You are on page 1of 12

104 Varietal aroma properties of Spanish white wines Australian Journal of Grape and Wine Research 14, 104–115,

14, 104–115, 2008

Aroma properties of young Spanish monovarietal white wines:


a study using sorting task, list of terms and frequency
of citation
E. CAMPO1,2, B.V. DO2, V. FERREIRA1 and D. VALENTIN2
1
Laboratory for Flavor Analysis and Enology, Department of Analytical Chemistry, Faculty of Sciences, University of
Zaragoza, 50009 Zaragoza, Spain
2
UMR CSG 5170 CNRS INRA UB, CESG, 15 Rue Hugues Picardet, 21000 Dijon, France
Corresponding author:Vicente Ferreira, fax +34 976 761292, email vferre@unizar.es

Abstract
Background and Aims: Little is known about the distinguishing role of variety on the sensory space of
commercial white monovarietal Spanish wines. The major aims of the present research were to define the
sensory space of these wines and to assess the actual influence of grape variety by means of a two-step
sensory methodology.
Methods and Results: Similarity between wines was first studied by a sorting task followed by
multidimensional scaling analysis. Then, sensory descriptive analysis (DA) based on citation frequencies
was performed on individual products. Descriptive data were analysed by correspondence analysis and
hierarchical cluster analysis. Verdejo and Sauvignon Blanc were the only varietals similarly perceived and
described by the panel, being tropical fruit as their major descriptor. No clear grouping on the basis of
grape variety was observed for the rest of the products, which were clustered into three different groups.
The first group contained wines with fruity (white fruit), spicy and floral properties, but the two
others contained wines defined with untypical and unfavourable descriptors (vegetal, leather, animal,
undergrowth).
Conclusions: Sorting task coupled with DA based on citation frequencies is a suitable tool to classify
samples and finely describe wine aroma. Leaving aside Verdejo and Sauvignon Blanc, grape variety is not
a major factor in defining the aroma properties of Spanish commercial wines; some of which develop
some unfavourable descriptors.
Significance of the Study: The role of neutral varietals to increase the diversity of the sensory space of
wines should be revisited.

Keywords: citation frequency, descriptive analysis, sorting task, Spanish varietal, wine

Introduction commercial wines prepared at relatively small scale. One


During the last decade, professionals from different of the first studies was carried out in the mid-1990s by De
Spanish ‘denominations of origin’ have carried out exten- la Presa-Owens and Noble (1995), who evaluated the
sive work on retrieving native white grape varieties for aroma properties of wines from Macabeo, Parellada and
producing distinctive wines. Such varieties are mainly Xarello from the Penedès region. Later, Gutierrez Afonso
employed for the elaboration of young 100% varietals, et al. (1998) described wines from Verdejo, Listán Blanco
which rarely undergo barrel ageing. Grape variety is, sup- and Gual, originating from the Canary Islands. More
posedly, a major factor determining the final character recently, Vilanova and Vilarino (2006) have presented a
and quality in these kinds of products, which explains the series of studies concerning wines from Albariño and
interest of scientists to characterise such varietals from a Godello (Vilanova 2006), both of them exclusively located
sensory point of view. in the Galician region. In all these studies, the samples
Compared with foreign varietals, especially Chardon- were analysed by descriptive analysis (DA), and data were
nay and Sauvignon Blanc, little data are available with processed by principal component analysis. However, and
regard to the aroma properties of traditional Spanish to the best of our knowledge, no comparative study
young white wines. Most of the published work has dealing with the aromatic sensory profiles of a wide range
focused on wines originating from a single and well- of Spanish commercial young white varietals (from native
delimited cultivar location using sometimes non- or foreign origin grapes) has been published to date.
doi: 10.1111/j.1755-0238.2008.00010.x
© 2008 Australian Society of Viticulture and Oenology Inc.
Campo et al. Varietal aroma properties of Spanish white wines 105

