability of RATA vs. Descriptive Analysis.
256 Study 1 applied the same attribute list generated by the DA panel for the RATA assessment which meant
257 a direct comparison of the discrimination ability between the two approaches could be made. The results
258 of a two-way ANOVA performed on the DA data showed that 40 out of 43 attributes (93%) significantly
259 discriminated between the samples presented (Table 2). Similarly, the two-way ANOVA showed the
260 RATA approach with naïve consumers resulted in a slightly higher number of attributes discriminating
261 between the samples (42 of 43, 98%). The non-discriminating attributes differed between the two
11
262 approaches. Whereas, herbaceous (aroma and flavor) and red berry flavor were not significant for the
263 DA, sweet oak flavor was the sole non-discriminating attribute for the RATA approach. The high
264 agreement suggests that RATA with untrained, naïve consumers discriminates samples similarly to DA
265 with trained panelists.
266 In Study 2 the attribute lists were different between the RATA and DA assessments and therefore, as
267 Oppermann et al. (2016) and Ares et al. (2015) pointed out, it is not possible to directly compare the
268 discriminative ability between the two approaches (Table 3). Therefore, only the 16 attributes shared by
269 the two approaches were directly compared with each other, with similar results seen as for Study 1. The
270 RATA method had a slightly higher number of discriminating attributes (14 out of 16, 88%) compared to
271 the DA (12 out of 16, 75%).
272 In summary, the results from both Study 1 and Study 2 show that when assessing commercial wine
273 samples, the discrimination ability of RATA with naïve consumers was very similar to that of DA using
274 trained panelists. These findings are in agreement with previous studies which indicate that consumers,
275 when given a suitable methodology, can adequately characterize and discriminate between samples based
276 on their sensory perceptions (Moussaoui and Varela, 2010, Ares et al. 2015, Giacalone and Hedelund
277 2016, Oppermann et al. 2016). Moreover, this work expands on the existing literature by investigating the
278 sensory discrimination of complex commercial wine samples which encompass a broad range of aroma,
279 flavor and textural attributes.
280 Configuration similarity.
281 Configuration similarity was evaluated for both studies in two ways: a) visual comparison of the sample
282 and attribute alignment within the MFA plots, and b) statistical quantification of the similarities in sample
283 configurations using RV coefficients.
12
284 Analysis of Study 1 RATA and DA mean intensity ratings of statistically significant attributes by MFA,
285 found that the first 2 factors explained 96% of the variation in sensory data. Overall the visual inspection
286 of the biplot (Figure 1) shows high similarities between RATA and DA methods, with a clear separation
287 between oaked and unoaked samples on the first dimension. Oak attributes such as sweet oak, savory,
288 spice and leather clustered in the top left quadrant, with these attributes indicative of wines fermented in
289 or aged with oak products (Crump et al. 2015) (Figure 1B). Similarly, the mouthfeel attributes such as
290 alcohol/heat, astringency and body, are all clustered in the bottom-left quadrant along with the majority
291 of taste attributes (i.e., acidity, bitterness) and berry characters. This result shows good alignment with the
292 sample set presented, i.e., full-bodied red and rosé styles. The sweetness taste attribute is correlated with
293 tropical/fruity/honey/floral characters located on the right-hand side of the plot, which are common
294 varietal descriptors for the wine styles presented e.g. a colored Moscato and rosé (Masson and Schneider
295 2009). The separation of attributes by consumers using the RATA method strongly aligns with the DA
296 panel assessment, therefore demonstrating that in this instance consumers were able to separate samples
297 based on processing conditions (i.e., oak maturation), wine style (i.e., full bodied, dry red table wine
298 versus rosé styles) and typical varietal characteristics. The bottom-right quadrant shows several attributes
299 from the RATA assessment that did not align with the DA descriptors i.e., cherry and red berry. Color
300 variation between samples may be a possible reason for the lack of correlation between these descriptors
301 in the RATA and DA assessments, with consumers potentially dismissing these attributes in the full-
302 bodied reds and associating them with lighter wine styles (Parr et al. 2003). However, the wine samples
303 separate into 3 distinct groups based on their distinct sensory attributes – fruity light bodied, oak-driven
304 and berry-driven full-bodied reds. The wines showing the most variability between assessments are MOS,
305 WSHZ and TMP, which were discriminated better in the second dimension (Fig. 1B). The remaining 3
306 wines (SHZ, LGN and MTE) showed very little variability in their discrimination by trained panelists or
307 consumers. All samples had their varietal characteristics well represented in the sensory space. The high
13
308 visual similarities can be confirmed by highly significant RV coefficient of 0.97, p<0.01. In comparison
309 with previous studies, similar or even lower RV values have been regarded as indictors of good
310 agreement between sample configurations (Lawless and Clatter 1990, Giacalone and Hedelund 2016,
311 Oppermann et al. 2016)
312 Figure 2 shows the first two dimensions of the consensus MFA sample map for Study 2, in which the
313 wine samples separate into 4 distinct groups based primarily on grape variety (Figure 2B). The wines
314 showing the most variability between assessments are V2, SB2 and SB1, which were discriminated
315 slightly better in the second dimension by the DA.
316 Analysis of Study 2 RATA and DA mean intensity ratings of statistically significant attributes by MFA,
317 found that the first 2 factors explained 76% of the variation in sensory data. The biplot shows a clear
318 separation of the attributes into 4 distinct sensory groups which differentiated the wines (Figure 2A).
