Food Quality and Preference 32 (2014) 317–329

Contents lists available at ScienceDirect

Food Quality and Preference

journal homepage: www.elsevier .com/locate / foodqual

Preference mapping of apple varieties in Europe

0950-3293/$ - see front matter � 2013 Elsevier Ltd. All rights reserved.http://dx.doi.org/10.1016/j.foodqual.2013.09.010

⇑ Corresponding author. Tel.: +34 972780275; fax: +34 9727801517.E-mail address: joan.bonany@irta.cat (J. Bonany).

J. Bonany a,⇑, C. Brugger b, A. Buehler b, J. Carbó a, S. Codarin c, F. Donati d, G. Echeverria e, S. Egger b,W. Guerra f, C. Hilaire c, I. Höller f, I. Iglesias e, K. Jesionkowska g, D. Konopacka g, D. Kruczynska g,A. Martinelli i, C. Petiot j, S. Sansavini d, R. Stehr k, F. Schoorl l

a IRTA-Mas Badia, E-17134 La Tallada, Spainb Agroscope Changins-Wädenswil Research Station ACW, Schloss 1, P.O. Box, 8820 Wädenswil, Switzerlandc CTIFL-Lanxade, 28 Route des Nebouts, 24130 Prigonrieux, Franced University of Bologna, Viale Fanin, 44, 40127 Bologna, Italye IRTA-Lleida, Av. Alcalde Rovira i Roure, 191, 25198 Lleida, Spainf Research Centre for Agriculture and Forestry Laimburg, Laimburg 6, I-39040 Posta Ora (BZ), Italyg Research Institute of Horticulture, Konstytucji 3 Maja 1/3, 96-100 Skierniewice, Polandi Consorzio Italiano Vivaisti, Stat. Romea n. 116, 44020 S. Giuseppe di Comacchio (Ferrara), Italyj NOVADI, 23 rue Jean Baldassini F-69364 Lyon Cedex 07, Francek Obstbau Versuchs-und Beratungszentrum Jork, Moorende 53, 21635 Jork, Germanyl WUR-PPO, Lingewal 1, 6668 LA Randwijk, Netherlands

a r t i c l e i n f o

Article history:Received 13 September 2012Received in revised form 25 September 2013Accepted 27 September 2013Available online 7 October 2013

Keywords:AppleVarietiesConsumer acceptancePreference mappingConsumer segmentationEuropean countries

a b s t r a c t

A consumer test carried out in 7 different European countries compared 3 standard apple varieties to 8new ones. A total of 4290 consumers took part in the test. Data from this test was used to develop a pref-erence map for apple. The preference map was constructed with 3 main dimensions (1 – sweetness, fruit-iness, flowery attributes, 2 – acidity, firmness, 3 – juiciness and crispness). Consumers were segmented in6 clusters according to their preferences. The 6 clusters were grouped into two main mega clusters A (68%of consumers) and B (32% of consumers). Megacluster A (Clusters 1, 2, 5 and 6) was characterized by pre-ferring sweet apples. Clusters 2 and 5 (41% of consumers) liked sweet apples independently of their acid-ity and firmness and moderate positive values on dimension of juiciness and crispness. Cluster 1 (21% ofconsumers) had an optimal point in positive values of the sweetness dimension, moderate negative valuefor acidity and firmness and moderate positive value for juiciness and crispness. Cluster 6 (6% of consum-ers) besides preferring sweet varieties disliked acid-firm varieties. As to regard to megacluster B (Clusters3 and 4) (32% of consumers), they preferred varieties that were acidic-firm and juiciy and crisp with val-ues in the mid range of the sweetness dimension. In spite of the difficulties in translating preferencedimensions into standard practical values for fruit quality and the fact of being a punctual measurementof consumer behaviour, this preference map could be of practical use of different actors on the fruit valuechain like marketers and breeders.

� 2013 Elsevier Ltd. All rights reserved.

1. Introduction

Many of the European countries do not achieve minimal recom-mended intake of fruit and vegetables proposed by FAO/WHO(Robertson et al., 2004). Increase fruit consumption is therefore apublic health objective (WHO, 2003) that has been translated intoseveral campaigns of promotion of fruit consumption (Departmentof Health. The NHS Plan – A plan for investment, 2000 and Centerfor Disease Control, 2002 cited in Gilmer, 2005; Lock, Pomerleau,Causer, Altmann, & McKee 2005; Subar et al., 1995) but also re-search projects promoted from the European Union to diminish

barriers to fruit consumption that could hamper the achievementof minimal fruit intake.

A possible barrier for increased fruit consumption is insufficientfruit quality (Briz et al., 2008; Cohen, Stoddard, Sarouhkhanians, &Sorensen, 1998; Yeh et al., 2008). On apple, in the recent years, anew generation of apple cultivars with improved fruit quality arenow making their way into the markets after their release (‘PINKLADY� Cripps Pink cov’, ‘KANZI� Nicoter cov’, ‘ARIANE cov’, amongothers). Many of these new apple varieties have a improvedtexture, higher soluble solids and higher total titratable aciditythan currently cultivated varieties like ‘GOLDEN DELICIOUS’ or‘JONAGOLD’, the two most cultivated varieties in Europe.

On the other hand, to increase fruit consumption it could behelpful to know what the preferences of the consumers are and

Table 2Temperature, O2 and CO2 concentration and Relative Humidity of cold storage rooms.

Variety Temperature(�C)

O2

(%)CO2

(%)Relativehumidity (%)

GOLDEN DELICIOUS 1.3 1.0 3.0 >95JONAGOLD na na na naFUJI 1.3 1.5 1.3 91–93PINK LADY� Cripps Pink cov 2.1 4.6 0.9 92ARIANE cov 1–1.5 1.5–2 <1.3 90–93RUBENS� CIVni cov 1 1.2 1.5 naKANZI� Nicoter cov 1.3 1.5 1.3 90–93JUNAMI� Milwa cov 1 1.2 2.5 naWELLANT� Cpro-47 1.5–2 1.3a 3 naLIGOL 1.7 2.0 1.8–2.0 Up to 90GOLDCHIEF� Gold Pink cov 1.5 2 3 90–95

a Some periods temporary higher concentration.

318 J. Bonany et al. / Food Quality and Preference 32 (2014) 317–329

how these consumers are segmented with regard to those prefer-ences. Targeting consumer preferences can lead to more satisfiedconsumers which in turn can result in increased fruit consumption.Consumer preference mapping is the technique most widely usedin the food and beverage industries to evaluate the preferencesof the consumers and the segmentation of these consumers onhomogeneous groups. Additionally, this methodology providesinformation on what are the food attributes that explain consumerpreference. On fruit this technique has been used on apples (Allan-Wotjas, Sanford, McRae, & Carby, 2003; Daillant-Spinnler, MacFie,Beyts, & Hedderley, 1996; Jaeger, Andani, Wakeling, & MacFie,1998; Villatoro, López, Echeverria, Graell, & Lara, 2009) , pears(Harker, Gunson, & Jaeger, 2003; Jaeger, Lund, Lau, & Harker,2003a) and kiwifruit (Jaeger, Rossiter, Wismer, & Harker, 2003b).

Consumer tests comparing consumer acceptance of differentapple varieties have been carried out in the past in the UnitedKingdom, Denmark, Holland, Germany, Switzerland, Poland,France, and other European countries (Barendsee, 1993; Daillant-Spinnler et al., 1996; Ellinger, 1987; Höhn & Güggenbühl, 1999;Jaeger et al., 1998; Kellerhals et al., 1999; Konopacka, Jesionkowska,Rutkowski, Płocharski, & Tomala, 2006; Mante, 1973; Tomala,Barylko-Pikielna, Jankowski, Jeziorek, & Wasiak-Zys, 2009; van deAbeele & Reijnders, 1980).

In spite of numerous consumer tests, many of them, be eitherconsumer tests comparing acceptance of varieties or using themethodology of consumer preference mapping, do not includethe major apple varieties consumed in Europe and logically didnot cover the newly released cultivars. Moreover, none of the testsinvolve consumers of a wide selection of European cities impedinga valid comparison among different areas in Europe.

A consumer test of 11 apple varieties carried out in differentEuropean countries was used as methodological tool to test thehypothesis that new cultivars with better quality attributes wouldincrease fruit consumption. In a previous paper, Bonany et al.,(2013), results on consumer eating quality acceptance as relatedto variety and demographic factors were reported. In this paper, fo-cus is on the results of applying the preference mapping method-ology on the same data set.

2. Materials and methods

2.1. Varieties included in the consumer test, origin of fruit samples andmanagement to simulate fruit chain

Varieties included in the apple consumer test are listed inTable 1. The varieties selected for the apple consumer test wereselected to represent on one side the most cultivated producedapple varieties in Europe (GOLDEN DELICIOUS’ and JONAGOLD)and some newly introduced varieties (FUJI and PINK LADY� CrippsPink cov) or in process of introduction (KANZI� Nicoter cov,

Table 1Varieties used for the apple consumer test.

Variety Country were the variety was bred

‘Golden Delicious’ United States‘Jonagold’ United States‘Fuji’ Japan‘Pink Lady�’ Cripps Pinkcov Australia‘Arianecov’ France‘Rubens�’ CIVnicov Italy‘Kanzi�’ Nicotercov Belgium‘Junami�’ Milwacov Switzerland‘Wellant�’ CPRO-47 The Netherlands‘Ligol�’ Poland‘Goldchief�’ ‘Gold Pinkcov’ Italy

a 2nd Pick.