There are several relevant considerations about the been recently pointed out by Koussissi et al. (2007).
role of variety in the Spanish wine market, some of which These authors have noted that, in general, most sensory
are also common to many other winemaking regions. The panels describing wine with classical sensory methodol-
first one is that the Spanish wine industry is fragmented ogy have difficulties with aroma terms and that the use of
into many relatively small winemaking regions, some of large lists of aroma descriptors does not guarantee a good
which share varietals but largely differ in climate and in discrimination between samples. In the present work, a
viticultural and enological practices. A second issue is the two-step methodology is proposed as an alternative to
lack of experience and tradition in the production and classical DA. The methodology includes a free sorting
marketing of monovarietal white wines, as most of them task, followed by a DA based on citation frequencies.
were sold as blends in the past. A third, but also very Such methodology has been recently applied to wine
important, consideration is that most of the Spanish vari- (Piombino et al. 2004, Parr et al. 2007), although in both
etals are non-floral and rather neutral. All these factors cases, the type of samples and the goals of the study were
acting concurrently may help in explaining why there is quite specific (looking for berry fruit notes or for typical-
no commonly accepted definition of what are the charac- ity), while here, the methodology will be used as a
teristic attributes of most of these varietals. In fact, even general screening technique.
the sensory space of common Spanish white wines is not The sorting procedure explores the similarity/
well defined, which means that the diversity of sensory dissimilarity among a set of products. Subjects are asked to
nuances is not really well known, nor of course is it examine a set of samples and to group them according to
possible to assess the contribution of variety to such a similar property (odour, flavour, colour, shape, etc.).
sensory diversity. As a result, it is not possible to make a Generally, subjects are also asked to briefly characterise
rational prospect about the role that variety could play in the groups once they have been established. This task is
the future, which seriously hinders the effectiveness of any ‘especially well adapted for the evaluation of food prod-
research and development plan undertake to increase the ucts as it is rapid and produces little fatigue and boredom’
‘personality’ and specificity of the aroma of white wines. (Bijmot and Wedel 1993). It permits the examination of
From this point of view, it is crucial to know the large sets of products in a single session and has been
sensory space of monovarietal Spanish white wines, to already applied to study the aroma properties of a variety
assess the unifying and distinguishing role of variety on of foods (Lawless et al. 1995, Saint-Eve et al. 2004, Cartier
such space and to verify whether there is some other et al. 2006) and beverages (Falahee and MacRae 1997,
parameter exerting a major influence on the sensory Chollet and Valentin 2000, 2001, Piombino et al. 2004,
space. The present research intended to address some of Ballester et al. 2005, Pages 2005). For example, Piombino
the above issues. The specific goals were to know whether et al. (2004) reported that a sorting task followed by
commercial young wines of average quality made with multidimensional scaling (MDS) analysis was an efficient
the same varietal can be identified and classified as way of exploring the flavour similarities among 22 red
members of the same group; to evaluate the degree of wines. The sorting procedure leads to a better understand-
difference existing between wines made with different ing of the similarities between products and provides
varietals; and finally, to evaluate, in depth, the aroma meaningful information regarding holistic perception of
properties that better characterise such wines. samples. However, as it generates only group data, an
Classical evaluation of wine aroma consists of estab- additional step is necessary to evaluate in greater depth
lishing sensory profiles by means of DA, as evidenced by the sensory characteristics of the individual samples.
the abundant scientific literature existing to this respect Such evaluation can be carried out with the frequency
(Guinard and Cliff 1987, Heymann and Noble 1987, of citation method, which has been previously applied to
Noble and Shannon 1987, De la Presa-Owens and Noble evaluate wine odour attributes (McCloskey et al. 1996, Le
1995, Gutierrez Afonso et al. 1998, Falque et al. 2004). Fur et al. 2003, Piombino et al. 2004). In this technique,
According to this methodology, a relatively low number subjects have to simply select the most pertinent odour
of aroma terms (between 5 and 16) are quantified using attributes from a list containing a relatively high number
an intensity rating scale by a panel of specifically trained of terms. Although this procedure is not exempt from
subjects. The two most frequent problems associated with problems related to the use of descriptors in a consensual
this methodology are the different interpretation that way, it avoids the use of intensity scales and makes it
subjects make of terms and the difficulty in the use of possible to obtain quite detailed descriptions of the prod-
intensity scales (Lawless 1999). These problems are partly ucts. We sought to establish whether this technique,
solved after thorough training, but such process can be when used in conjunction with a sorting task, would
time consuming and difficult when the samples being allow the determination of the real effect exerted by the
measured are heterogeneous and do not show specific grape variety on the aroma properties of a selection of
and easily identifiable descriptors. In this case, a large commercial Spanish white wines.
number of descriptors, many of which can be very diffi-
cult to evaluate in many of the samples, have to be
defined and used, which makes the work extremely com- Materials and methods
plicated. Reducing the number of descriptors makes the Panel
task easier, but reduces the accuracy of the analysis. A total of 36 people were recruited on the basis of their
Indeed, this is the case of wine aroma evaluation, as it has interest and their availability for approximately 10
© 2008 Australian Society of Viticulture and Oenology Inc.
106 Varietal aroma properties of Spanish white wines Australian Journal of Grape and Wine Research 14, 104–115, 2008

Table 1. Grape variety, origin, brand and code of the studied wines.

Grape varieties Origin Brand Code

Albariño Rías Baixas Alargo Albala


Albariño Rías Baixas Lusco Alblus
Albariño Rías Baixas Estación Enológica Ablest
Chardonnay Cariñena Care Chacar
Chardonnay Costes del Segre Raimat Charai
Chardonnay Somontano Enate Chaena
Godello Valdeorras Estación Enológica Godest
Godello Valdeorras Gaba do Xil Godsil
Godello Valdeorras Guitián Godgui
Macabeo Calatayud Viña Aldaba Macald
Macabeo (Viura) Rioja Sierra Cantabria Macsie
Macabeo Rioja El Coto Maccot
Sauvignon Blanc Rueda Marqués de Riscal Saumar
Sauvignon Blanc Rueda Sanz Sausan
Verdejo Rueda Marqués de Riscal Vermar
Verdejo Rueda Naia Vernai
Verdejo Rueda Villa Narcisa Vernar
Treixadura, Torrontés Ribeiro Gran Solbeira Tresol
Treixadura Ribeiro Villa Mein Tremei
Treixadura Ribeiro Estación Enológica Treest
Xarello Penedés Albert i Noya Xarain
Xarello Penedés Anna Vallformosa Xarann
Palomino Fina Cádiz Antonio Barbadillo Palbar

weeks (two 1-h sessions per week). A 1-h selection medium quality range (between 4 and 12 Euros/bottle).
session was devoted to choose the best panellists suited to All the bottles were sealed with natural cork stoppers.
performing descriptive tasks. Screening tests included the The wines were stored in a climate-controlled, dark
identification of ten common odours (grapefruit, apple, cellar maintained at 13°C. Four bottles of each wine were
coconut, rose, orange blossom, honeysuckle, geranium, available for this study. Wine bottles were opened 30 min
asparagus and pepper) and free odour description of a prior to test sessions. After wines were verified free of
white and a red wine. Those candidates showing sensory cork taint by the panel leader, 20 mL samples were
deficiencies to identify basic odours or a manifest inability poured at room temperature (around 19°C) into dark
to verbal description were not selected to participate in INAO-approved wineglasses labelled with random three-
the study. The final panel of 32 judges (14 women, 18 digit codes and covered by plastic Petri dishes to allow the
men, median age 32 years) consisted of students and staff volatiles to equilibrate in the headspace. The sessions took
of the University of Burgundy. They were habitual or place in a ventilated and air-conditioned tasting room.
occasional consumers of white wines (several times a Because we were interested only in the aroma profile of
month) and were not paid for their participation. wines and that previous works (Aubry et al. 1999, Ball-
ester et al. 2005, Lillo et al. 2005) showed a good corre-
Wines lation between orthonasal and retronasal evaluation of
Twenty-three commercial monovarietal young white wine aromas, the study was limited to an orthonasal
wines made from different grape varieties produced in evaluation. This had the advantage of reducing judge
different Spanish regions were studied (see Table 1). In all fatigue as well as alcohol consumption.
cases, fermentation and maturation were carried out in
stainless steel tanks without contact with wood. Wines Sorting task
were all non-sparkling from the 2004 vintage and with
less than 2 g/L of residual sugar. Three different wines Procedure. Each panellist participated individually in one
from each grape variety were included in the study, 45-min session. First, the panellists were required to smell
except in the case of Xarello (two samples), Sauvignon each of the 23 wines once in the proposed order. After-
Blanc (two samples) and Palomino Fino (one sample) for ward, they could smell the samples as many times as they
which only reduced numbers of commercial monovari- wanted and in any order. The panellists were asked to sort
etals were available. The selection of samples was made as the wines into groups on the basis of odour similarity.
a random survey from commercial wines belonging to the No other instruction was given, except that they were
© 2008 Australian Society of Viticulture and Oenology Inc.
Campo et al. Varietal aroma properties of Spanish white wines 107