319 Tropical and green attributes are clustered on the top left-hand side of the plot. This result shows good
320 alignment with typical varietal characteristics observed in Sauvignon Blanc wines (Benkwitz et al. 2012).
321 Oak-related attributes such as woody, nutty, and spice are clustered in the top right quadrant. The
322 Chardonnay and Viognier wines, which according to the labels had undergone contact with oak, can be
323 found towards the right of the biplot, which confirms the ability of consumers to discriminate between
324 wines based on production methods. The bottom left quadrant contains acidity-related attributes such as
325 citrus, apple and sour, which shows good alignment with the varietal characteristics of the Riesling,
326 Gewürztraminer and Semillon wine samples. Interestingly, there was good correlation between the DA
327 and RATA descriptors located in the bottom right quadrant, such as medicinal and chemical. These terms
328 would be less familiar to the general consumer for describing wine sensory attributes, however in this
329 instance, they have been understood and used effectively to discriminate between wines in the sample set.
330 Further to this, 'floral' is a subtle wine aroma typically used to describe a particular varietal characteristic
331 of Gewürztraminer wines (Ebeler 2001). In the current sample set, consumers have used the floral term to
14
332 effectively differentiate Gewürztraminer wines from the other samples. The high visual similarities
333 between RATA and DA are supported by a highly significant RV value of 0.92, p<0.001, indicating that
334 RATA with naïve consumers using a generic (pre-defined) attribute list can be a valid method for rapid
335 sensory profiling.
336 Limitations and future perspectives.
337 Tasting 12 wine samples in one session, despite enforced breaks, can be a challenging task for untrained
338 consumers. With RATA questions not being forced-choice, a potential risk of the methodology is that the
339 number of terms used to describe samples may decrease during assessment as the participants begin to
340 fatigue. However, in this study no significant (F (11,770) = 1.296, p = 0.222) reduction on average term
341 usage was found (ranging between 31.6 for the first sample to 29.8 for the last sample), indicating
342 consistent attribute usage throughout the assessment.
343 Although this study found similar discrimination ability and configuration similarity between DA with
344 trained panelists and RATA with naïve consumers, it is worth noting that discrimination ability and
345 configurational similarities are highly dependent on sample number and product differences. In the
346 present study, commercial wine samples encompassing a number of different varieties were used,
347 therefore future work should further focus on the discriminative performance of RATA for samples of the
348 same variety or similar wine styles.
349 In both studies, the DA panelists tasted the wine in black glasses, whereas the naïve consumers received
350 the wine in clear glasses during the RATA tasting. The use of black glasses is the standard procedure for
351 DA at the University of Adelaide, as this minimizes color bias, but even more importantly, prevents
352 potential identification of wines which the panelists have seen previously during the training or earlier
353 evaluation sessions. As the naïve consumers saw each wine only once during the RATA testing, it was
354 decided to use clear glasses. This presentation is more familiar to the participants and more closely
15
355 replicates the actual consumption situation. Although very similar discrimination ability and sample
356 configurations were observed for RATA and DA, it might be beneficial for future studies to consider
357 potential color and visual biases when presenting samples to naïve consumers.
358 Meyners, Jaeger, & Ares (2016), who in investigating the analysis of RATA derived data, came to the
359 conclusion that transforming the absence of a check for an attribute to a score of 0 (and subsequent
360 parametric inference based on F-tests) provides valid statistical inference for RATA data, even though not
361 all assumptions for ANOVA are met. Future research to validate this analysis and how additional factors
362 could be incorporated (e.g. gender, consumer segments) within the model needs to be investigated.
363 It is important to note that rapid sensory methods, in this case RATA, are not intended as a replacement to
364 traditional consensus approaches performed by highly trained assessors, such as Descriptive Analysis
365 (Varela and Ares, 2012). Instead, it is anticipated that the knowledge gained from the current study may
366 demonstrate whether RATA can provide similar quantitative information to trained panel assessments,
367 thus offering a feasible option for profiling complex products with limited resources. Furthermore, using
368 consumers for sensory profiling can provide valuable insights into how consumers perceive the wines and
369 simultaneously allows the collection of additional information (i.e., preference data) (Jaeger et al. 2016,
370 Schouteten et al. 2016), not possible with highly trained panels (Lawless and Heymann, 2010). In
371 agreement with Ares et al. (2015), our recommendation is to decide, on a case-by-case basis, which
372 approach for sensory profiling aligns best with the specific research objectives while delivering the
373 required precision.
374 Conclusion
375 The two studies presented here show that RATA methodology using naïve consumers can result in very
376 similar sample discrimination and sample configurations as DA with trained panelists, when evaluating
377 commercial wine samples. This suggests that RATA can be a valid method for rapid sensory profiling
16
378 within industry and research applications which aim to describe the sensory characteristics of wines (or
379 other similarly complex products), making it particularly relevant when resources are limited and/or
380 additional consumer responses i.e., hedonic ratings or willingness-to-pay, are of interest.
381 Literature Cited
