Food Chemistry 114 (2009) 15–19

Contents lists available at ScienceDirect

Food Chemistry

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

Sensory and analytical re-evaluation of ‘‘Brett character”

Andrea Romano a,*, Marie Claire Perello a, Aline Lonvaud-Funel a, Gilles Sicard b, Gilles de Revel a

a UMR 1219 �nologie, Université Victor Segalen Bordeaux 2, INRA, ISVV, 351 Cours de la Libération, 33405 Talence, Franceb UMR5170, Centre Européen des Sciences du Goût, 21000 Dijon, France

a r t i c l e i n f o

Article history:Received 2 June 2008Received in revised form 28 June 2008Accepted 1 September 2008

Keywords:EthylphenolsIsobutyric acidIsovaleric acidDetection thresholdWine

0308-8146/$ - see front matter � 2008 Elsevier Ltd. Adoi:10.1016/j.foodchem.2008.09.006

* Corresponding author. Tel.: +33 (0)5 40 00 64 69E-mail address: andrea.romano@u-bordeaux2.fr (A

a b s t r a c t

Worldwide wine production has been significantly affected by Brettanomyces bruxellensis spoilage. Thisalteration, sometimes referred to as ‘‘Brett character”, results in the production of several volatile com-pounds and a large spectrum of flavours and aromas. Ethylphenols (namely 4-ethylphenol and 4-ethyl-guaiacol) are the best-known markers of this defect with a commonly used aggregate detection thresholdof about 400 lg/l. Fifty-one Bordeaux red wines were tasted with the aim of wine profiling for commer-cial purposes. Ethylphenol concentrations of wines were very poorly correlated to the corresponding tast-ing notes. Sensory analysis was employed to demonstrate the complexity of ‘‘Brett character”. A maskingeffect of isobutyric acid and isovaleric acid on the detection of ethylphenols in wine was proven. Thispartly explained the poor correspondence between ethylphenol concentrations and presence of ‘‘Bretty”descriptors.

� 2008 Elsevier Ltd. All rights reserved.

1. Introduction

Among all wine spoilage microorganisms the yeast Brettanomy-ces bruxellensis is one of the most feared by winemakers (Loureiro& Malfeito-Ferreira, 2006; Renouf, Lonvaud-Funel, & Coulon, 2007).The development of B. bruxellensis may cause ‘‘Brett character”,which occurs mainly in red wines. Brett character produces a widespectrum of flavours and aromas that include barnyard-like, min-eral, ink, tobacco, leathery, pharmaceutical and smoky descriptors(Boulton, Singleton, Bisson, & Kunkee, 1996; Etievant, 1991). Brettcharacter also entails the suppression of desirable fruity and flow-ery notes of wine (Gerbaux & Vincent, 2001).

B. bruxellensis spoilage activity is linked to the synthesis of vi-nyl- and ethylphenols (Heresztyn, 1986). Ethylphenols, andnamely 4-ethylphenol and 4-ethylguaiacol, are the best-knownmarkers of the alteration. To date B. bruxellensis is the only micro-bial species whose presence and development in wine has beenunmistakably related to the synthesis of ethylphenols. Some lacticacid bacteria (Couto, Campos, Figueiredo, & Hogg, 2006) and theyeast Pichia guillermondii (Barata, Nobre, Correia, Malfeito-Ferreira,& Loureiro, 2006) are able to synthesise ethylphenols in culturemedia but the oenological relevance of these activities was neverproven so far.

The aggregate detection threshold for ethylphenols wasreported at 426 lg/l (Chatonnet, Dubourdieu, Boidron, & Pons1992). This value was determined with data from a jury of 20

ll rights reserved.