allowed to form as many groups as they wanted, and to their odour properties with the aid of the list. The session
put as many wines as they wished in each group. After ended by a discussion during which the panel leader
the completion of the task, the panellists were asked to highlighted the terms most frequently cited to describe
provide a few words to describe each of the groups they each wine. The wines selected for this training phase
had formed. The panellists were not informed about the presented intense and easily recognisable odour proper-
nature of the samples. ties and included young and barrel-aged red and white
wines of diverse grape varieties and origins. During the
Statistical analysis: MDS and hierarchical cluster analy- training, the panellists modified the initial list of terms by
sis (HCA). Sorting data were first explored by MDS eliminating those terms they considered irrelevant,
analysis. For each possible pair of wines, the number of ambiguous or redundant and by adding additional
times that two wines were sorted into the same group attributes they considered pertinent. At the end of the
was first counted across all subjects. The resulting general training, the list included 101 terms.
co-occurrence matrix was submitted to an MDS analysis The specific training consisted of three 1-h familiari-
in order to derive a spatial representation of wines (Schiff- sation sessions in which panellists described a total of 15
man 1976). A non-parametric scaling algorithm was used young white wines of similar characteristics (grape
(alternating least-square scaling). This procedure seeks variety and origin) as those of the study. The responses
iteratively a solution preserving the rank-order of the were compiled for all 15 wines, and those terms cited by
perceived similarities between products. The quality of fit less than 15% of the panel were eliminated from the list.
is measured by the stress value, which is the sum of the The final list included a total of 73 terms, together with
squares of differences between observed dissimilarities their corresponding odour reference standards, which is
and distances in the final configuration. Finally, coordi- presented in Table 2.
nates of samples in the MDS configuration were submit-
ted to HCA with the Ward criteria to aid the interpretation Wine evaluation. The 46 samples (23 wines ¥ two repeti-
of groupings of wines in the MDS charts. All the analyses tions) were divided into two blocks of 12 wines and two
were performed with the software SAS/STAT version 6.0 blocks of 11 wines with approximately the same propor-
(SAS Institute Inc., Cary, New York, USA). tion of varietals in each. Within a block, the samples
were presented simultaneously according to the William
Descriptive task Latin-square arrangement (FIZZ Biosystems, Couternon,
France) in order to reduce biases related to the order,
Procedure. Panel training. The panellists received des- carry-over or expectation effects (MacFie et al. 1989).
criptive sensory training in eight 1-h sessions over a Each member of the panel took part in four 1-h sessions
period of 2 weeks. They were provided with an initial list over 2 weeks, assessing one block per session. A 10-min
of 108 terms. The list was elaborated by compiling terms break was enforced in the middle of each session to limit
from other lists employed in the description of wines from panellist fatigue. The panellists were asked to smell the
different varieties and geographical origins (Noble et al. odour of wines and then to allocate to each wine a
1987, Peynaud and Blouin 1999, Aznar et al. 2003, Le maximum of six odour terms from the list of 73 terms. A
Fur et al. 2003, Valentin et al. 2003, Piombino et al. different bottle of wine was employed in each replication.
2004). Terms were arranged in the list by odour families: The panellists were not informed about the nature of the
fruity, floral, spices, vegetal, burnt, undergrowth and wines to evaluate.
others. The fruit family was subdivided into three catego-
ries ranging from general to specific terms (e.g. fruity, Data analysis. Individual panellists’ performance. To as-
white fruits and pear). The other families were not sub- sess the individual performance, an average reproducibil-
divided in categories. ity index (Ri) was calculated for each of the panellists:
During training, different reference standards repre-
sentative of aroma terms were presented. Commercially Ri = Σ [ 2 × descom (desrep1 + desrep2 )] n
available standards were taken from Sentosphère (Paris,
France), ‘Le Nez du Vin’ (Jean Lenoir, Provence, France) where descom is the number of common terms given by the
and Firmenich (Geneva, Switzerland). The standards not judge in the two replicates of a wine, desrep1 and desrep2 are
commercially available were prepared with natural prod- the number of terms given by the judge in the first and
ucts. The sensory properties of the former were stable second repetitions, respectively, and n is the number of
during the whole period of the study. The latter were wines.
prepared at the beginning of each day in order to guar-
antee good aroma quality. Terms citation frequency (Cf). All 73 terms were ranked
The training period was composed of two phases: according to their Cf percentage to identify the most rele-
general and specific training. Each of the five sessions of vant terms of each wine. Only terms cited by a minimum
general training was divided into two parts. In the first of five judges (15% of the panel) in, at least, one wine/re-
part (20 min), the panellists became familiar with the petition were considered for subsequent statistical analy-
specific vocabulary of the list and smelled different stan- ses. A chi-square (c2) analysis performed on the average Cf
dard aromatic references. In the second part, the panel- (two repetitions) of each term in the 23 wines was carried
lists had to evaluate four different wines and to describe out in order to identify the most discriminant terms.
© 2008 Australian Society of Viticulture and Oenology Inc.
108 Varietal aroma properties of Spanish white wines Australian Journal of Grape and Wine Research 14, 104–115, 2008

Table 2. List of terms used for descriptive measures and odour reference standards presented
during the training period.