382 Ares G, et al. 2015. Comparison of sensory product profiles generated by trained assessors and consumers
383 using CATA questions: Four case studies with complex and/or similar samples. Food Qual Prefer 45:75–
384 86.
385 Ares G, Barreiro C, Deliza R, Giménez A and Gámbaro, A. 2010. Application of a Check-All-That-Apply
386 Question to the Development of Chocolate Milk Desserts. J Sens Stud 25:67–86.
387 Ares G, Bruzzone F, Vidal L, Cadena RS, Giménez A, Pineau B, Hunter DC, Paisley AG and Jaeger SR.
388 2014. Evaluation of a rating-based variant of check-all-that-apply questions: Rate-all-that-apply (RATA).
389 Food Qual Prefer 36:87–95.
390 Ares G and Jaeger SR. 2015. Check-all-that-apply (CATA) questions with consumers in practice:
391 experimental considerations and impact on outcome. In Rapid Sensory Profiling Techniques. J. Delarue,
392 B. Lawlor and M. Rogeaux (eds.), 1st ed., pp. 227–245. Woodhead Publishing, New York, USA.
393 Ares G, Mawad F, Giménez A and Maiche A. 2014. Influence of rational and intuitive thinking styles on
394 food choice: Preliminary evidence from an eye-tracking study with yogurt labels. Food Qual Prefer
395 31:28–37.
396 Benkwitz F, Nicolau L, Lund C, Beresford M, Wohlers M and Kilmartin PA. 2012. Evaluation of key
397 odorants in sauvignon blanc wines using three different methodologies. J Agric Food Chem 60:6293–
398 6302.
399 Crump AM, Johnson TE, Wilkinson KL and Bastian SEP. 2015. Influence of Oak Maturation Regimen
400 on Composition, Sensory Properties, Quality, and Consumer Acceptability of Cabernet Sauvignon Wines.
401 J Agric Food Chem 63:1593–1600.
402 Delarue J. 2015. The use of rapid sensory methods in R&D and research: an introduction. In Rapid
403 sensory profiling techniques and related methods. J. Delarue, B. Lawlor and M. Rogeaux (eds.), 1st ed.,
404 pp. 3–25. Woodhead Publishing Limited, New York, USA.
405 Ebeler SE. 2001. Analytical Chemistry: Unlocking the Secrets of Wine Flavor. Food Rev Int 17:45–64.
17
406 Franco-Luesma E, Sáenz-Navajas MP, Valentin D, Ballester J, Rodrigues H and Ferreira V. 2016. Study
407 of the effect of H2S, MeSH and DMS on the sensory profile of wine model solutions by Rate-All-That-
408 Apply (RATA). Food Res Int 87:152–160.
409 Giacalone D and Hedelund PI. 2016. Rate-all-that-apply (RATA) with semi-trained assessors: An
410 investigation of the method reproducibility at assessor-, attribute- and panel-level. Food Qual Prefer
411 51:65–71.
412 Jaeger SR and Ares G. 2015. RATA questions are not likely to bias hedonic scores. Food Qual Prefer
413 44:157–161.
414 Jaeger SR, Chheang SL, Yin J, Bava CM, Gimenez A, Vidal L and Ares G. 2013. Check-all-that-apply
415 (CATA) responses elicited by consumers: Within-assessor reproducibility and stability of sensory product
416 characterizations. Food Qual Prefer 30:56–67.
417 Jaeger SR, Kim KO, Min Lee S, Hunter DC, Kam K, Chheang SL, Jin D, Lee PY, Xia Y, and Ares G.
418 2016. Concurrent elicitation of hedonic and CATA/RATA responses with Chinese and Korean
419 consumers: Hedonic bias is unlikely to occur. Food Qual Prefer 56:130–137.
420 Lawless HT and Clatter S. 1990. Consistency of multidimentsional scaling models derived from odor
421 sorting. J Sens Stud 5:217–230.
422 Lawless HT and Heymann H. 2010. Sensory evaluation of food, principles and practices (2nd ed.). New
423 York: Springer.
424 Masson G and Schneider R. 2009. Key compounds of provence rosé wine flavor. Am J Enol Vitic
425 60:116–122.
426 Meyners M, Castura JC and Carr BT. 2013. Existing and new approaches for the analysis of CATA data.
427 Food Qual Prefer 30:309–319.
428 Meyners M, Jaeger SR and Ares G. 2016. On the analysis of Rate-All-That-Apply (RATA) data. Food
429 Qual Prefer 49:1–10.
430 Moussaoui KA and Varela P. 2010. Exploring consumer product profiling techniques and their linkage to
431 a quantitative descriptive analysis. Food Qual Prefer 21:1088–1099.
432 Noble AC, Arnold RA, Buechsenstein J, Leach EJ, Schmidt JO and Stern, PM. 1987. Modification of a
433 standardized system of wine aroma terminology. Am J Enol Vitic 38:143–146.
18
434 Oppermann AKL, de Graaf C, Scholten E, Stieger M and Piqueras-Fiszman B. 2016. Comparison of
435 Rate-All-That-Apply (RATA) and Descriptive sensory Analysis (DA) of model double emulsions with
436 subtle perceptual differences. Food Qual Prefer 56:55–68.
437 Parr WV, Geoffrey White K and Heatherbell DA. 2003. The nose knows: influence of colour on
438 perception of wine aroma. J Wine Res 14:79–101.
439 Parr WV, Heatherbell D and White KG. 2002. Demystifying wine expertise: olfactory threshold,
440 perceptual skill and semantic memory in expert and novice wine judges. Chem Senses 27:747–755.
441 Reinbach HC, Giacalone D, Ribeiro LM, Bredie WLP and Frøst MB. 2014. Comparison of three sensory
442 profiling methods based on consumer perception: CATA, CATA with intensity and Napping®. Food
443 Qual Prefer 32:160–166.
444 Sáenz-Navajas MP, Avizcuri, J-M Ferrero-del-Teso S, Valentin D, Ferreira V and Fernández-Zurbano P.
445 2017. Chemo-sensory characterization of fractions driving different mouthfeel properties in red wines.
446 Food Res Int 94:54–64.
447 Schouteten JJ, Gellynck X, Bourdeaudhuij ID, Sas B, Bredie WLP, Perez-Cueto FJA and Steur, H. De.
448 2016. Comparison of response formats and concurrent hedonic measures for optimal use of the
449 EmoSensory® Wheel. Food Res Int 93:33–42.
450 Stone H. and Sidel J. 1993. Sensory Evaluation Practices. (3rd ed.). Elsevier, London, UK
451 Stone H, Sidel JL, Oliver S, Woolsey A and Singleton CR. 1974. Sensory evaluation by quantitative
452 descriptive analysis. Food Technol 28:24–34.
453 Thorngate, JH. 1997. The physiology of human sensory response to wine: A review. Am J Enol Vitic
454 48:271–279.
455 Valentin D, Chollet S, Lelièvre M and Abdi H. 2012. Quick and dirty but still pretty good: A review of
456 new descriptive methods in food science. Int J Food Sci Technol 47:1563–1578.
457 Varela P and Ares G. 2012. Sensory profiling, the blurred line between sensory and consumer science. A
458 review of novel methods for product characterization. Food Res Int 48:893–908.
459
19
Table 1 Basic information of samples used in Study 1 and Study 2.