JUNAMI� Milwa cov, WELLANT�, LIGOL, ARIANE cov’, LIGOL,GOLDCHIEF� Goldpink cov or RUBENS� CIVni cov). On the otherside, they were also selected because it was anticipated fromknowledge of the varieties that they would provide a good repre-sentation of the sensory space (flavour and texture mainly). Fruitsof these varieties were harvested from a single representativecommercial orchard with standard management practices for eachvariety. Harvest dates and location and country of origin for eachvariety can also be found in Table 1.

After harvest, fruits were sorted and stored under the appropri-ate conditions for each variety for long term storage. Locations forstorage and detailed conditions are described in Table 2. All fruitsamples were removed from cold storage on 15 January 2007 andtransported to consumer test locations by means of refrigeratedvehicle where they were held between 3 �C and 4 �C until 48 h priorutilization in consumer test. The last 48 h before the test wascarried out, fruit samples were maintained at room temperature.

2.2. Fruit quality measurements

Fruit quality (Soluble Solids Content, SSC, �Brix; Total TitratableAcidity, TTA, g/L equivalent malic acid; Firmness, F, kg) on 25 fruitsfor all cultivars included in the test, was measured in differentpoints in time throughout the simulation of the fruit chain: at har-vest, at the end of the cold storage period, after the transportationto the location of the consumer test and just prior the moment ofconsumer test. Only this later measurement was used in the statis-tical analysis. The rest were used as quality control of the evolutionof the different parameters. Starch Pattern Index was also mea-sured at harvest time.

2.3. Trained panel sensory evaluation of fruit samples

Additionally to the fruit quality analysis, all varieties were sub-jected to a sensory evaluation by the AGROSCOPE trained panel in

Location of orchard Harvest date

Merano (Italy) 30/9/2006Wijk bij Duurstede (The Netherlands) 11/10/2006a

Tramin (Italy) 16/10/2006Nîmes (France) 2/11/2006Saint Laurent des Vignes (France) 8/9/2006Randwijk (The Netherlands) 27/09/2006Schenna (Italy) 6/10/2006Randwijk (The Netherlands) 5/10/2006Randwijk (The Netherlands) 3/10/2006Zalesie (Poland) 25/09/2006Coredo (Italy) 16/10/2006

J. Bonany et al. / Food Quality and Preference 32 (2014) 317–329 319

Wädenswil (Switzerland). The trained panel consisted of 12 peoplefrom the staff of AGROSCOPE who were selected after checking forhomogeneity with the other panellists and repeatability of samplescoring. All panellists had previous experience and were previouslytrained and involved in sensory testing of apples. The performanceof the panel was checked using PanelCheck software (2006). Thesensorial evaluation was replicated twice and in each replicatetwo sessions were held in two consecutive days (replicate 1: Jan25 and 30, replicate 2: Feb 1 and 6, 2007). In each session, 5 or 6apple varieties samples were presented using three digit codesand according to incomplete block design (latin square). Samplesof fruits were served cut in halves. The panellists were advised tobite in from the side, having the fruit flesh side on the top andthe skin side below when placing the apple sample in the mouth.A total of 18 attributes were evaluated using a non-structured10 cm line scale verbally anchored at both ends using the FIZZ Soft-ware (Biosystèmes, F-Courternon) (Table 3).

2.4. Consumer tests

Consumer tests were carried out in different sites located inFrance, Netherlands, Germany, Poland, Switzerland, Italy andSpain. The test was organized in 39 sessions involving 110 con-sumers each with a total of 4290 consumers participating. All ses-sions took place within a relatively short period of time tominimize the changes in fruit quality attributes ranging from24/01/2007 the earliest to 10/02/2007 the latest. Site, date, partnercarrying out the test, number of consumers and the type of locationcan be found in Table 4. Target population was people between 15and 70 years old located in major representative European citiesregardless of their apple consumption habit for the objective of in-crease fruit consumption is for all population.

All varieties were tested in each session. Fruit samples wereprepared so two fruits wedges from the transition from the wellexposed part of the fruit to the non well exposed part were becut from each fruit. Each fruit wedge piece, unpeeled, was placedin a previously coded labelled disposable Styrofoam cup andplaced into a tray with 6 varieties. Fruit wedges were maintainedonly for a short period of time to avoid browning. Experimental de-sign was incomplete blocks derived from Mutually Orthogonal La-tin Squares (MOLS) with 110 consumers in each session tasting 6

Table 3Attributes evaluated by the trained panel in AGROSCOPE (Wadesnwil, Switzerland).

Definition

SmellOverall odour intensity (O/OS) Overall odour intensityGrassy (O/GR) Odour intensity of fresh cut grassFruity (O/FR) Intensity of fruity odoursFlowery (O/FL) Intensity of flowery odours

TextureJuiciness (T/JU) Amount of liquid released during chewingFirmness (T/FN) Force required to bite the appleMealiness (T/ME) Mealy mouth-feeling while chewing the appleCrispness (T/CR) Force for first bite and its noise intensityChewiness (T/CH) Time and number of mastication steps required toToughness of the skin (T/TP) Force needed to penetrate the skinFineness (T/FI) Smoothness of the fruit flesh

FlavourOverall flavour intensity (F/OF)Sweetness (F/SW)Sourness (F/SO)Grassiness (F/GR) Flavour intensity of fresh cut grassFloweriness (F/FL) Intensity of flowery flavourFruitiness (F/FR) Intensity of fruity flavourApple flavour (F/AF) Intensity of apple flavour

varieties out of the 11 varieties to minimize the order and carryover effects (Callier & Schlich, 1997; Wakeling & MacFie, 1995).

People were invited, according to the target population, to par-ticipate in the test and if they accepted they were given a tray with6 varieties according to pre-established statistical design. After abrief explanation about the purpose of the European research pro-ject ISAFRUIT, under which the tests were carried out, the inter-viewer did shortly explain how to perform the test and give anindication of the total amount of time estimated to carry it out.Invitation to take part in the test was done so proportions repre-senting the desired combination of ages and genders wereachieved. Interviewers worked to prevent any cross consumerinfluence.

Each consumer was given a questionnaire in paper format inwhich the samples were identified in the same order as in the plas-tic tray. The questionnaire contained 4 different parts. Part A re-corded information about age class, gender and classification forconsumer habit for apples. Part B included a 9 scale hedonic testof the 6 fruit samples for eating quality acceptance. Part C consistedof a question addressed to estimate the potential increase of appleconsumption potentially brought about by the consumption of thatspecific variety. Finally, Part D included a 9 scale hedonic test forfruit visual acceptance for all 11 varieties included in test.

Consumers were asked to taste each variety individually andanswer the questions in Part B and C separately for each variety.After parts A, B and C of the test were completed, they wereprompted to move to separate but contiguous part where theywere shown a display with fruits of the 11 varieties under testand ask to answer Part D of the questionnaire. In this paper, onlydata from parts A and B will be analyzed.

2.5. Statistical analysis

Preference and perceptual data was analyzed using ExternalPreference Mapping approach as described by Schlich andMcEwan, (1992). The first part of this procedure makes uses ofPrincipal Component Analysis applied to a matrix of values withattributes evaluated by the expert panel (sensory) and instrumen-tal variables for each one of the 11 varieties as variables and thevarieties as observations. The output of this procedure is whatis called a sensory map or biplot where projections of the

Beginning scale End of scale

Not applicable Very highNot applicable Very highNot applicable Very highNot applicable Very high

Not applicable Very highNot applicable Very highNot applicable Very highNot applicable Very high

the prepare the apple piece before swallowing Short LongNot applicable Very toughFine Rough

Not applicable Very highNot applicable Very highNot applicable Very highNot applicable Very highNot applicable Very highNot applicable Very highNot applicable Very high

Table 4Location, sessions, date, partners, number of consumers and type of location of the apple varieties consumer test developed.