Term Odour reference (source/brand) Quantity or concentration

Fruity
White fruits
Quince Quince jelly (Bon Maman) 1 teaspoon†
Pear Solution (Firmenich) 400 mg/L‡
Apple Solution (Firmenich) 400 mg/L‡
Yellow fruits
Apricot Solution (Firmenich) 500 mg/L‡
Peach Solution (Firmenich) 500 mg/L‡
Melon Solution (Firmenich) 400 mg/L‡
Citrus fruits
Bergamot Solution (Firmenich) 500 mg/L‡
Lemon Lemon natural aroma (Vahiné) Diluted 1/3‡
Orange Orange juice (Granini) Diluted 1/3‡
Grapefruit Solution (Firmenich) 300 mg/L‡
Red fruits
Cherry N° 15 (Le Nez du Vin)
Strawberry N° 24 (Le Nez du Vin)
Raspberry Solution (Firmenich) 200 mg/L‡
Redcurrant Redcurrant jelly (Bon Maman) 1 teaspoon†
Black fruits
Blackcurrant N° 14 (Le Nez du Vin)
Blackberry Blackberry yogurt (Yoplait) 1 teaspoon†
Plum N° 7 (Le Nez du Vin)
Dry fruits
Date Dry date 1 unit†
Fig Dry fig 1 unit†
Prune Prune 1 unit†
Exotic fruits
Banana Solution (Firmenich) 200 mg/L‡
Kiwi Kiwi juice (Caraïbos) Diluted 1/3‡
Pineapple Solution (Firmenich) 200 mg/L‡‡
Passion fruit Passion fruit juice (Caraïbos) Diluted 1/3‡
Lychee Solution (Firmenich) 300 mg/L‡
Mango Solution (Firmenich) 300 mg/L‡
Crystallised fruit Crystallised fruit (Vahiné) 2 units†
Muscat Rivesaltes muscat 5 mL†
Stone fruit Solution (Firmenich) 5000 mg/L‡
Floral
Acacia N° 2 (Le Nez du Vin)
Lime N° 50 (Le Nez du Vin)
Orange blossom Solution (Firmenich) 100 mg/L‡
Jasmine Solution (Firmenich) 100 mg/L‡
Lilac Solution (Firmenich) 200 mg/L‡
Violet Solution (Firmenich) 5000 mg/L‡
Chamomile Chamomile (Lipton) 1 tea sachet in 200 mL water
Rose Solution (Firmenich) 300 mg/L‡
Honeysuckle Solution (Firmenich) 100 mg/L‡
Geranium Geranium petals 4 units†
Honey Honey (Carrefour) 2 mL†

© 2008 Australian Society of Viticulture and Oenology Inc.


Campo et al. Varietal aroma properties of Spanish white wines 109

Table 2. Continued

Term Odour reference (source/brand) Quantity or concentration

Spicy
Fennel Fennel grains Five grains†
Liquorice Liquorice Stick (2 ¥ 1 cm)†
Clove Clove grains (Amora) 2 units†
Vanilla Vanilla extract (Vahiné) 1 mL†
Nutmeg Nutmeg (Amora) 1 teaspoon†
Black pepper Black pepper grains (Amora) 2 units†
Burnt
Smoky Standard (Sentosphère) §
Caramel Liquid caramel (Vahiné) 1 mL†
Toasted bread Standard (Sentosphère) §
Silex Flint 1 unit†
Vegetal
Vegetables
Artichoke Juice of artichoke in conserve 1 mL (diluted 1/10)†
(Carrefour)
Asparagus Juice of asparagus in conserve 1 mL (diluted 1/10)†
(Carrefour)
Cauliflower Juice of cauliflower in conserve 1 mL (diluted 1/10)†
(Carrefour)
Green beans Juice of green beans in conserve 1 mL (diluted 1/10)†
(Carrefour)
Olives Olive juice 10 mL (diluted 1/3)†
Bay leaf Bay leaf One chopped leaf†
Herbaceous Standard (Firmenich) 1 mL†
Animal
Musk Standard (Sentosphère) §
Cat urine Standard (Sentosphère) §
Wet dog Wet dog hair One lock†
Leather Standard (Sentosphère) §
Undergrowth
Mushroom Standard (Sentosphère) §
Humus Humus 10 g†
Earthy Earth 10 g†
Others
Lactic Solution (Firmenich) 1 mL†
Biscuit Standard (Sentosphère) §
Cider Cider (Carrefour) Diluted 1/3‡
Cooked meat Fumet (Viandox-Maggi) Two drops‡
Iode Solution (Firmenich) 1 mL†
Baking powder Solution (Firmenich) 1 mL†
Mouldy Solution (Firmenich) 1 mL†
Oak Oak essence 1 mL†
Ink Liquid Ink (Inoxcrom) 1 mL†

Descriptor, reference (chemical or natural) and employed quantities.


† Contained in a glass amber flask of 125 mL.
‡ Glass amber flask (60 mL) containing an absorbent paper support (5 ¥ 11 cm) impregnated with 10 mL of the odourant solution.
1 Solution (Firmenich): concentration is expressed in relation to the concentration of the commercial solution. The solvent employed
was mineral oil from Sigma (St. Louis, Pennsylvania, USA).
2 Natural product: diluted with purified water (Milli-Q, Millipore, Massachusetts, USA).
§ Standard Sentosphère directly placed in a glass amber flask of 125 mL.

© 2008 Australian Society of Viticulture and Oenology Inc.