Sample Codes Variety Region Vintage Retail Price ($AUD)
Study 1
LGE Lagrein Riverland 2013 40
MTE Montepulciano Riverland 2015 26
MOS Moscato (colored red) Riverland NV 14
TMP Tempranillo Riverland 2015 26
SHZ Shiraz Riverland 2015 14
WSHZ Shiraz rosé Riverland 2015 7
Study 2
C1 Chardonnay Yarra Valley 2013 35
C2 Chardonnay Adelaide Hills 2014 39
G1 Gewürztraminer Eden Valley 2015 31
G2 Gewürztraminer Tamar Valley 2014 34
R1 Riesling Clare Valley 2015 52
R2 Riesling Eden Valley 2015 32
S1 Semillon Hunter Valley 2014 25
S2 Semillon Hunter Valley 2014 15
SB1 Sauvignon Blanc Marlborough 2014 11
SB2 Sauvignon Blanc Marlborough 2014 15
Languedoc-
V1 Viognier 2014 9
Rousillon
V2 Viognier Eden Valley 2014 18
20
Table 2 Comparison of the discrimination ability of RATA analyzed as frequency (Cochran's Q test) or
intensity (ANOVA), and DA (ANOVA) for Study 1. Significant p-values are bold.
RATA frequency RATA intensity DA
Terms
Q-value (df 5) p F (4,415) p F (5,50) p
Aroma
Cherry 33.212 <0.001 12.276 <0.001 10.43 <0.001
Citrus 40.486 <0.001 21.122 <0.001 10.69 <0.001
Confectionary 20.727 <0.001 13.961 <0.001 7.60 <0.001
Dark Berry 54.537 <0.001 34.915 <0.001 25.20 <0.001
Dried Fruit 13.429 0.020 5.332 <0.001 13.16 <0.001
Earthy 25.000 <0.001 15.843 <0.001 6.56 <0.001
Floral 47.764 <0.001 32.880 <0.001 16.14 <0.001
Herbaceous 7.934 0.160 2.322 0.042 1.20 0.323
Honey 27.203 <0.001 15.582 <0.001 6.78 <0.001
Jammy 3.295 0.655 2.839 0.016 4.61 0.002
Leather 32.954 <0.001 21.526 <0.001 7.71 <0.001
Lychee 51.478 <0.001 39.946 <0.001 13.33 <0.001
Oak 61.142 <0.001 29.337 <0.001 10.93 <0.001
Pepper 26.865 <0.001 14.037 <0.001 8.92 <0.001
Red Berry 21.387 0.001 10.196 <0.001 3.46 0.009
Savory 12.232 0.032 5.756 <0.001 5.96 <0.001
Spice 35.564 <0.001 20.078 <0.001 9.12 <0.001
Sweet Oak 5.000 0.416 4.729 <0.001 9.99 <0.001
Tobacco 45.627 <0.001 27.435 <0.001 5.85 <0.001
Tropical 38.502 <0.001 32.269 <0.001 49.75 <0.001
Flavor
Cherry 35.491 <0.001 13.187 <0.001 7.06 <0.001
Confectionary 11.485 0.043 13.631 <0.001 26.93 <0.001
Dark Berry 32.344 <0.001 20.254 <0.001 35.89 <0.001
Earthy 31.722 <0.001 21.913 <0.001 7.09 <0.001
Floral 51.295 <0.001 36.304 <0.001 14.09 <0.001
Herbaceous 7.407 <0.001 4.764 <0.001 1.22 0.314
Honey 56.317 <0.001 39.737 <0.001 13.41 <0.001
Jammy 3.295 0.655 2.837 0.016 5.11 <0.001
Lychee 64.242 <0.001 46.511 <0.001 11.63 <0.001
Oak 68.323 <0.001 33.536 <0.001 10.30 <0.001
Pepper 45.323 <0.001 28.590 <0.001 13.52 <0.001
Red Berry 33.333 <0.001 13.199 <0.001 1.46 0.22
Savory 6.394 0.270 3.563 0.004 4.65 0.002
Spice 44.694 <0.001 25.475 <0.001 7.35 <0.001
Sweet Oak 3.061 0.691 1.247 0.286 7.35 <0.001
Tropical 50.511 <0.001 44.029 <0.001 41.32 <0.001
Mouthfeel
Alcohol content/Heat 7.558 0.182 46.047 <0.001 23.19 <0.001
Astringency 6.744 0.241 52.589 <0.001 33.90 <0.001
Body 2.821 0.728 17.866 <0.001 5.57 <0.001
Tannins 14.412 0.013 59.725 <0.001 66.96 <0.001
Taste
Acidity 1.444 0.919 2.565 0.027 2.72 0.03
Bitterness 29.595 <0.001 20.978 <0.001 8.93 <0.001
Sweetness 42.580 <0.001 75.660 <0.001 55.67 <0.001
Number and percentage of significantly 34/43 42/43 40/43
different terms 79% 98% 93%
21
Table 3 Comparison of the discrimination ability of RATA analyzed as frequency (Cochran's Q test) or
intensity (ANOVA), and DA (ANOVA) for Study 2. Significant p-values are bold. Empty cells indicate
that this attribute was not used in this part of the study.
RATA frequency RATA intensity DA
Terms
Q-stat (df 11) p F (11,770) p F (11,110) p
Appearance
Brown 33.314 <0.001 8.563 <0.001
Clarity 3.244 0.987 2.326 0.008
Color (green-yellow) 23.469 <0.001
Color Intensity 77.043 <0.001
Green 16.691 0.117 6.606 <0.001
Yellow 7.441 0.762 42.883 <0.001
Aroma
Apple-Pear 35.153 <0.001 4.265 <0.001
Bubblegum 19.182 0.058 0.923 0.517
Butterscotch 3.875 <0.001
Buttery 41.619 <0.001 9.077 <0.001
Chemical 11.025 0.441 1.301 0.219
Citrus 28.284 0.003 4.865 <0.001 3.775 <0.001
Dried Apricot 7.806 <0.001
Dried Fruits 18.665 0.067 2.402 0.006
Floral 38.449 <0.001 5.162 <0.001 1.748 0.072
Green-Grassy 48.076 <0.001 7.633 <0.001 4.437 <0.001
Herbaceous 31.346 <0.001 5.001 <0.001
Honey 18.467 0.071 2.752 0.002 1.746 0.073
Kerosene-Burnt-Rubber 2.184 0.020
Medicinal 2.701 0.004
Milky 32.220 <0.001 5.157 <0.001
Nutty 37.190 <0.001 6.650 <0.001
Passionfruit 10.660 <0.001
Petroleum 24.747 0.010 2.312 0.009
Spice 28.077 0.003 2.753 0.002
Stone Fruits 16.338 0.129 1.306 0.216
Sulfidic 13.379 0.269 1.179 0.297
Toasted 84.627 <0.001 14.017 <0.001
Tropical 40.013 <0.001 8.427 <0.001 6.144 <0.001
Vanilla 17.655 <0.001
Wood 82.127 <0.001 15.643 <0.001 15.937 <0.001
Flavor
Apple-Pear 17.960 0.083 3.378 <0.001
Butter 57.771 <0.001 11.845 <0.001
Butterscotch 4.793 <0.001
Chemical 20.518 0.039 2.608 0.003
Citrus 33.597 <0.001 7.159 <0.001 6.396 <0.001
Dried Apricot 9.716 <0.001
Dried Fruits 28.519 0.003 2.866 0.001
Floral 18.900 0.063 4.612 <0.001 1.819 0.059
Green-Grassy 37.837 <0.001 7.339 <0.001 5.173 <0.001
22
Herbaceous 37.345 <0.001 6.197 <0.001