; fax +33 (0)5 40 00 64 68.. Romano).

tasters employing a methodology developed by Boidron, Chaton-net, and Pons (1988). The authors used a Bordeaux red wine thatwas supplemented with graduated amounts of a 10:1 mixture of4-ethylphenol and 4-ethylguaiacol. The 10:1 concentration ratiocorresponds to the average found in Bordeaux wines. Researchershave subsequently used this threshold value as the ‘‘borderline”between spoiled and unspoiled wine. This detection threshold alsocarries significant economic importance: winemakers have made itthe basis for treating wines thought to be at risk of B. bruxellensisspoilage.

In 2007, the practical significance of this threshold value wasevaluated using a panel of 6 expert judges and 17 Bordeaux redwines of the 2003 vintage. The wines contained from 53 to1417 lg/l ethylphenols. Quite interestingly some of the wines thatpresented above-threshold amounts of ethylphenols (up to 668 lg/l) were not described as ‘‘Bretty” by any of the tasters (de Revel,personal communication). These results are in accordance withanecdotal reports (Goode, 2005) about scarce correlations betweenBrett character and ethylphenol contents of wines.

Acetyl-tetrahydropyridine, carboxylic acids and some of theirethyl-esters can also participate to Brett character in wines (Roma-no, Perello, de Revel, & Lonvaud-Funel, 2008) but their sensoryimplications are not fully known to date. This work aimed at ananalytical and sensory re-evaluation of Brett character on the basisof more refined criteria that take into consideration its complexity.Wines were analysed by gas chromatography–mass spectrometry(GC–MS) and sensorial experiments were carried out to identifyproducts of B. bruxellensis metabolism that interact with ethyl-phenols to mask their aromas.

16 A. Romano et al. / Food Chemistry 114 (2009) 15–19

2. Materials and methods

2.1. Reagents and standards

4-vinylphenol, 4-vinylguaiacol, 4-ethylphenol and 4-ethyl-guaiacol, were supplied by Lancaster (Ward Hill, MA). Deuterated4-ethylphenol was purchased from Cluzeau (Paris, France). Ethanolwas from Merck (Darmstadt, Germany). All other standards (car-boxylic acids, ethyl-esters and oak wood volatile compounds)and solvents (dichloromethane, diethyl-ether and isohexane) wereobtained from Sigma Aldrich (Saint Quentin Fallavier, France). Allstandards were at least 97% pure, all solvents were HPLC-grade(at least 99.7% pure).

2.2. Wine samples

The wines inoculated with B. bruxellensis were 2006 red Bor-deaux (pH 3.4–3.5, ethanol 12.4–13.0% (v/v), total reducing sugars0.8–1.2 g/l, volatile acidity 0.12–0.35 g/l expressed as acetic acid,total phenolics 39–57 expressed as absorbance at 280 nm, totalvolatile phenols 8–14 lg/l). A 2006 Bordeaux red wine was usedfor detection threshold calculation (pH 3.5, ethanol 12.2% (v/v), to-tal reducing sugars 2.0 g/l, volatile acidity 0.55 g/l expressed asacetic acid, total phenolics 49 expressed as absorbance at280 nm, total volatile phenols 13 lg/l). All the other red wines(pH 3.4, ethanol 12.5–13.7% (v/v), total reducing sugars 1.2–2.4 g/l, volatile acidity 0.51–0.84 g/l, 58–98 total phenolics) were com-mercial samples that originated from wineries throughout the Bor-deaux region and belonged to the 2005 vintage.

4-vinylphenol, 4-vinylguaiacol, 4-ethylphenol and 4-ethyl-guaiacol, isobutyric acid and isovaleric acid pure standards weredissolved at a concentration of 10% w/v in ethanol. When neededthese concentrated ethanolic solutions were employed to performadditions to wine.