110 Varietal aroma properties of Spanish white wines Australian Journal of Grape and Wine Research 14, 104–115, 2008

Correspondence analysis (CA) and HCA. Data were the sorting task showed that two different wines were
explored by CA. CA requires that data are organised into sorted in the same group a maximum of 17 times, which
a contingency table, the goal of CA being to transform corresponds to 52% of the panel (Verdejo Naia and
such a table into a graphical display. The resulting CA Verdejo Marqués de Riscal). On the contrary, several
map can then be interpreted using HCA. HCA makes it wines were never sorted together by any of the panellists,
possible to check the spatial arrangement of CA data and for example, Godello Gaba do Xil and Macabeo Coto, or
refine the interpretation of the map by identifying the Treixadura Villa Mein and Treixadura Gran Solbeira.
wines that belong to the same cluster (Lebart 1994). The resulting three-dimensional MDS chart (figure
A first contingency table, in which rows were the not shown) gave a stress value of 0.16, which indicated
wines (two replicates) and columns were the terms, was an acceptable representation of the original data. The
submitted to CA to explore the global reproducibility of examination of the MDS chart did not suggest a varietal
the panel. A second contingency table, containing the origin clustering as most of the wines were scattered over
average Cf of terms given by the most reproducible judges the map. These observations were corroborated by the
(according to the Ri value), was then constructed to study clusters yielded by the HCA (Figure 1). Four stable groups
the relationship between wines and terms. HCA with the emerge from the sorting task. Within the first cluster, all
Ward criteria was applied to the factorial coordinates of the Verdejo and Sauvignon Blanc wines appear clearly
the wines in the space defined by CA. The clusters iden- differentiated. No pattern seems to have emerged in the
tified by truncating the tree diagram were consolidated by second and third clusters as each comprised wines from
aggregation around mobile centres. The terms that best different varieties. The fourth cluster included the three
characterised each of the clusters were identified by using wines from Albariño, as well as a Chardonnay, a Treixa-
the test-value parameter (Morineau 1984). The test value dura and a Xarello. In Figure 1, each cluster has been
corresponds to a statistical criterion akin to a standardised labelled by a few generic descriptors, which roughly rep-
variable (zero mean and unit variance). Significance is resent the most usual terms given by the panellists. The
obtained when the absolute test value is ⱖ1.96, which first cluster was mainly described by ‘fruity’, and more
corresponds to an error threshold of 5%. By ranking the precisely, by ‘exotic fruits’. The second cluster was
terms according to their test values, it is possible to labelled by terms like ‘animal’ and ‘undergrowth’ as well
quickly characterise each of the clusters (Morineau as by terms with a hedonic connotation such as ‘unpleas-
1984). All analyses were performed with SPAD software ant’ or ‘bad odour’. Finally, the third and fourth clusters
(version 5.5, CISIA-CERESTA, Montreuil, France). were defined by non-specific terms such as ‘fruity’ or
‘floral’.
Results
Sorting task Descriptive test
The number of wine groups formed by the panellists
varied from 4 to 13 (median = 7). The maximum number Discriminant terms. The most discriminant terms,
of wines per group was nine, while groups constituted by according to the c2 criterion, are presented in Table 3.
a single wine were 21. The similarity matrix derived from Data in the table reveal that three different fruity families
Semi-partial R-squared

Cluster 1 Cluster 2 Cluster 3 Cluster 4


Figure 1. Composition of the
clusters derived from the
fruity, exotic fruits animal, undergrowth, fruity, floral fruity, floral sorting task (clusters were
“unpleasant”, “bad odour”
selected on the basis of the
first large change in levels).
saumar

maccot
macald

sausan

chaena
macsie
godest
godgui
vermar

xarann
chacar

parbar
vernar

tremei
charai

xarain
vernai

albest

alblus
godsil

albala

treest
tresol

© 2008 Australian Society of Viticulture and Oenology Inc.


Campo et al. Varietal aroma properties of Spanish white wines 111

Table 3. Most discriminant terms according to chi-square (white, citrus and exotic fruits) varied significantly among
(c2) analysis. the wines. On the contrary, no specific floral term, with
the single exception of honey, was used in a discriminant
Descriptor P Significance way.

White fruits 0.057 * Panel reproducibility. The overall panel reproducibility


Citrus fruits 0.042 ** was explored using the projection of wine replicates on
Exotic fruits 0.000 **** the two-dimensional CA map (Figure 2). As can be seen
Pineapple 0.001 *** in the figure, most of the wine replicates are close to each
Passion fruit 0.000 **** other on the map. Taking into account that a different
Mango 0.000 **** bottle was used for each repetition, it can be said that,
Honey 0.098 * globally, the panel was reproducible when describing
Smoky 0.009 **** wines. Concerning individual reproducibility, the
Vegetables 0.037 ** maximum value of the Ri index (which varies from 0 to 1)
was 0.51, which corresponds to 51% of common terms
Cauliflower 0.052 *
between the two replicates for a subject. The minimum
Undergrowth 0.000 ****
value was 0.17 (median = 0.32). The responses from the
Mushroom 0.000 ****
subjects showing an Ri < 0.2 were left out from the study
Animal 0.000 **** so that the final CA map was run over the responses of a
Leather 0.090 * total of 28 panellists.
Wet dog 0.015 **
Mouldy 0.025 ** CA and hierarchical clustering. The projection of wines
and terms (average of two repetitions) into a bidimen-
* P < 0.1. sional CA map (60% inertia) can be seen in Figure 3.
** P < 0.05.
The interpretation of the dimensions of the CA map
*** P < 0.01.
**** P < 0.001. was established by statistical indicators measuring the

Figure 2. Projection of wine replicates on the two-dimensional correspondence analysis map.


© 2008 Australian Society of Viticulture and Oenology Inc.
112 Varietal aroma properties of Spanish white wines Australian Journal of Grape and Wine Research 14, 104–115, 2008

Figure 3. Projection of descriptors and samples in the correspondence analysis map (dimensions 1 and 2). Generic terms are written in
capital letters, and specific terms are in plain text.