Honey 1.599 0.109
Kerosene-Burnt Rubber 2.017 0.033
Length 12.892 0.300 1.645 0.082 3.477 <0.001
Medicinal 2.469 0.008
Milky 33.563 <0.001 4.347 <0.001
Mineral 15.269 0.171 1.402 0.576
Nutty 27.146 0.004 5.820 <0.001
Passionfruit 10.699 <0.001
Savory 7.677 0.742 1.243 0.254
Spice 28.129 0.003 2.560 0.003
Stone Fruits 12.876 0.302 1.042 0.406
Sulfidic 9.965 0.534 1.179 0.298
Toasted 87.949 <0.001 14.301 <0.001
Tropical 39.168 <0.001 8.509 <0.001 7.144 <0.001
Vanilla 12.621 <0.001
Woody 57.085 <0.001 13.504 <0.001 21.192 <0.001
Taste
Bitter 10.909 0.451 1.291 0.225 5.136 0.001
Salty 13.952 0.236 1.984 0.027
Sour 28.391 0.003 3.299 <0.001 3.790 <0.001
Sweet 33.822 <0.001 4.748 <0.001 0.988 0.462
Mouthfeel
Body 14.776 0.193 6.315 <0.001 3.548 <0.001
Creamy 50.500 <0.001 7.701 <0.001
Crisp 20.493 0.039 5.207 <0.001
Dry 7.971 0.716 1.676 0.074
Heat 2.244 0.017
Spritz 8.144 0.700 1.139 0.327
Watery 15.831 0.148 3.430 <0.001
Number and percentage of 32/52 40/52 27/32
significantly different terms 62% 77% 84%
23
Figure 1 Representation of the (A)