Location Session Date Partner Number consumers Type of location

Espai Gironés, Girona, Spain 1 24/1/2007 IRTA 110 Commercial mallValencia, Spain 2 2/2/2007 IRTA 110 Commercial mallBarcelona, Spain 3 7/2/2007 IRTA 110 Commercial mallBarcelona, Spain 4 9/2/2007 IRTA 110 Fruit shopZaragoza, Spain 5 27/1/2007 IRTA 110 Commercial mallMadrid, Spain 6 2/2/2007 IRTA 110 Commercial mallMadrid, Spain 7 3/2/2007 IRTA 110 Commercial mallSan Sebastian, Spain 8 10/2/2007 IRTA 110 Commercial mallWarsaw, Poland 1 24/1/2007 RIPF 110 Commercial mallWarsaw, Poland 2 26/1/2007 RIPF 110 Commercial mallWarsaw, Poland 3 31/1/2007 RIPF 110 Commercial mallWarsaw, Poland 4 2/2/2007 RIPF 110 Commercial mallQuickborn–Hamburg, Germany 5 24/1/2007 JORK 110 Phone banking companyStade–Hamburg, Germany 6 25/1/2007 JORK 110 Restaurant/schoolQuickborn–Hamburg, Germany 7 1/2/2007 JORK 110 Phone banking companyBuxtehude–Hamburg, Germany 8 30/1/2007 JORK 110 School, meeting roomRome, Italy 1 24/1/2007 LAIMBURG 110 Consumer test facilityRome, Italy 2 26/1/2007 LAIMBURG 110 Consumer test facilityRome, Italy 3 31/1/2007 LAIMBURG 110 Consumer test facilityRome, Italy 4 2/2/2007 LAIMBURG 110 Consumer test facilitySan Lazzaro–Bologna, Italy 5 24/1/2007 UNIBO, CIV 110 Commercial mallSan Lazzaro–Bologna, Italy 6 31/1/2007 UNIBO, CIV 110 Commercial MallCentro Lame, Bologna, Italy 7 7/2/2007 UNIBO, CIV 110 Commercial mallCentro Lame, Bologna, Italy 8 9/2/2007 UNIBO, CIV 110 Commercial mallWinkelcentrum Overvecht, Utrecht, Netherlands 1 24/1/2007 WUR-PPO 110 Commercial mallWinkelcentrum Overvecht, Utrecht, Netherlands 2 25/1/2007 WUR-PPO 110 Commercial mallWinkelcentrum Schalkwijk,Haarlem, Netherlands 3 31/1/2007 WUR-PPO 110 Commercial mallWinkelcentrum Schalkwijk,Haarlem, Netherlands 4 1/02/2007 WUR-PPO 110 Commercial mallLa Ville du Bois–Paris, France 6 25/01/2007 CTIFL 110 Commercial mallLa Ville du Bois–Paris, France 7 26/01/2007 CTIFL 110 Commercial mallLa Ville du Bois–Paris, France 8 31/01/2007 CTIFL 110 Commercial mallFrench speaking area, Switzerland 1 26/01/2007 ACW 110 SupermarketFrench speaking area, Switzerland 2 7/02/2007 ACW 110 SupermarketGerman speaking area, Switzerland 3 26/01/2007 ACW 110 Commercial mallGerman speaking area, Switzerland 4 2/02/2007 ACW 110 SupermarketGerman speaking area, Switzerland 5 9/02/2007 ACW 110 Commercial mallCastelssarrasin, France 6 31/01/2007 CTIFL 110 Commercial mallBordeaux–Sainte Eulalie, France 7 9/02/2007 CTIFL 110 Commercial mallBordeaux–Sainte Eulalie, France 8 10/02/2007 CTIFL 110 Commercial mall

1 SSC: Soluble Solid Contents %, TTA: Total Titrable Acidity, g/L, F: Firmness, kg, F/AF: Flavour-Apple Flavour, F/FL: Flavour-Flowery, F/FR: Flavour-fruitness, F/GR:Flavour: grassiness, F/OF: Flavour-Overall Flavour, F/SO: Flavour-Sourness, F/SW:Flavour-Sweetness, O/FL: Odour-Flowery, O/FR:Odour-Fruity, O/GR: Odour-Grassi-ness, O/OS: Odour-Overall Smell, T/CH: Texture-Chewiness, T/CR: Texture-Crispness,T/FN: Texture-Firmness, T/FI: Texture-Fineness, T/JU: Texture-Juiciness, T/ME:Texture-Mealiness, T/TP: Texture-Toughness Peel.

320 J. Bonany et al. / Food Quality and Preference 32 (2014) 317–329

observations (varieties) and variables (sensory attributes andinstrumental quality variables) are plotted in. XLSTAT-Pro module(XLSTAT, New York, NY, USA) was used for this procedure.

Afterwards, a second step consisting of grouping consumersaccordingly to their preferences was undertaken. More specifically,given that the data set was incomplete (every consumer tastedonly 6 out of the 11 varieties), a specific technique was used: CLIP(Clustering of Incomplete Preferences) developed by Callier andSchlich, (1997). With this technique, a similarity matrix is pro-duced. In this matrix, each element i, j of the matrix representsan index of similarity between consumer i and consumer j, thehigher the index, the higher the similarity of the ratings of the vari-eties that consumer i and consumer j have tasted in common. Thissimilarity matrix was obtained using a specifically written SAS pro-gram ( SAS Institute Inc., Cary, NC, USA) following the method de-scribed in Callier and Schlich, (1997). The similarity matrix is thenused as the input data for the Weighed Multidimensional Scalingprocedure using PROC MDS (SAS Institute Inc., Cary, NC, USAtware). The result is a two dimension map of consumers with sim-ilar preferences. Based on the position in this map, clusters of con-sumers with similar varietal preferences are created using PROCFASTCLUS (SAS Institute Inc., Cary, NC, USA).

Finally using the average acceptance values for each cluster, 4different models are tested to predict the acceptance of each con-sumer group based on the attributes of the varieties: (1) vectormodel, (2) circular ideal point model, (3) elliptical ideal point mod-el and (4) quadratic surface model. This part was carried out usingPREFMAP method in XLSTAT-MX module (XLSTAT, New York, NY,USA).

3. Results

3.1. Descriptive and correlation analysis

Overall mean, standard deviation, minimum and maximum ofall sensory and instrumental quality variables and means for eachvariety are listed in Table 5. In Table 6, the correlation matrixamong all variables can be observed. Quite an important numberof significant correlations were found among variables.

Among instrumental variables TTA1 and F were significantly cor-related (r = 0.757). The relationship between instrumental qualityvariables (TTA, F, SSC) and their corresponding sensory variables(F/SO, F/FN, F/SW) was also significant. TTA and F/SO correlationwas the highest with an r value of 0.967. The relationship betweenF and T/FN was 0.767, being the lowest between SSC and F/SW withan r value of 0.710. Additionally, SSC was positively correlated withF/AF, F/FR, F/OF and negatively with F/GR. TTA was found positivelycorrelated with sensory texture variables T/FN and T/FI. Finally, F,was positively correlated with T/CH, T/CR and T/FI.

In general, the sensory variables show some degree of correla-tion with other varieties of the same group (Flavour, Texture or

Table 5Mean for each apple cultivar of instrumental quality variables and sensory scores from trained panel used in the principal component analysis. (SSC: Soluble Solid Contents%, TTA: Total Titrable Acidity, g/L, F: Firmness, N, F/AF: Flavour-Apple Flavour, F/FL: Flavour-Flowery, F/FR: Flavour-fruitness, F/GR: Flavour: grassiness, F/OF: Flavour-Overall Flavour, F/SO: Flavour-Sourness, F/SW: Flavour-Sweetness, O/FL: Odour-Flowery, O/FR:Odour-Fruity, O/GR: Odour-Grassiness, O/OS: Odour-Overall Smell, T/CH: Texture-Chewiness, T/CR: Texture-Crispness, T/FN: Texture-Firmness, T/FI: Texture-Fineness, T/JU: Texture-Juiciness, T/ME: Texture-Mealiness, T/TP: Texture-Toughness Peel).

SSC(�Brix)

TTA(g/L)

F(N)

F/AF(0–10)

F/FL(0–10)

F/FR(0–10)

F/GR(0–10)

F/OF(0–10)

F/SO(0–10)

F/SW(0–10)

O/FL(0–10)

O/FR(0–10)

O/GR(0–10)

O/OS(0–10)

T/CH(0–10)

T/CR(0–10)

T/FN(0–10)

T/FI(0–10)

T/JU(0–10)

T/ME(0–10)

T/TP(0–10)

ARIANE cov 16.4 8.5 82.3 4.7 1.1 4.4 1.0 6.0 6.6 4.5 1.0 3.2 1.4 4.1 5.9 6.8 4.8 7.2 5.2 1.0 4.5FUJI 14.3 2.8 53.3 3.4 1.7 3.5 0.7 4.6 1.4 5.8 1.2 4.0 1.2 5.4 4.0 6.9 3.2 4.9 7.6 0.8 3.9JONAGOLD 11.6 4.5 54.9 2.9 0.8 2.5 2.0 4.2 3.3 3.4 0.8 3.1 1.8 4.8 2.9 5.1 2.7 2.9 6.5 1.5 3.9WELLANT� CPRO-

4713.7 4.6 54.9 4.2 0.9 4.0 0.7 4.8 3.7 4.8 1.0 2.9 1.7 4.8 3.2 5.7 3.1 4.1 6.2 0.8 4.9

RUBENS� CIVni cov 14.6 5.2 61.7 5.0 1.6 5.2 0.5 5.7 3.9 5.7 0.9 3.3 1.2 4.2 2.9 5.9 3.0 5.1 6.3 1.0 2.6GOLDEN

DELICIOUS14.7 4.1 48.0 4.1 0.9 4.2 0.4 5.0 2.1 5.3 1.3 4.5 1.0 5.4 1.9 3.8 3.1 2.4 4.3 3.3 2.9

GOLDCHIEF�

Goldpink cov16.4 5.0 59.8 5.3 1.3 5.7 0.4 5.9 3.4 6.6 0.8 4.4 0.7 5.4 3.8 6.0 3.3 4.3 6.0 1.5 3.6

JUNAMI� Milwacov

14.5 5.8 67.6 5.0 1.0 4.5 1.0 5.9 4.3 5.5 0.5 2.8 2.0 4.8 4.1 6.9 3.0 5.7 6.5 1.1 4.1