contributions of each term to the inertia on such dimen- showed statistically significant test values for each of the
sions. Only those attributes showing a contribution clusters, ranked in decreasing order. These terms are
higher than the average contribution were considered. those that better define the aromatic properties of wines
The first dimension, explaining 45% of the total variance, appearing in each of the four groups. As can be seen, most
reveals an aromatic opposition that could be summarised of the terms are specific to each cluster and only some
by ‘unpleasant’ (e.g. cauliflower, mushroom or wet dog) terms from the vegetal or animal families (vegetables,
versus ‘fruity/floral’ terms. The second dimension, green beans and leather) are shared by C1 and C2.
explaining 15% of the total variance, opposes the ‘exotic However, while C1 is richer in terms from the burnt
fruit’ attributes (mango, passion fruit) to the ‘sweet family (smoky, caramel, toasted bread), C2 is richer in
spices’ such as cinnamon or liquorices. terms from the undergrowth (mushroom), animal (wet
The tree diagram resulting from the HCA (figure not dog) and vegetal (cauliflower, green beans, asparagus and
shown) revealed that three partition options, from ten artichoke) families. C3 was mainly characterised by terms
possible, were the most appropriate from a statistical from the spicy and floral families. Finally, C4 was mainly
point of view, the number of clusters in each of these defined by attributes such as passion fruit, mango and
partitions being, respectively, three, two and four. Even if blackcurrant.
this option (three clusters) was the most natural partition For each cluster, the wine closest to the centre of
of the tree diagram, we chose the partition containing a gravity was identified as the most typical exemplar of this
total of four clusters as it permitted to obtain more precise cluster. These were Albariño Alargo (C1), Palomino Fino
descriptions of wines belonging to each of the groups. The (C2), Albariño Lusco (C3) and Sauvignon Marqués de
first and second clusters (C1 and C2) were constituted, Riscal (C4). The individual profiles of these wines are
respectively, by four (Chardonnay Raimat, Albariño presented in Table 5. In light of the terms hierarchy
Alargo, Treixadura Gran Solbeira and Treixadura obtained for those wines, the list of attributes employed
Estación) and three wines (Palomino Fino, Macabeo Coto seems to be very appropriate, as it has led to detailed
and Godello Guitian). Eleven wines were placed in the descriptions of wines. Furthermore, the list of terms used
third cluster (C3) including varietals from Xarello, in this study seems to be particularly efficient when
Godello, Macabeo, Chardonnay, Treixadura and describing wines with diverse fruity nuances. For
Albariño. Finally, the fourth cluster (C4) was formed by example, both Macabeo Albada and Verdejo Naia wines
the two wines from Sauvignon Blanc and the three wines showed high values for the fruity term (78 and 83%,
of Verdejo. respectively). However, while the former was richer in
To examine precisely which terms contribute most to white fruits (44%; 27% pear), the latter was extremely
the building of the clusters, we employed the test values rich in exotic fruits (61%; 42% passion fruit), which
associated with each cluster. Table 4 lists the terms that indicates two sensory profiles radically different.
© 2008 Australian Society of Viticulture and Oenology Inc.
Campo et al. Varietal aroma properties of Spanish white wines 113

Table 4. Terms showing statistically significant test value (ⱖ1.96) and contributing most to the building of each cluster.

Cluster 1 Cluster 2 Cluster 3 Cluster 4


Descriptor Test value Descriptor Test value Descriptor Test value Descriptor Test value

Smoky 3.99 Mushroom 6.09 Spicy 3.03 Passion fruit 8.24


Burnt 3.53 Undergrowth 5.45 Floral 2.99 Exotic fruits 7.20
Vegetal 3.20 Animal 4.49 White fruits 2.82 Mango 6.38
Animal 3.04 Mouldy 3.78 Liquorice 2.80 Citrus fruits 2.60
Vegetables 2.99 Cauliflower 3.73 Cinnamon 2.53 Blackcurrant 2.58
Caramel 2.92 Wet dog 3.71 Orange blossom 2.20 Yellow fruits 2.58
Bay leaf 2.35 Vegetables 3.67 Apricot 2.09 Fruity 2.07
Leather 2.32 Vegetal 3.39 Pear 2.06
Green beans 2.14 Humus 3.39
Oak 2.09 Green beans 3.07
Toasty bread 1.96 Leather 2.92
Asparagus 2.90
Earthy 2.68
Artichoke 2.06

Table 5. Sensory profiles (% of citation frequency) of wines that better represent the characteristics of each cluster.

Cluster 1 Cluster 2 Cluster 3 Cluster 4


Descriptor Albariño Descriptor Palomino Descriptor Albariño Descriptor Sauvignon
Alargo Barbadillo Lusco Marqués

Fruity 53 Vegetal 55 Fruity 69 Fruity 83


Vegetal 52 Animal 47 Floral 49 Exotic fruits 47
Burnt 45 Fruity 34 Spicy 30 Citrus fruit 33
Smoky 31 Undergrowth 33 Citrus 28 Floral 33
Animal 31 Vegetables 31 Vegetables 28 Passion fruit 28
Spicy 30 Burnt 28 White fruits 25 White fruits 27
Dried fruits 27 Mushroom 27 Burnt 20 Vegetal 27
Floral 27 Floral 22 Dried fruits 19 Black fruits 22
Vegetables 27 Caramel 19 Rose 19 Spicy 19
Caramel 17 Wet dog 19 Exotic fruits 17 Yellow fruits 17
Oak 16 Mouldy 19 Yellow fruits 16 Blackberry 17
Leather 16 Leather 17 Quince 14 Burnt 17
Prune 11 Dried fruits 16 Cider 14 Mango 16
Bay leaf 11 Cauliflower 16 Red fruits 13 Dried fruits 14
Undergrowth 11 Grass 16 Black fruits 13 Quince 13
Earth 16 Prune 13 Pineapple 13
Nuts 11 Peach 11 Pear 11

Discussion defining the aroma characteristics of Spanish varietals.


The first aim of the present study was to know if the This does not mean that the varietals lack specificity, but
subjects were able to group together, on the basis of their that at present, such specificity is blurred by other factors
aroma characteristics, wines made with the same variety. with a stronger influence on the aroma profile, such
The mapping that results from the sorting task makes it as those related to geographical origin and to different
possible to rapidly evaluate the odour similarities per- viticultural and enological practices.
ceived by the panel. In our study, no classification based In agreement with the sorting task, the descriptive
on varieties was observed excepting Verdejo, Sauvignon task shows that varietals from Verdejo and Sauvignon
Blanc and, to a lesser extent, Albariño. This clearly means Blanc are similar from an aromatic point of view. To our
that, at present, grape variety is not the major factor knowledge, there are no reports describing the similarities
© 2008 Australian Society of Viticulture and Oenology Inc.
114 Varietal aroma properties of Spanish white wines Australian Journal of Grape and Wine Research 14, 104–115, 2008