significantly discriminating sensory
attributes, and (B) samples of Study
1 in the first two dimensions of the
MFA performed on the significantly
discriminating attributes of the DA
with trained assessors and RATA
with naive consumers. A = aroma, T
= taste, M = mouthfeel and F =
flavor. Wine sample abbreviations
include TMP = Tempranillo, SHZ =
Shiraz, LGE = Lagrein, MTE =
Montepulciano, WSHZ = Shiraz
rosé, MOS = Moscato, colored pink
with commercial grape skin extract.
24
Figure 2 Representation of the (A)

significantly discriminating sensory
attributes and (B) samples of Study 2
in the first two dimensions of the MFA
performed on the significantly
discriminating attributes of the DA
with trained assessors and RATA with
naive consumers. Ap = appearance, A
= aroma, T = taste, M = mouthfeel and
F = flavor. C1 & C2 Chardonnay, G1
& G2 Gewürztraminer, R1 & R2
Riesling, S1 & S2 Semillon, SB1 &
SB2 Sauvignon Blanc and V1 & V2
Viognier.
25
Supplemental Table 1 Attribute list for the DA and RATA for Study 1, including description and
reference standards. Reference standards were only used for the DA.
Attribute Description Reference
Aroma/Flavor
5g John West passionfruit pulp, 2 x 2cm2 pineapple pieces, 10 ml
Tropical Passionfruit, guava, pineapple, melon
Golden Circle Guava Nectar
Lychee Fresh or canned lychee 2 x lychees (Coles brand canned pitted lychee’s in syrup)
Citrus Lemon, grapefruit 3cm2 piece ruby grapefruit with skin
Red berry Strawberry, raspberry 1 x fresh strawberry
Cherry Cherry or sour cherry Sandhurst Amarena cherries in syrup
Blackcurrant, blackberry, mulberry,
Dark berry
plum, blueberry
Quince paste, dried fig, fruit cake, 1 Tblsp Beerenberg Quince paste, 2cm2 piece Black & Gold dark
Dried fruit
sultana, dried apricot fruit cake
Jammy Preserved or cooked fruit 1 Tblsp Cottee’s Fruits of the Forest jam
Confectionary Strawberries and cream or red lollies 2 x Allen’s red raspberries
Floral Violet, rose, hibiscus, orange blossom Alwadi Al Akhdar Sal orange blossom water
Honey Honey, floral 2 Tblsp Buzz Honey Orange Blossom
Herbaceous Grassy, tomato leaf, capsicum 2 x tomato stalks, 1cm2 green capsicum
Oak Oak, woody, saw dust French oak chips (medium toast)
Sweet oak Vanilla, chocolate, coffee, caramel, nutty 1 square Lindt 85% dark chocolate, 20ml 1% vanillin solution
Leather Leather 3cm x 1cm piece tanned leather
Jose L. Piedra Petit Cazadores Cuba cigar, 5g Erinmore Mixture
Tobacco Tobacco, cigar
pipe tobacco
Clove, cinnamon, nutmeg, mixed spice,
Spice 5g Masterfoods mixed spice, 3 cloves
mulled wine
Pepper Black pepper 5g black peppercorns
Earthy Mushroom, dusty, beetroot 1 x mushroom
10ml Kikkoman soy sauce, 10g Always Fresh sliced Kalamata
Savory Savory, meaty, soy sauce, black olive
olives
Taste
Sweetness Level of sweetness perceived Low = 10g/L sucrose in water, High = 25g/L sucrose in water
Low = 0.5g/L tartaric acid in water, High = 1.4g/L tartaric acid in
Acidity Level of acid perceived
water
Bitterness Level of bitterness perceived Low = 6mg/L quinine in water, High = 12mg/L quinine in water
Mouthfeel
Low = 0.5g/L tannin in water, High = 1.5g/L tannin in water
Astringent Sensation of drying or mouth puckering
(Laffort Nadine)
Level of heat perceived due to ethanol High (14.5%) = 20.8g/L ethanol (Tarac 96%) in red wine (Berri
Alcohol/Heat
content Estates Traditional Dry Red)
Body Perceived weight of wine in the mouth
Tannin structure e.g. coarse = fine-grade
Tannins
sand paper
26
Supplemental Table 2 Attribute list for the DA for Study 2, including description and reference
standards.
Attribute Description Reference
Appearance
Green to yellow Green to yellow
Colour intensity Low to high
Aroma Perception when smelling the sample
10 ml canned passionfruit (John West,
Tropical Fruit Any combination of tropical fruit. Mentone, Australia) plus one slice of fresh
pineapple
Passionfruit Passionfruit aroma 10 ml canned passionfruit
Dried apricot Dried apricot 2 slices of dried apricot
Citrus Any combination of citrus fruit including lemon and lime 1 slice of lemon and lime
1mL/L rose water (Queen Fine Foods Pty.
Floral/Perfume Any combination of flowers Ltd., Alderley, Australia) diluted in
Yalumba Classic Dry White Cask wine
Grassy/Green Fresh cut grass Fresh cut grass
0.5mL/L vanilla essence (Queen Fine Foods
Vanilla Any of combination vanilla, coconut, caramel, brown sugar Pty. Ltd., Alderley, Australia) diluted in
Yalumba Classic Dry White Cask wine
Woody-a Any of combination woody and oak flavours Light toast oak chips in white wine
Butterscotch Butterscotch, Caramel One crushed butterscotch
Honey Honey 3 ml of honey
Medicinal Medicinal, Band-Aid 2 cm piece of band aid