LIGOL 12.3 5.0 62.7 3.3 0.6 3.1 1.4 4.4 4.1 3.3 0.7 3.0 1.4 4.4 4.4 7.2 4.0 6.1 6.4 0.8 3.7KANZI� Nicoter

cov12.7 6.8 61.7 4.0 1.2 3.5 1.4 5.6 5.5 4.5 1.0 3.2 2.9 5.3 3.4 7.7 3.4 5.8 7.7 0.3 3.9

PINK LADY� CrippsPink cov

14.4 5.2 65.7 4.7 1.9 5.2 0.4 6.1 3.4 6.0 1.6 5.0 0.9 5.8 4.0 5.8 3.4 5.1 5.7 1.8 4.1

Mean 14.1 5.2 61.7 4.2 1.2 4.2 0.9 5.3 3.8 5.0 1.0 3.6 1.5 4.9 3.7 6.2 3.3 4.9 6.2 1.3 3.8Standard Deviation 1.5 1.5 8.8 0.8 0.4 1.0 0.5 0.7 1.4 1.0 0.3 0.8 0.6 0.5 1.0 1.1 0.6 1.4 1.0 0.8 0.7Minimum 11.6 2.8 49.0 2.9 0.6 2.5 0.4 4.2 1.4 3.3 0.5 2.8 0.7 4.1 1.9 3.8 2.7 2.4 4.3 0.3 2.6Maximum 16.4 8.5 82.3 5.3 1.9 5.7 2.0 6.1 6.6 6.6 1.6 5.0 2.9 5.8 5.9 7.7 4.8 7.2 7.7 3.3 4.9

Table 6Correlation matrix among variables (sensory and instrumental quality variables) included in principal component analysis.

Variables SSC TTA F F/AF F/FL F/FR F/GR F/OF F/SO F/SW O/FL O/FR O/GR O/OS T/CH T/CR T/FN T/FI T/JU T/ME T/TP

SSC 1 0.286 0.437 0.781 0.364 0.813 �0.733 0.711 0.117 0.710 0.154 0.386 �0.564 0.010 0.344 �0.079 0.352 0.240 �0.468 0.236 �0.036TTA 0.286 1 0.757 0.420 �0.147 0.198 0.214 0.588 0.967 �0.208 �0.179 �0.323 0.352 �0.473 0.595 0.426 0.646 0.688 �0.151 �0.280 0.269F 0.437 0.757 1 0.367 0.201 0.251 0.043 0.557 0.710 �0.008 �0.065 �0.190 0.030 �0.426 0.920 0.628 0.767 0.896 0.017 �0.436 0.370F/AF 0.781 0.420 0.367 1 0.341 0.939 �0.658 0.914 0.311 0.723 �0.013 0.207 �0.272 �0.017 0.164 0.030 0.098 0.297 �0.313 0.087 �0.084F/FL 0.364 �0.147 0.201 0.341 1 0.510 �0.562 0.481 �0.246 0.711 0.647 0.577 �0.327 0.450 0.065 0.119 �0.098 0.146 0.215 �0.089 �0.101F/FR 0.813 0.198 0.251 0.939 0.510 1 �0.811 0.839 0.068 0.840 0.198 0.465 �0.542 0.127 0.076 �0.119 0.052 0.160 �0.378 0.222 �0.212F/GR �0.733 0.214 0.043 �0.658 �0.562 �0.811 1 �0.514 0.346 �0.875 �0.479 �0.620 0.648 �0.345 0.123 0.279 0.022 0.065 0.379 �0.358 0.241F/OF 0.711 0.588 0.557 0.914 0.481 0.839 �0.514 1 0.448 0.646 0.159 0.281 �0.123 0.090 0.320 0.201 0.234 0.459 �0.202 0.008 0.009F/SO 0.117 0.967 0.710 0.311 �0.246 0.068 0.346 0.448 1 �0.349 �0.311 �0.508 0.473 �0.584 0.588 0.529 0.614 0.718 0.004 �0.457 0.321F/SW 0.710 �0.208 �0.008 0.723 0.711 0.840 �0.875 0.646 �0.349 1 0.349 0.621 �0.512 0.500 �0.119 �0.149 �0.241 �0.075 �0.117 0.243 �0.215O/FL 0.154 �0.179 �0.065 �0.013 0.647 0.198 �0.479 0.159 �0.311 0.349 1 0.757 �0.349 0.595 �0.124 �0.281 0.077 �0.145 �0.231 0.352 �0.018O/FR 0.386 �0.323 �0.190 0.207 0.577 0.465 �0.620 0.281 �0.508 0.621 0.757 1 �0.664 0.748 �0.211 �0.430 �0.072 �0.327 �0.371 0.623 �0.295O/GR �0.564 0.352 0.030 �0.272 �0.327 �0.542 0.648 �0.123 0.473 �0.512 �0.349 �0.664 1 �0.159 �0.017 0.483 �0.091 0.219 0.568 �0.528 0.277O/OS 0.010 �0.473 �0.426 �0.017 0.450 0.127 �0.345 0.090 �0.584 0.500 0.595 0.748 �0.159 1 �0.364 �0.248 �0.407 �0.447 0.077 0.384 0.029T/CH 0.344 0.595 0.920 0.164 0.065 0.076 0.123 0.320 0.588 �0.119 �0.124 �0.211 �0.017 �0.364 1 0.684 0.824 0.857 0.109 �0.504 0.554T/CR �0.079 0.426 0.628 0.030 0.119 �0.119 0.279 0.201 0.529 �0.149 �0.281 �0.430 0.483 �0.248 0.684 1 0.459 0.851 0.710 �0.881 0.364T/FN 0.352 0.646 0.767 0.098 �0.098 0.052 0.022 0.234 0.614 �0.241 0.077 �0.072 �0.091 �0.407 0.824 0.459 1 0.749 �0.211 �0.241 0.307T/FI 0.240 0.688 0.896 0.297 0.146 0.160 0.065 0.459 0.718 �0.075 �0.145 �0.327 0.219 �0.447 0.857 0.851 0.749 1 0.270 �0.669 0.324T/JU �0.468 �0.151 0.017 �0.313 0.215 �0.378 0.379 �0.202 0.004 �0.117 �0.231 �0.371 0.568 0.077 0.109 0.710 �0.211 0.270 1 �0.808 0.212T/ME 0.236 �0.280 �0.436 0.087 �0.089 0.222 �0.358 0.008 �0.457 0.243 0.352 0.623 �0.528 0.384 �0.504 �0.881 �0.241 �0.669 �0.808 1 �0.413T/TP �0.036 0.269 0.370 �0.084 �0.101 �0.212 0.241 0.009 0.321 �0.215 �0.018 �0.295 0.277 0.029 0.554 0.364 0.307 0.324 0.212 �0.413 1

Bold values are statistically sifnificant from 0 with significance level alfa = 0,05.(SSC: Soluble Solid Contents%, TTA: Total Titrable Acidity, g/L, F: Firmness, N, F/AF: Flavour-Apple Flavour, F/FL: Flavour-Flowery, F/FR: Flavour-fruitness, F/GR: Flavour: grassiness, F/OF: Flavour-Overall Flavour, F/SO: Flavour-Sourness, F/SW: Flavour-Sweetness, O/FL: Odour-Flowery, O/FR:Odour-Fruity, O/GR: Odour-Grassiness, O/OS: Odour-Overall Smell, T/CH: Texture-Chewiness, T/CR: Texture-Crispness, T/FN: Texture-Firmness, T/FI: Texture-Fineness, T/JU: Texture-Juiciness, T/ME: Texture-Mealiness, T/TP: Texture-Toughness Peel).

J.Bonanyet

al./FoodQ

ualityand

Preference32

(2014)317–

329321

Table 7Percentage of variance and % accumulated variance for each principal componentafter VARIMAX transformation.

D1 D2 D3

Variance (%) 30,885 29,594 16,072Accumulated variance (%) 30,885 60,479 76,551

Table 8Contribution of each variable (%) to the different principal components afterVARIMAX transformation.

D1 D2 D3

SSC 2,535 8,950 3,559TTA 12,510 0,023 0,102F 12,232 0,660 1,285F/AF 3,047 8,097 1,732F/FL 0,309 10,510 4,821F/FR 0,945 11,760 2,379F/GR 0,195 12,174 1,831F/OF 4,343 8,554 0,075F/SO 12,421 0,664 0,086F/SW 0,370 14,061 0,021O/FL 1,544 4,981 0,031O/FR 2,634 8,523 1,321O/GR 0,434 5,222 6,932O/OS 6,183 3,800 1,066T/CH 10,120 0,112 2,588T/CR 4,982 0,067 17,653T/FN 9,319 0,011 0,016T/FI 11,198 0,193 5,325T/JU 0,150 0,587 25,794T/ME 2,809 0,779 18,988T/TP 1,720 0,273 4,396

SSC: Soluble Solid Contents %, TTA: Total Titrable Acidity, g/L, F: Firmness, kg, F/AF:Flavour-Apple Flavour, F/FL: Flavour-Flowery, F/FR: Flavour-fruitness, F/GR: Fla-vour: grassiness, F/OF: Flavour-Overall Flavour, F/SO: Flavour-Sourness, F/SW:Flavour-Sweetness, O/FL: Odour-Flowery, O/FR:Odour-Fruity, O/GR: Odour-Grassi-ness, O/OS: Odour-Overall Smell, T/CH: Texture-Chewiness, T/CR: Texture-Crisp-ness, T/FN: Texture-Firmness, T/FI: Texture-Fineness, T/JU: Texture-Juiciness, T/ME:Texture-Mealiness, T/TP: Texture-Toughness Peel.