existing between these two varietals, a fact probably In parallel with the sensory characterisation of wines,
because of the scarce number of studies dealing with an additional scope of the present paper was to evaluate
Verdejo wine aroma characteristics (Gutierrez Afonso the suitability of the proposed strategy to study wine
et al. 1998, Lopez et al. 2003, Campo et al. 2005) A recent aroma. The final list of terms is deemed appropriate as it
study carried out by Campo et al. (2005) pointed out that allows fine descriptions of the products to be established.
Verdejo wines were rich in tropical fruit notes mainly Less cited terms were those of the ‘animal’ (musk and cat
because of the presence of high levels of the powerful urine), ‘red fruit’ (cherry, strawberry, raspberry and red-
aromatic thiol 3-mercaptohexylacetate, as seen in Sauvi- currant) and ‘spicy’ families (fennel, clove and nutmeg).
gnon Blanc wines (Tominaga et al. 1996). It is worth mentioning that up to six terms of the ‘floral’
The second aim of this work was to evaluate the category, a well-known descriptor for white varietals,
degree of diversity potentially introduced by the variety were below the 15% citation threshold, which indicates
of grape. Within the limited scope of this study, only that the number of floral terms of the list was in fact
wines from Verdejo and Sauvignon Blanc were found to excessive for young white wine description. The sorting
have a well delimited and specific sensory space, while task followed by MDS proved to be a successful qualita-
the remaining varietals share a common area of the tive instrument, allowing us to explore the holistic per-
sensory space. Interestingly, there is no single sample ceptions of the samples. Then, DA based on citation
showing distinctive and positive attributes. This suggests frequencies followed by CA and HCA (test values) made
that there is, in fact, a little diversity in aroma nuances it possible to successfully delimit the sensory spaces
among average quality young wines. This result is sur- where the different wines were pooled. Both sensory
prising in the case of Chardonnay, as recent studies have methodologies are based on substantially different prin-
shown the typicality associated with wines produced ciples but lead to similar conclusions, thus indicating a
from this grape (Ballester et al. 2005). These same good performance of the panellists to discriminate among
authors have, however, also indicated that not all the wines.
wines from Chardonnay are actually good examples of It is important to bear in mind that any method based
the so-called ‘Chardonnay wine concept’. on frequencies needs a larger number of people than
The third goal of the present work was to evaluate the classic DA. However, this kind of approach presents a
odour characteristics of this set of wines. The results clear advantage with respect to classic DA. As panellists
confirm that only the wines made with Verdejo and Sau- follow a general training, the trained panel is ready to
vignon Blanc have specific terms, such as exotic and evaluate quite different wine types with only two or three
citrus fruits. The majority of wines clustered in cluster 3 additional specific training sessions. This permits time
(Table 5) share several common attributes such as fruity, saving compared with classical DA, which require com-
floral and spicy. plete training each time different wine types are evalu-
An unexpected result of the present work was the ated. Concerning data analysis, it has been shown that,
presence of three wines with relatively unpleasant even if the subjects do not rate perceived intensities, it is
nuances (cluster 2) and of another four with some possible to establish a hierarchy of the terms that best
attributes belonging to the burnt and animal families define individual products (frequency citation percent-
(cluster 1), far from the general idea of young white age) or clustered wines (test values). A hierarchy of
wines without barrel ageing. In these cases, the four the most discriminant terms is also possible when using
bottles of each wine showed similar characteristics, chi-square values.
which indicates that such a problem cannot be associated
with the sampling of a single deficient bottle. An inter- Acknowledgements
esting observation is that such flavours were not associ- This work has been funded by the Spanish government,
ated to the most common wine faults such as cork taint projects AGL2004-06060 and AGL2007-65139/ALI. The
or oxidation. The presence of such aroma nuances is authors thank the people who kindly participated in this
probably masking the existence of other positive aroma study.
features of these samples, which clearly constitutes an
additional factor for blurring the role of variety. The
References
existence of these ‘problematic’ samples is particularly Aubry, V., Schlich, P., Issanchou, S. and Etievant, P. (1999) Com-
important, because the consumer will experience a parison of wine discrimination with orthonasal and retronasal
‘deception’ as the aromatic properties of the samples are profilings. Application to Burgundy Pinot Noir wines. Food
far from those expected. To illustrate this argument with Quality And Preference 10, 253–259.
a real example, the aroma profile of one of the studied Aznar, M., Lopez, R., Cacho, J. and Ferreira, V. (2003) Prediction
of aged red wine aroma properties from aroma chemical compo-
wines, Chardonnay Raimat, given by a Spanish sition. Partial least squares regression models. Journal of Agricul-
commonly visited wine guide web site (http:// tural and Food Chemistry 51, 2700–2707.
elmundovino.elmundo.es), was described as ‘floral, Ballester, J., Dacremont, C., Le Fur, Y. and Etievant, P. (2005) The
fruity and aromatic herbs’. This profile is the opposite of role of olfaction in the elaboration and use of the Chardonnay
the descriptions given by our panel (see Figure 3 and wine concept. Food Quality and Preference 16, 351–359.
Bijmot, T.H.A. and Wedel, M. (1993) An investigation into multi-
Table 4). Identifying the chemicals responsible for these dimensional scaling: Task effects on similarity data and solutions
unexpected terms and finding the causes leading to their of MAXSCAL. Doctoral Thesis, Institute of Economic Research,
formation should be a priority. University of Groningen, Groningen, the Netherlands.
© 2008 Australian Society of Viticulture and Oenology Inc.
Campo et al. Varietal aroma properties of Spanish white wines 115