Kerosene/Burnt
Any combination of Kerosene, petrol and burnt rubber Small piece of burnt rubber
Rubber
Flavour Taste and retro-nasal perception when tasting a sip of the sample
Tropical Fruit Any combination of tropical fruit
Passionfruit Passionfruit
Dried apricot Dried apricot
Citrus Any combination of citrus fruit including banana
Floral/Perfume Any combination of flowers
Green/Grassy Fresh cut grass
Vanilla Any of combination vanilla, coconut, caramel, brown sugar
Woody- Any of combination woody and oak flavours
Butterscotch Butterscotch, Caramel
Honey Honey
Medicinal Medicinal, Band-Aid
Kerosene/Burnt
Any combination of Kerosene, petrol and burnt rubber
Rubber
Sour How sour is the wine?
Sweet How sweet is the wine?
Bitter How bitter is the wine?
Overall flavour
Duration of flavour perception after expectoration.
length
Mouthfeel Perception when tasting the wine
Heat How hot in terms of alcohol perception do you perceive this wine?
Body How would you describe the wine body in your mouth?
27
Supplemental Table 3 List of attributes and description for Study 2, presented to the consumer during
the RATA study.
Attribute Description
Appearance
Brown The color from that of pale cardboard through to intense dark chocolate
Clarity The ability to see through the wine
Green From a pale grass to an intense leafy green
Yellow From a pale straw to a rich gold
Aroma Perception when smelling the sample
Apple/Pear The scent of fresh cut apples and pears
Bubblegum A sweet aroma like candies/lollies
Buttery Rich dairy product from cows
Chemical Cleaning products used in the home
Citrus Lemon, orange, lime, grapefruit and any other citrus fruits, or any combination of these
Dried Fruit Prune, date, sultana and any other dried fruits, or any combination of these
Floral Rose, jasmine, honeysuckle, and any other perfumed flowers, or any combination of these
Green/Grassy Freshly mowed/cut grass, green pepper/capsicum or any combination of these
Herbaceous Basil, thyme, oregano, mint, and any other fresh garden herbs, or any combination of these
Honey Honey
Milky Odor of cow based dairy products
Nutty Almond, cashew, hazelnut, walnut, chestnut, any other nut, or any combination of these
Petroleum Kerosene, gas station, petrol pump, or any combination of these
Spice Liquorice, star anise, fennel, cardamom or any other spice used in cooking, or any
combination of spices
Stone Fruits Peach and nectarine, or any combination of these
Sulfidic Rubber, onion, egg, cabbage, or any combination of these
Toasted Like fresh toast, grilled/hot bread, or a rich woody aroma
Tropical Pineapple, passionfruit, mango, or any combination of these
Wood Flavor of oak in wine, e.g. vanilla
Flavor Flavor perception when tasting the wine
Apple/Pear Freshly cut apples and pears
Bitter A harsh feeling like that given by flat tonic water or strong coffee
Butter Rich dairy product from cows
Chemical Cleaning products used in the home
Citrus Lemon, orange, lime, grapefruit and any other citrus fruits, or any combination of these
Dried Fruits Prune, date, sultana and any other dried fruits, or any combination of these
Floral Rose, jasmine, honeysuckle, and any other perfumed flowers, or any combination of these
Green/Grassy Freshly mowed/cut grass, green pepper/capsicum, or any combination of these
Herbaceous Basil, thyme, oregano, mint, and any other fresh garden herbs, or any combination of these
Milky Odor of cow based dairy products
Mineral Stoney, chalky feeling in wines without strong tannin/astringency
Nutty Almond, cashew, hazelnut, walnut, chestnut, any other nut, or any combination of these
Salt Basic taste of salt
Savory Preserved meats like pastrami, salami, sausage or any combination of these
Sour Basic taste of sour
Spice Licorice, star anise, fennel, cardamom or any other spice used in cooking, or any combination
of spices
Stone Fruits Peach and nectarine, or any combination of these
Sulfidic Rubber, onion, egg, cabbage, or any combination of these
Sweet Presence of sugar
Toasted Like fresh toast, grilled/hot bread, or a rich woody aroma
Tropical Pineapple, passionfruit, mango, or any combination of these
Woody Vanilla, or the flavor of oak in wine
Length The time which the flavor/feeling of the wine persists in the mouth
28
Mouthfeel Perception of the wine. How does the wine feel in your mouth?
Body The weight or thickness of the wine in the mouth
Creamy A feeling of smoothness, like cream
Crisp A quick finish like a fresh apple
Dry The presence of astringency
Spritz The feeling of bubbles or sparkle in the mouth
Watery The absence of weight or thickness of the wine in the mouth
29