322 J. Bonany et al. / Food Quality and Preference 32 (2014) 317–329

Odour groups). Within the Flavour variables group, F/AF andF/FR, F/AF and F/OF, F/AF and F/SW, F/FL and F/SW, F/FR andF/OF, F/FR and F/SW, F/OF and F/SW were positively correlatedwhereas F/GR and F/AF, F/GR and F/FR, F/GR and F/SW were nega-tively correlated.

When considering the group Odour sensory variables, the fol-lowing positive correlations were found: O/FL and O/FR, O/FR andO/OS while O/FR was negatively correlated with O/GR.

When Texture sensory variables are considered, the followingcouple of variables were statistically significantly positivelyrelated: T/CH and T/CR, T/CH and T/FN, T/CH and T/FI, T/CR andT/FI, T/CR and T/JU, T/FN and T/FI. On the contrary the followingpairs T/CR and T/ME, T/FI and T/ME, T/JU and T/ME were negativelycorrelated.

Some sensory variables showed also a correlation outside oftheir own group. Thus, between variables of flavour and odourwere found the following positive correlations: F/FL and O/FL,F/SW and O/FR, F/GR and O/GR whereas a negative correlationbetween F/GR and O/FR was found. Among variables of Flavourand Texture very few correlations were found. Only F/SO waspositively correlated with T/FN and T/FI. Likewise between odoursensory variables and texture sensory variables only a positivecorrelation was observed between O/FR and T/ME.

3.2. Sensory map

From the results of Principal Component Analysis it can be ob-served that all instrumental and sensory variables measured on the11 varieties can be summarized in 3 main orthogonal componentsaccounting for 76.55% of the total variance. In Fig. 1, a scree plotcan be observed and relative importance of the 3 main componentscan be assessed. In order to simplify the interpretation of thedifferent dimensions of the principal component analysis, theVARIMAX transformation (Kaiser, 1958) was applied. A new distri-bution of the total variance explained by each component whilemaintaining the accumulated variance for the first 3 components(D1, D2, D3) was obtained (Table 7). Although the varianceaccumulated by the first 3 components remains to 76.55% of thetotal, the distribution among the 3 components is morehomogeneous.

Observing Table 8, it can be interpreted that the first componentrepresents mainly the following variables: TTA, F/SO, F, T/FI, T/CHand T/FN. The second component would basically represent F/SW,F/GR, F/FR, F/FL. Finally the third component would represent

Fig. 1. Scree plot of principal component analysis.

T/JU, T/ME/ and T/CR. Therefore the sensory map will be repre-sented by 3 major dimensions. The first component or dimension(ACIDIY–FIRMNESS) is representing more acid and firm fruit inthe positive range and less acid and less firm in the negative range.The second dimension (SWEETNESS) is correlated with sweetness(F/SW) and fruity flavour/odour (F/FR, F/FL) in the positive rangeand grassiness, unripe or not matured fruit in the negative range(F/GR). Finally the third dimension (JUICINESS–CRISPNESS), mostprobably, is representing juiciness and crispness (T/JU, T/CR) inthe positive range and mealiness in the negative range (T/ME).

In Fig. 2 (depicting first and second components) and Fig. 3 (firstand third dimension), the position of the 11 cultivars tested in the3 dimensional sensory map can be observed.

ARIANE cov is well distinguished since it situated in an area inthe map indicating high values for the first component (acidity andfirmness) while situated in the mid range values for perceivedsweetness (first dimension).

PINK LADY� Cripps Pink cov, GOLDCHIEF� Goldpink cov andRUBENS� CIVni cov are varieties that are situated in the centre re-gion for acidity (both instrumental TTA and perceived F/SO) andFirmness (F) while at the same time are situated in the high valuesregion for the second dimension of perceived sweetness, fruity fla-vour and low values of perceived grassiness. As it concerns thethird component positively related to T/JU and T/CR and negativelycorrelated to F/GR, these varieties are located in the middle region.

KANZI� Nicoter cov, JUNAMI� Milwa cov, WELLANT� and LIGOLare located in the middle range values of the acidity–firmnessdimension and in the negative range of the second dimension

Fig. 2. Biplot of instrumental quality and sensory variables and cultivars in the first and second component sensory map after VARIMAX transformation. (SSC: Soluble SolidContents%, TTA: Total Titrable Acidity, g/L, F: Firmness, kg, F/AF: Flavour-Apple Flavour, F/FL: Flavour-Flowery, F/FR: Flavour-fruitness, F/GR: Flavour: grassiness, F/OF:Flavour-Overall Flavour, F/SO: Flavour-Sourness, F/SW: Flavour-Sweetness, O/FL: Odour-Flowery, O/FR:Odour-Fruity, O/GR: Odour-Grassiness, O/OS: Odour-Overall Smell,T/CH: Texture-Chewiness, T/CR: Texture-Crispness, T/FN: Texture-Firmness, T/FI: Texture-Fineness, T/JU: Texture-Juiciness, T/ME: Texture-Mealiness, T/TP: Texture-ToughnessPeel) (ARI: Ariane; FUJ: Fuji; GC: Gold Pink; GD: Golden Delicious; JON: Jonagold; JUN: Milwa; KAN: Nicoter; LIG: Ligol; PL: Cripps Pink; RUB: CIVni; WEL: Cpro-47).

J. Bonany et al. / Food Quality and Preference 32 (2014) 317–329 323

(sweetness), indicating relatively low sweetness values. At thesame time, KANZI� Nicoter cov is located in the high end valuesof third dimension (positively correlated to T/JU, T/CR), JUNAMI�

Milwa cov, WELLANT� in the middle range whereas LIGOL on thelower range of this component.

Finally, although GOLDEN DELICIOUS, JONAGOLD and FUJI aresituated in the negative range for the first dimension (acidity andfirmness) as regards to their classification for sweetness, GOLDENDELICIOUS is located in the mid range for sweetness; JONAGOLDis falling in the negative part and FUJI in the positive range of it.They also fall in different categories regarding third component(T/JU; T/CR). Whereas FUJI falls into the category of high valuesfor third dimension (juicy varieties), JONAGOLD is falling in themiddle range for this component, whereas GOLDEN DELICIOUS isfalling into the category of low values for juiciness and crispness.

A summarized description of the projection of the varieties inthe sensory map can be found in Table 9.

3.3. Cluster analysis. Grouping consumers

Consumers were grouped in clusters based in similarity of thescores given to the cultivars tested in common as it has been de-scribed in the methodology. Cluster analysis (SAS FASTCLUS proce-dure) was performed based on the Euclidean distances betweenconsumers calculated through the SAS MDS procedure. Six clusterswere selected based on the analysis of the percentage of explainedvariance from the total for each additional cluster considered.Frequencies in each cluster are described in Table 10. Four major

clusters, considering the size of each cluster, were detected:Cluster 2, 4, 1 and 3 with two other minor clusters: Clusters 5 and 6.

Average hedonic scores for each cultivar for all consumers ineach cluster and the cultivar preferences within each cluster canbe observed in Table 11. Cluster 1 (21% of consumers), Cluster 2(38%) and Cluster 5 (3%) show the same varieties for the 4 mostpreferred varieties: GOLDCHIEF� Goldpink cov, RUBENS� CIVnicov, FUJI, PINK LADY� Cripps Pink cov although with changes inthe order of preference among them. Cluster 6 (6%) is showing alsoFUJI, GOLDCHIEF� Goldpink cov and PINK LADY� Cripps Pink covwithin the first four positions but includes GOLDEN instead ofRUBENS� CIVni cov in the four most preferred. These clusters havealso in common that they show JONAGOLD and ARIANE cov as theleast preferred varieties, with the exception of Cluster 6 that isKANZI� Nicoter cov and ARIANE cov the worst appreciated culti-vars for the consumers in this group. Flavour acceptance scoresfor Cluster 1, 2, 5 and 6 for cultivars assessed show a similar pat-tern. However, consumers in cluster 5 tend to score the varietieslower than consumers in clusters 1 or 2. Consumers in cluster 6,although showing a similar pattern to those in cluster 1, 2 and 5are somehow different in the sense that consumers in this clusterseem to show a higher preference for GOLDEN DELICIOUS thanconsumers in cluster 1, 2 or 5. These four clusters of similar pattern(Cluster 1, 2, 5, 6) are considered to be part of a larger group of con-sumers called megacluster A with 68% of consumers that took partin the test.