Campo, E., Ferreira, V., Escudero, A. and Cacho, J. (2005) Pre- Canary Islands. Journal of Agricultural and Food Chemistry 51,
diction of the wine sensory properties related to grape variety 3419–3425.
from dynamic-headspace gas chromatography-olfactometry data. MacFie, H.J., Bratchell, N., Grennhoff, K. and Vallis, L.V. (1989)
Journal of Agricultural and Food Chemistry 53, 5682–5690. Designs to balance the effects of order of presentation and first-
Cartier, R., Rytz, A., Lecomte, A., Poblete, F., Krystlik, J., Belin, E. order carry-over effects in hall test. Journal of Sensory Studies 4,
and Martin, N. (2006) Sorting procedure as an alternative to 129–148.
quantitative descriptive analysis to obtain a product sensory map. McCloskey, L.P., Sylvan, M. and Arrhenius, S.P. (1996) Descriptive
Food Quality and Preference 17, 562–571. analysis for wine quality experts determining appellations by
Chollet, S. and Valentin, D. (2000) Le degré d’expertise a-t-il une Chardonnay wine aroma. Journal of Sensory Studies 11, 49–67.
influence sur la perception olfactive? Quelques éléments de Morineau, A. (1984) Note sur la caractérisation statistique d’une
réponse dans le domaine du vin. Année Psycologique 100, 11– classe par les valeurs-test. In: Bulletin technique du centre
36. international de statistique et d’informatique appliquées. (Saint-
Chollet, S. and Valentin, D. (2001) Impact of training on beer Mandé France) pp. 20–27.
flavor perception and description: Are trained and untrained sub- Noble, A.C. and Shannon, M. (1987) Profiling Zinfandel wines by
jects really different? Journal of Sensory Studies 16, 601–618. sensory and chemical analyses. American Journal of Enology and
De la Presa-Owens, C. and Noble, A.C. (1995) Descriptive analysis Viticulture 38, 1–5.
of three white wine varieties from Penedès. American Journal of Noble, A.C., Arnold, R.A., Buechsenstein, J., Leach, E.J., Schmidt,
Enology and Viticulture 46, 5–9. J.O. and Stern, P.M. (1987) Modification of a standardized
Falahee, M. and MacRae, A.W. (1997) Perceptual variation among system of wine aroma terminology. American Journal of Enology
drinking waters: The reliability of sorting and ranking data for and Viticulture 38, 143–146.
multidimensional scaling. Food Quality and Preference 8, 389– Pages, J. (2005) Collection and analysis of perceived product inter-
394. distances using multiple factor analysis: Application to the study
Falque, E., Ferreira, A.C., Hogg, T. and Guedes-Pinho, P. (2004) of 10 white wines from the Loire Valley. Food Quality and Pref-
Determination of aromatic descriptors of Touriga Nacional wines erence 16, 642–649.
by sensory descriptive analysis. Flavour and Fragrance Journal 19, Parr, W.V., Green, J.A., Geoffrey White, K. and Sherlock, R.R.
298–302. (2007) The distinctive flavour of New Zealand Sauvignon Blanc:
Guinard, J.X. and Cliff, M. (1987) Descriptive analysis of Pinot noir Sensory characterisation by wine professionals. Food Quality and
wines from Carneros, Napa and Sonoma. American Journal of Preference 18, 849–861.
Enology and Viticulture 38, 211–215. Peynaud, E. and Blouin, J. (1999) Découvrir le goût du vin
Gutierrez Afonso, V.L., Darias, J., Armas, R., Medina, M.R. and (Dunod: Paris).
Diaz, E. (1998) Descriptive analysis of three white wine varieties Piombino, P., Nicklaus, S., le Fur, Y., Moio, L. and Le Quéré, J.L.
cultivated in the Canary Islands. American Journal of Enology (2004) Selection of products presenting given flavor characteris-
and Viticulture 49, 440–443. tics: An application to wine. American Journal of Enology and
Heymann, H. and Noble, A.C. (1987) Descriptive analysis of com- Viticulture 55, 27–34.
mercial Cabernet-Sauvignon wines from California. American Saint-Eve, A., Kora, E.P. and Martin, N. (2004) Impact of the
Journal of Enology and Viticulture 38, 41–44. olfactory quality and chemical complexity of the flavouring agent
Koussissi, E., Paterson, A. and Piggott, J.R. (2007) Sensory profil- on the texture of low fat stirred yogurts assessed by three different
ing of aroma in Greek dry red wines using rank-rating and sensory methodologies. Food Quality and Preference 15, 655–
modanic scoring related to headspace composition. European 668.
Food Research and Technology 225, 749–756. Schiffman, S. (1976) Multidimensional scaling, a useful tool to
Lawless, H.T. (1999) Descriptive analysis of complex odours: measure flavor. Cereal Foods World 21, 64–68.
Reality, model or illusion? Food Quality and Preference 10, 325– Tominaga, T., Darriet, P. and Dubourdieu, D. (1996) Identification
332. of 3-mercaptohexyl acetate in Sauvignon wine, a powerful aro-
Lawless, H.T., Sheng, N. and Knoops, S. (1995) Multidimensional matic compound exhibiting box-tree odour. Vitis 35, 207–210.
scaling of sorting data applied to cheese perception. Food Quality Valentin, D., Chollet, S. and Abdi, H. (2003) Les mots du vin:
and Preference 6, 91–98. experts et novices diffèrent-ils quand ils décrivent des vins?
Le Fur, Y., Mercurio, V., Moio, L., Blanquet, J. and Meunier, J.M. Corpus 2, 183–200.
(2003) A new approach to examine the relationships between Vilanova, M. (2006) Sensory descriptive analysis and consumer
sensory and gas chromatography-olfactometry data using gener- acceptability of Godello wines from Valdeorras Apellation Origen
alized procruster analysis applied to six French Chardonnay Controlee (Northwest Spain). Journal of Sensory Studies 21,
wines. Journal of Agricultural and Food Chemistry 51, 443–452. 362–372.
Lebart, L. (1994) Complementary use of correspondence analysis Vilanova, M. and Vilarino, F. (2006) Influence of geographic origin
and cluster analysis. In: Complementary use of correspondence on aromatic descriptors of Spanish Albarino wine. Flavour and
analysis in the social sciences: Recent developments and applica- Fragrance Journal 21, 373–378.
tions. Eds. M. Greenacre and J. Blasius (Academic Press: London,
San Diego) pp. 162–178.
Lillo, M.P.Y., Latrille, E., Casaubon, G., Agosin, E., Bordeu, E. and
Martin, N. (2005) Comparison between odour and aroma profiles Manuscript received: 28 September 2007
of Chilean Pisco spirit. Food Quality and Preference 16, 59–70.
Lopez, R., Ortin, N., Perez-Trujillo, J.P., Cacho, J. and Ferreira, V. Revised manuscript received: 25 March 2008
(2003) Impact odourants of different young white wines from the Accepted: 30 March 2008

© 2008 Australian Society of Viticulture and Oenology Inc.

You might also like