Comparison of Rate-All-That-Apply (RATA) And: Descriptive Analysis (DA) For The Sensory Profiling of Wine

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Comparison of Rate-All-That-Apply (RATA) And: Descriptive Analysis (DA) For The Sensory Profiling of Wine

Uploaded by

Copyright:

Available Formats

AJEV Papers in Press. Published online September 7, 2017.

American Journal of Enology and Viticulture (AJEV). doi: 10.5344/ajev.2017.17052

38 Rate-All-That-Apply (RATA), sensory characterization

60 Profiling, Check-All-That-Apply (CATA), and Rate-All-That-Apply (RATA) (Moussaoui and Varela

65 discrimination of products (Jaeger and Ares 2015).

80 attributes present across products.

87 Bruzzone, et al. 2014).

102 applications. The project focused on two specific research questions:

107 results were found.

110 Wine Aroma Wheel© is used for RATA.

112 discrimination ability of RATA undertaken by consumers compared to traditional DA methodology, in

118 Methods and Materials

122 Study one.

123 Wine Samples.

132 Descriptive Analysis.

174 Study two.

175 Wine Samples.

179 described for Study 1.

180 Descriptive Analysis.

195 Assessment Sessions.

203 Statistical analysis.

205 Descriptive analysis.

210 RATA analyses.

221 analyses reported herein.

234 used per sample.

237 All statistical analyses were performed at 5% level of significance.

238 Results and Discussion

254 RATA intensity ratings are expected to be more discriminative.

255 Comparison of discriminative ability of RATA vs. Descriptive Analysis.

265 with trained panelists.

271 the DA (12 out of 16, 75%).

279 flavor and textural attributes.

280 Configuration similarity.

283 configurations using RV coefficients.

311 Oppermann et al. 2016)

315 slightly better in the second dimension by the DA.

335 sensory profiling.

336 Limitations and future perspectives.

342 consistent attribute usage throughout the assessment.

348 same variety or similar wine styles.

373 required precision.

381 Literature Cited

Table 1 Basic information of samples used in Study 1 and Study 2.

Herbaceous 37.345 <0.001 6.197 <0.001

Figure 1 Representation of the (A)

Figure 2 Representation of the (A)

You might also like