On the other hand, Cluster 3 (10% of consumers) and Cluster 4(22%) show also the same four varieties in the same order of

Fig. 3. Biplot of instrumental quality and sensory variables and cultivars in the first and third component sensory map after VARIMAX transformation. (SSC: Soluble SolidContents %, TTA: Total Titrable Acidity, g/L, F: Firmness, kg, F/AF: Flavour-Apple Flavour, F/FL: Flavour-Flowery, F/FR: Flavour-fruitness, F/GR: Flavour: grassiness, F/OF:Flavour-Overall Flavour, F/SO: Flavour-Sourness, F/SW: Flavour-Sweetness, O/FL: Odour-Flowery, O/FR:Odour-Fruity, O/GR: Odour-Grassiness, O/OS: Odour-Overall Smell,T/CH: Texture-Chewiness, T/CR: Texture-Crispness, T/FN: Texture-Firmness, T/FI: Texture-Fineness, T/JU: Texture-Juiciness, T/ME: Texture-Mealiness, T/TP: Texture-ToughnessPeel) (ARI: Ariane; FUJ: Fuji; GC: Gold Pink; GD: Golden Delicious; JON: Jonagold; JUN: Milwa; KAN: Nicoter; LIG: Ligol; PL: Cripps Pink; RUB: CIVni; WEL: Cpro-47).

324 J. Bonany et al. / Food Quality and Preference 32 (2014) 317–329

preference as the most four preferred cultivars: KANZI� Nicotercov, ARIANE cov, JUNAMI and RUBENS. They also coincide withthe fact that is GOLDEN DELICIOUS the least preferred variety. Bothgroups of consumers show a similar pattern of acceptance ofvarieties with small differences in the order of preference. Themain difference is that Cluster 4 tends to score higher the varietiesthan Cluster 3 (Table 11). These two clusters are considered to bepart also of a larger group called megacluster B (32% of consumers).

3.4. Preference Map

Average preferences for each cultivar were modelled for eachcluster as a function of the main components (D1, D2 and D3)using PREFMAP method included in XLSTAT software. Linear, circu-lar, elliptical and quadratic model were tested. For cluster 1, circu-lar model was retained whereas for the other clusters a linear orvector model was the only model statistically significant. A statis-tical summary for each cluster can be found in Table 12.

In Figs. 4 and 5, bi-dimensional Preference Maps can be ob-served. These maps include position of the ideal point and isopre-ference curves for the circular model for cluster 1 and vectorsindicating the direction of maximum preference for each one ofthe linear models for clusters 2, 3, 4, 5 and 6. The position or direc-tion for each one of the clusters, confirm the existence of two maingroups or megaclusters as already stated previously. One group ofconsumer clusters, megagroup A includes C1, C2, C5 and C6whereas the other group is made of C3 and C4 clusters.

Within the first megagroup, models of acceptance prediction forclusters C2 and C5, accounting for 41% of consumers, are linearmodels. These models are represented in the different bi-dimen-sional plots as vectors indicating direction of maximum accep-tance. As it can be observed in Fig. 4, the vector models are quitealigned with D2 (sweetness component) pointing to the positiverange of it, indicating that consumers in these clusters prefer applecultivars with high values of attributes positively correlated (F/SW,F/FR, F/FL) and negatively correlated (F/GR) being relatively inde-pendent of D1 (acidity–firmness) and D3 (juiciness–crispness)and its correlated variables TTA, F/SO, F, T/FI and T/CH for D1 andT/JU, T/ME/ and T/CR for D3 respectively. The varieties whichprojection into the bi–dimensional preference map plot (D1, D2)is closer to the vector of preference for these consumer clustersare GOLDCHIEF� Goldpink cov, RUBENS� CIVni cov, FUJI, PINKLADY� Cripps Pink cov. These two clusters of consumers prefervarieties with high values of sweetness (F/SW and SSC), fruitiness(F/FR) and flowery (F/FL) and low values of grassiness (F/GR). Atthe same time their preference is relatively independent ofcomponent D1 (acidity–firmness) and D3 (juiciness–crispness)provided these values are in a middle range values. Comparingthe two clusters C2 and C5, in fact C5 is very minor clusterregarding number of consumers (3%), and besides a slight changein the preference direction compared to C2, consumers in C5scored on average much lower all varieties compared to consumersin C2.

Cluster C1 with of 21% of consumers is the only clusterwhere the retained model is a circular model with the optimum

Table 9Classification of each of the 11 varieties included in the consumer test regarding their position in the sensory map.

First component-dimension (D1)Total Tritable Acidity/TTA (+)Firmness (F) (+)Flavour Sourness F/SO (+)Texture Fineness (F/FI) (+)Texture Chewiness (+)

Second component-dimension (D2)Flavour Sweetness (F/SW) (+),Flavour Fruitness (F/FR) (+)Flavour Floweriness (F/FL) (+)Flavour Grassiness (F/GR) (�)

Third component-dimension (D3)Texture Juiciness (T/JU) (+)Texture Crispness (T/CR) (+)Texture Mealiness (T/ME) (�)

+ ARIANE cov + +±�

± ARIANE cov +± ARIANE cov�

� +±�

± PINK LADY� Cripps Pink covRUBENS� CIVni covKANZI� Nicoter covJUNAMI� Milwa covWELLANT� CPRO47LIGOLGOLDCHIEF� Goldpink cov

+ PINK LADY� Cripps Pink covGOLDCHIEF� Goldpink covRUBENS� CIVni cov

+± PINK LADY� Cripps Pink cov

GOLDCHIEF� Goldpink covRUBENS� CIVni cov

�± KANZI� Nicoter cov

JUNAMI� Milwa covWELLANT� CPRO47

+ KANZI� Nicoter cov± JUNAMI� Milwa cov

WELLANT� CPRO47�

� LIGOL +±�

� FUJIGOLDEN DELICIOUSJONAGOLD

+ FUJI + FUJI±�

± GOLDEN DELICIOUS +±� GOLDEN DELICIOUS

� JONAGOLD +± JONAGOLD�

Table 10Cluster analysis results.

Cluster Frequency % RMS Closest cluster

1 902 21 0.0624 22 1631 38 0.0725 63 436 10 0.0709 24 944 22 0.0660 25 121 3 0.0700 36 256 6 0.0760 2

J. Bonany et al. / Food Quality and Preference 32 (2014) 317–329 325

situated in high values of D2 (sweetness), low values for D1(acidity–firmness) and mid range values for D3 (juiciness–crispness).For these consumers, contrarily to C2 and C5 clusters where thehigher the values for sweetness component higher the acceptance,

Table 11Average score of each cultivar for all consumers in the different clusters.

Cultivar Clusters

C1 C2

ARIANE cov 6.42 fa 5.08 fFUJI 7.87 ab 6.78 abJONAGOLD 7.07 e 5.10 fWELLANT� CPRO-47 7.60 abcd 5.93 cdRUBENS� CIVni cov 7.92 a 6.57 bGOLDEN DELICIOUS 7.76 abc 6.51 bGOLDCHIEF� Gold Pink cov 8.00 a 6.98 aJUNAMI� Milwa cov 7.49 bcd 6.21 cLIGOL 7.40 cde 5.85 deKANZI� Nicoter cov 7.21 de 5.58 ePINK LADY� Cripps Pink cov 7.79 abc 6.66 b

a LSMEANS in the same column correspond to comparisons of the different varieties in

there is an optimal value after which the acceptance diminishes.The same comment is also valid for the acidity–firmnesscomponent. In this case, GOLDCHIEF� Goldpink cov, FUJI, and PINKLADY� Cripps Pink cov and RUBENS� CIVni cov are the varietieswith projected values on the bi-dimension (D1, D2) preferencemap closer to preference vector for this cluster.

Regarding consumer cluster C6 (6% of consumers), the vector ofmaximum preference has a significant projection in the positiverange of D2 (sweetness component) and in the negative range ofcomponent D1 (acidity–firmness) being relatively independent ofD3 (juiciness–crispness). FUJI and GOLDEN DELICIOUS are the vari-eties with a projection closer to this vector of preference.

As it concerns the clusters included in megagluster B, clusters 3and 4 with a total of 32% of consumers, the directions of maximumpreference are completely different from the previously described

C3 C4 C5 C6

6.05 b 7.35 b 3.11 c 3.12 g4.29 efg 5.53 f 5.09 a 7.73 a3.91 fg 5.90 f 2.93 c 5.10 ef4.41 ef 6.41 e 3.38 c 5.65 de5.74 b 7.04 bc 4.85 ab 6.22 cd3.85 g 5.01 g 4.72 ab 7.12 ab5.67 bc 6.90 cd 5.26 a 6.48 bc5.85 b 7.33 b 4.84 ab 4.92 ef5.12 cd 7.02 bc 3.41 c 4.90 f6.67 a 7.78 a 3.89 bc 3.59 g4.75 de 6.52 de 4.86 ab 6.32 cd

the same cluster. Means with different letters are significantly different (a = 0.05).

Table 12Statistical parameters of models obtained with PREFMAP (XLSTAT Software) method.

Cluster Model df SS MS R2 F Pr > F F-ratio Pr > F

C1 Circulara 4 9.114 2.279 0.911 15.434 0.003 13.716 0.010C2 Linearb 3 7.709 2.570 0.771 7.851 0.012C3 Linear 3 6.757 2.252 0.676 4.861 0.039C4 Linear 3 7.342 2.447 0.734 6.446 0.020C5 Linear 3 7.443 2.481 0.744 6.790 0.018C6 Linear 3 8.365 2.788 0.836 11.936 0.004

a Circular model: Y = a0 + a1 D1 + a2 D2 + a3 D3 + a4 D12 + a2 D22 + a3 D32, where Y is average acceptance score for each variety and D1, D2 and D3, main components fromPrincipal Component Analysis. Ideal point at D1 = 1,505, D2 = 1,109, D3 = �0,314.

b Linear model : Y = a0 + a1 D1 + a2 D2 + a3 D3, where Y is average acceptance score for each variety and D1, D2 and D3, main components from principal componentanalysis.

Fig. 4. Bidimensional preference map of components D1 and D2 with representation of each cultivar, the vector of maximum preference for clusters C2, C3, C4, C5 and C6 andisopreference curves according to circular model with ideal point for cluster C1. Additionally contour plot of estimated percentage of clusters above average preference isdepicted. (ARI: Ariane; FUJ: Fuji; GC: Gold Pink; GD: Golden Delicious; JON: Jonagold; JUN: Milwa; KAN: Nicoter; LIG: Ligol; PL: Cripps Pink; RUB: CIVni; WEL: Cpro-47).

326 J. Bonany et al. / Food Quality and Preference 32 (2014) 317–329

clusters. The main characteristic of these clusters is that the direc-tion of preference is in the direction of the positive range for D1(acidity–firmness), i.e. high values of the correlated variablesTTA, F/SO, F, T/FI and T/CH. These clusters are basically indepen-dent of D2 (sweetness) and relatively positively correlated withD3 (juiciness–crispness). KANZI� Nicoter cov, ARIANE cov,JUNAMI� Milwa cov and RUBENS� CIVni cov are the varieties withprojections more aligned with the preference vector in theseclusters. Besides an slight difference in the direction of preference,C3 cluster consumers tend to score lower the varieties than con-sumers in C4.

Bi-dimensional representations (Figs. 4 and 5) of preferencemap show also contour areas of estimated percentage of clusterswith above average preference. Observing the figures it can benoticed an area of maximum percentage (80–100%) of clusterswith preference above average. This area covers all positiverange of D2 (sweetness component) and small positive rangeof D1 (acidity–firmness). In the D1 (acidity–firmness), D3

(juiciness–cripsness) plane, area of maximum preference covers aportion of the positive range of D3 and short portion of negativerange of D3. Finally, on the D2, D3 plane, the area is defined bythe whole positive range of D2 (sweetness) and a significantportion of D3 (juiciness–crispness). GOLDCHIEF� Goldpink cov,and PINK LADY� Cripps Pink cov and RUBENS� CIVni cov fall intothe area of maximum acceptance.

4. Discussion

The sensory map with 3 main components explained 76.6% ofthe total variance of the sensorial and instrumental variables usedto describe the 11 varieties under test. First component is aboutacidity and firmness and shows a high correlation with bothinstrumental and sensorial acidity and firmness. The second com-ponent is related with sweetness and it is highly positively corre-lated with sensorial sweetness, flowery, fruitiness scores and to a

Fig. 5. Bidimensional preference map of components D1 and D3 with representation of each cultivar, the vector of maximum preference for clusters C2, C3, C4, C5 and C6 andisopreference curves according to circular model with ideal point for cluster C1. Additionally contour plot of estimated percentage of clusters above average preference isdepicted. (ARI: Ariane; FUJ: Fuji; GC: Gold Pink; GD: Golden Delicious; JON: Jonagold; JUN: Milwa; KAN: Nicoter; LIG: Ligol; PL: Cripps Pink; RUB: CIVni; WEL: Cpro-47).

J. Bonany et al. / Food Quality and Preference 32 (2014) 317–329 327

lesser extent to Soluble Solids content and negatively correlatedwith sensorial grassiness. Finally, the third component is positivelyrelated to juiciness and crispness and negatively with mealiness.These findings are at large, in concordance with previouspreference maps for apple published so far on apple preferencemap (Carbonell, Izquierdo, Carbonell, & Costell, 2008; Daillant-Spinnler et al., 1996; Jaeger et al., 1998; Popper, 1998). Attributeson sweetness, firmness and juiciness or related conform the mainaxis of the preference maps available. Daillant-Spinnler et al.,1996 in a study involving 12 varieties and 120 consumers in theUK, found two main components. First component was associatedto juiciness, crispness, hardness, green apple and fresh flavor andsecond dimension was related to sweetness and acidity (latterone also related to the 1st component). Jaeger et al., 1998, in astudy including Danish (117) and English (127) consumers with3 varieties at various levels of mealiness, found that the firstdimension was related flavor (sweetness, acidity) differences andsecond to texture differences mostly mealiness level. Popper,1998, found that two dimensions accounted for a significant partof the variability. The first component combined both taste attri-butes like sweetness and acidity with texture attributes like hard-ness (interpreted as firmness) while the second dimension wasrelated to juiciness and crispness. Finally, Carbonell et al., (2008)in a methodological trial involving 3 apple varieties and 3 levelsof 2 months storage temperature in Spain with 99 consumers,found also two main dimensions. All attributes of taste and texturemeasured where associated to both dimensions except for juici-ness, flouriness, granularity and taste intensity that were basicallyrelated to only the second dimension. Across different countrieswith different varieties, most of preference maps dimensions areassociated with flavor, sweetness and acidity, texture like hardnessor firmness and also juiciness and crispness. When only twodimensions are considered, sweetness and acidity might be partof the same dimension although in opposite directions. Also acidity

and firmness are often part of the same axis since there are biolog-ical reasons for this association (Harker et al., 2003).

Clustering of consumers yielded 6 consumer segments. Four ofthese clusters, megacluster A (cluster 1, 2, 5 and 6) accounting for68% of the consumers that took part of the test have in common forthe preference of a sweet apple whereas only 32% preferred a moreacidic variety. Clusters 2 and 5 are very similar in preference direc-tions and account 41% of the population. Consumers in these clus-ters are characterized by better acceptance for varieties withhigher sweetness, fruitiness and flowery scores, relatively indepen-dent of the acidity and firmness scores and mid range for juicinessand crispness. Cluster 1 (21% of consumers) is defined by an opti-mal point of preference on the range of high values for sweetness,relatively low values for acidity and firmness and mid range forjuiciness and crispness. Cluster 6 (6% of consumers) is character-ized by a group of people that dislikes acidity but maintain prefer-ence for high sweetness and mid range to high values for crispnessand juiciness. On the other hand in megacluster B, (clusters 3 and4) representing 32% of consumers prefers an acidic, firm, juicy andcrisp variety with mid range values for sweetness. These results arequite difficult to compare with previously reported preferencemaps because lack of reporting on relative importance of consumersegments, differences of countries and varieties. Nevertheless,Daillant and Spinnler et al., 1996, reported two segments of con-sumers, one of the groups preferring sweet, hard and juicy andthe other one liking more juicy and acidic apple. These two groupscould be roughly assimilated to megagroup A and B respectively.However, no report on relative importance of each group as per-centage of consumers is given. Results about consumer segmentsdescribed in Popper, 1998 and Carbonell et al., 2008, are also com-patible with results described in this paper in the sense that theyboth describe a group of consumers liking more acidic, hard andjuicy products and other groups with more prevalence of sweet-ness, aromatic, fruity products even if they are less firm and juicy.

328 J. Bonany et al. / Food Quality and Preference 32 (2014) 317–329

Carbonell et al., 2008 reports that the acidic, juicy and hard pre-ferred variety group accounts to 29% of the people in Spain. Resultson consumer segments reported by Jaeger et al., 1998 are less com-parable since the two consumer segments are based on one likingmost fresh apples (sweet, floral, fruity, hard, juicy and crisp) andthe other segment preferring also sweet but mealier apples. No re-port on percentages and relative importance of each consumer seg-ment is given either.

Preference mapping can be of practical use in different parts ofthe fruit value chain. With 62% of consumers, Cluster 1 (21% of con-sumers), Cluster 2 (38%) and Cluster 5 (3%), showing the same 4varieties as the most preferred varieties: GOLDCHIEF� Goldpinkcov, RUBENS� CIVni cov, FUJI, PINK LADY� Cripps Pink cov, and32% (Clusters 3 and 4), liking ARIANE cov and KANZI� Nicoter, itindicates very clearly the segmentation of the market in differentcountries in Europe. This would allow the positioning of these vari-eties or varieties with equal characteristics in the supermarketshelf and it will need a communication effort regarding their char-acteristics of flavour and texture according to consumer needs. An-other practical implication is related to fruit breeding programs.Either the breeding aims are concentrated in an area of the prefer-ence map with higher percentage of groups showing acceptanceabove average (Figs. 4 and 5) described as having positive valuesof dimension 1 (sweetness, fruitness, flowery), slightly positivevalues for dimension 2 (acidity and firmness) and moderate posi-tive values for dimension 3 (juiciness and crispness) or else thebreeding aims are concentrated in producing varieties withcharacteristics for each one of the consumer clusters, speciallythe larger ones.

Although, in this analysis instrumental measurements havebeen used as descriptors of variety products together with senso-rial evaluations by panelists, the preference dimensions are linearcombinations of both and in spite of good correlations betweeninstrumental and sensorial evaluations, it is not a simple task totranslate the preference map dimension values for preferred prod-ucts into practical standard measurements of quality (Harker et al.,2003). On the other hand, the preference map has been obtainedfrom a consumer test developed at a specific time in the season.Certainly, results would have changed if the test would have beencarried out earlier or later in the season although scale and direc-tion of change remains to be observed and it does not precludeutility of this point in time measurement.

5. Conclusions

A preference map for apple in different European countries hasbeen developed. The preference map encompasses 3 main dimen-sions (1-sweetness, fruitiness, flowery attributes, 2-acidity, firm-ness, 3-juiciness and crispness). The eleven varieties included inthe test have been placed into the map and consumers have beensegmented in 6 clusters according to their preferences. The 6 clus-ters are grouped into two main mega clusters A (68% of consumers)and B (32% of consumers). Megacluster A (Clusters 1, 2, 5 and 6) ischaracterized by preferring sweet apples. Clusters 2 and 5 (41% ofconsumers) liked sweet apple independently of acidity and firm-ness and moderate positive values on dimension of juiciness andcrispness. Cluster 1 (21% of consumers) has an optimal point inpositive values of sweetness dimension, moderate negative valuefor acidity and firmness and moderate positive value for juicinessand crispness. Cluster 6 (6% of consumers) besides preferring sweetvarieties dislike acid-firm varieties. As to regard to megacluster B(32% of consumers), they prefer varieties that are acidic-firm andjuiciy and crisp with values in the mid range of sweetnessdimension. In spite of the difficulties in translating preferencedimensions into standard practical values for fruit quality and

the fact of beign a punctual measurement of consumer behaviour,this preference map can be of practical use of different actors onthe fruit value chain likle marketers and breeders.

Acknowledgements

This work was supported through the ISAFRUIT project, fundedby the European Commission under the Thematic Priority 5 – FoodQuality and Safety of the 6th Framework Programme of RTD (Con-tract No. FP6-FOOD-CT-2006-016279). The views and opinions ex-pressed in this publication are purely those of the writers and maynot, in any circumstances, be regarded as stating an official posi-tion of the European Commission.

References

Allan-Wotjas, P., Sanford, K. A., McRae, K. B., & Carby, S. (2003). An integratedmicrostructural and sensory approach to describe apple texture. Journal of theAmerican Society for Horticultural Science, 128, 381–390.

Barendsee, V. (1993). Effect of harvesting time and handling period on quality ofapple. Fruit Belge, 48(392), 275–278.

Bonany, J., Buehler, A., Carbó, J., Codarin, S., Donati, F., Echeverria, G., et al. (2013).Consumer eating quality acceptance of new apple varieties in differentEuropean countries. Food Quality and Preference, 30(2), 250–259. ISSN0950-3293, http://dx.doi.org/10.1016/j.foodqual.2013.06.004, <http://www.sciencedirect.com/science/article/pii/S0950329313001018>.

Briz, T., Sijtsema, S. J., Jasiulewicz, A., Kyriakidi, A., Guardia, M. D., van Berg, I., et al.(2008). Barriers to fruit consumption: Driving forces behind consumerbehaviour. Scripta Horticulturae, 8(1), 7–18.

Callier, P., & Schlich, P. (1997). La cartographie des preferences incompletes –Validation par simulation. Sciences des Aliments, 17, 155–172.

Carbonell, L., Izquierdo, L., Carbonell, I., & Costell, E. (2008). Segmentation of foodconsumers according to their correlations with sensory attributes projected onpreference spaces. Food Quality and Preference, 19, 71–78.

Center for Disease Control (2002). National 5 a day partnership plan 2001–2004.Atlanta (GA), USA: National Center for Chronic Disease Prevention and HealthPromotion. <http://www.cdc.gov/nccdphp/5aday/>.

Cohen, N. L., Stoddard, A. M., Sarouhkhanians, S., & Sorensen, G. (1998). Barrierstoward fruit and vegetable consumption in a multiethnic worksite population.Journal of Nutrition Education, 30(6), 381–386.

Daillant-Spinnler, B., MacFie, H. J. H., Beyts, P. K., & Hedderley, D. (1996).Relationships between perceived sensory properties and major preferencedirections of 12 varieties of apples from the Southern Hemisphere. Food Qualityand Preference, 7, 113–126.

Department of Health. The NHS Plan – A plan for investment (2000). A plan forreform. London: The Stationery Office.

Ellinger, W. (1987). By which criteria are apples bought and how fruit from theSouth Tyrol evaluated? An analysis from the view point of the german market.Obstbau-Weinbau, 241, 4–8.

Gilmer, R. (2005). A new day for 5-A-day (vol. 98(2)). Willoughby: Florida Grower, p.31.

Harker, F. R., Gunson, F. A., & Jaeger, S. R. (2003). The case for fruit quality: Aninterpretive review of consumer attitudes, and preferences for apples.Postharvest Biology and Technology, 28, 333–347.

Höhn, E., & Güggenbühl, B. (1999). Golden Delicious: Festigkeit, Zuckergehalt undAkzeptanz bei der Konsumentenschaft. Schweizer Zeitschrift für Obst-undWeinbau, 14, 339–342.

Jaeger, S. R., Andani, Z., Wakeling, I. N., & MacFie, H. J. H. (1998). Consumerpreferences for fresh and aged apples: A cross-cultural comparison. Food Qualityand Preference, 9, 335–366.

Jaeger, S. R., Lund, C. M., Lau, K., & Harker, F. R. (2003). In search of the ‘ideal’ pear(Pyrus spp.): Results of a multidisciplinary exploration. Journal of Food Science,1108–1117.

Jaeger, S. R., Rossiter, K. L., Wismer, W. V., & Harker, F. R. (2003). Consumer-drivenproduct development in the kiwifruit industry. Food Quality and Preference, 14,187–198.

Kaiser, H. F. (1958). The varimax criterion for analytic rotation in factor analysis.Psychometrika, 23, 187–200.

Kellerhals, M., Duerr, P., Wittmann, P., Goerre, M., Hoehn, E., & Daetwyler, D. (1999).Qualität neuer Apfelsorten: Sensorische und analytische Aspekte.Agrarforschung, 6(2), 53–56.

Konopacka, D., Jesionkowska, K., Rutkowski, K., Płocharski, W., & Tomala, K. (2006).A comparison of the quality of scab resistant apples in expert and consumerevaluation. Vegetable Crop Research Institute, 65, 185–194.

Lock, K., Pomerleau, J., Causer, L., Altmann, D. R., & McKee, M. (2005). The globalburden of disease attributable to low consumption of fruit and vegetables:Implications for the global strategy on diet. Bulletin of the World HealthOrganization, 83(2), 100–108. <http://www.scielosp.org/scielo.php?script=sci_arttext&pid=S0042-96862005000200010&lng=en&nrm=iso>, ISSN 0042-9686,http://dx.doi.org/10.1590/S0042-96862005000200010.

Mante, W. (1973). Consumer preferences in apple purchase by 762 Berlinhouseholds. Erwerbobstbau, 157, 101–102.

J. Bonany et al. / Food Quality and Preference 32 (2014) 317–329 329

PanelCheck software. (2006). Nofima Mat, A � s, Norway. <http://www.panelcheck.com>.

Popper, R. (1998). Consumer segmentation in sensory research. In Measurement ofConsumer Attitudes and their Influence on Food Choice, Report series from EUproject AIR-CAT, pp. 34–38.

Robertson, A., Tirado, C., Lobstein, T., Jermin, M., Knai, C., Jensen, H. J., et al. (2004).Food and health in Europe: A new basis for action. WHO regional publications,European series (No. 96). Copenhagen: World Health Organization. <http://www.euro.who.int/document/E82161.pdf>.

Schlich, P., & McEwan, J. A. (1992). Cartographie des Préférences. Un outil statistiquepour l’industrie agro-alimentaire. Science des Aliments, 12, 339–355.

Subar, A. F., Heimendinger, J., Patterson, B. H., Krebs-Smith, S. M., Pivonka, E., &Kessler, R. (1995). Fruit and vegetable intake in the United States: The baselinesurvey of the five a day for better health program. American Journal of HealthPromotion, 9(5), 352–360.

Tomala, K., Barylko-Pikielna, N., Jankowski, P., Jeziorek, K., & Wasiak-Zys, G. (2009).Acceptability of scab-resistant versus conventional apple cultivars by Polishadult and young consumers. Journal of the Science of Food and Agriculture, 89,1035–1045.

van de Abeele, P., & Reijnders, C. (1980). Survey on consumption of apples inBelgium during a promotion campaign in schools. Fruit Belge, 48(392),271–274.

Villatoro, C., López, M. L., Echeverria, G., Graell, J., & Lara, I. (2009). Effect ofcontrolled atmospheres and shelf life period on concentrations of volatilesubstances released by ‘Pink Lady�’ apples and on consumer acceptance. Journalof the Science of Food and Agriculture, 89, 1023–1034.

Wakeling, I. N., & MacFie, H. J. H. (1995). Designing consumer trials balanced forfirst and higher orders of carry-over effect when only a subset of k samples fromt may be tested. Food Quality and Preference, 6, 299–308.

WHO. (2003). Diet, nutrition and the prevention of chronic diseases. Report of aJoint WHO/FAO Expert Consultation. WHO technical report series (Vol. 916).World Health Organization, Geneva.

Yeh, M., Ickes, S. B., Lowenstein, L. M., Shuval, K., Ammerman, A. S., Farris, R., et al.(2008). Understanding barriers and facilitators of fruit and vegetableconsumption among a diverse multi-ethnic population in the USA. HealthPromotion International, 23(1), 42–51.