2.3. Protocol of the experiments with B. bruxellensis

Strain B. bruxellensis IOEB L0506 (culture collection of the Fa-culté d’�nologie, Bordeaux, France) was inoculated in three sterilered wines. Before being inoculated wines were centrifuged twice(15800g for 30 min) and sterilized by filtration on a 0.45 lm poresize membrane (Sartorius, Goettingen, Germany). After a pre-adap-tation step the microbial strain was inoculated in sterile wine at aconcentration of 5 � 103 viable cells/ml (measured by direct epi-fluorescence method, Divol & Lonvaud-Funel, 2005). Inoculatedwines were kept at 25 �C in 2.5 l screw capped bottles with noheadspace volume. For each wine a parallel run without inocula-tion was carried out. After one month all wines were filtered againand submitted to sensory analysis. Prior to sensory analysis non-inoculated blanks were added with the required amounts of vola-tile phenol pure standards (see further).

2.4. Sensory analysis

Fifty-one red Bordeaux wines of the 2005 vintage were tastedwith the aim of wine profiling for commercial purposes. Four pro-fessionals of the wine sector evaluated samples both orthonasallyand retronasally and a list of freely perceived descriptors was ob-tained for each wine. Samples were presented to the tastersemploying tulip-shaped tasting glasses (ISO 3591:1977).

Triangle tests (ISO 4120:2004) were employed to compare inoc-ulated and non-inoculated wines (three sets of samples in total).The panel consisted of 21 judges belonging to the faculty staff.Samples were presented within tulip-shaped glasses marked byrandom three-digit codes and were evaluated both orthonasally

and retronasally. The inoculated wine samples were taken asexamples of ‘‘natural” Brett character. Non-inoculated blanks wereadded with volatile phenol (4-vinylphenol, 4-vinylguaiacol, 4-eth-ylphenol and 4-ethylguaiacol) pure standards so that their finalconcentrations matched those of spoiled wines. The samples wherethe sensory defect was reproduced by mere addition of its markers(i.e. volatile phenols) were taken as example of ‘‘artificial” Brettcharacter.

Detection thresholds were calculated following ISO guidelines(ISO 13301:2002). Six sets of three-alternative forced-choice (3-AFC) tests were performed. Each series contained one positivesample supplemented with ascending (17 – 34 – 68 – 137 – 275– 550 lg/l) concentrations of ethylphenols in a 4-ethylphenol: 4-ethylguaiacol 10:1 concentration ratio. The panel was a subset ofthe previous one and consisted of 10 expert judges belonging tothe laboratory staff. The judges were already trained in the percep-tion of the sensory notes of 4-ethylphenol and 4-ethylguaiacol inwine but they were not informed about the purpose of the test.Samples were evaluated orthonasally and consisted in 4 ml winealiquots presented within 20 ml screw capped bottles with 2 cmneck diameter. The wine samples were introduced into the bottlesand these were capped at least 1 h before the experiment in orderto allow for equilibration. All tests were performed in a roomequipped with individual tasting booths (ISO 8589:2007).

2.5. Chemical analysis

Ethanol, volatile acidity and total reducing sugars were deter-mined according to the international methods for wine and mustanalysis published by the Organisation Internationale de la Vigneet du Vin (downloadable at http://news.reseau-concept.net/images/oiv/Client/RECUEIL_2007_Vol1.pdf).

Volatile phenols were determined by GC–MS analysis coupledto solid-phase micro-extraction (SPME) on polyacrylate fibers.Samples (10 ml of wine) were placed in 25 ml vials containing NaCl(3.5 g). Deuterated 4-ethylphenol was added as internal standardat a concentration of 100 lg/l. Samples were submitted to GC–MS analysis (Romano et al., 2008). The electron impact (EI) massdetector operated in the selected ion monitoring (SIM) mode andanalytes were measured by comparing peak areas of specific ions(4-vinylphenol m/z 120, 4-vinylguaiacol m/z 150, 4-ethylphenolm/z 107, 4-ethylguaiacol m/z 137) with that of [2H10] 4-ethylphe-nol (m/z 113) used as internal standard.

Carboxylic acids and ethyl-esters were quantified by GC–FIDanalysis of wine liquid extracts. Samples (10 ml of wine) wereadded with 1 mg/l internal standard (3-octanol) and extractedtwice with 2 ml of diethyl-ether:isohexane 1:1 (v/v). The organicfractions were collected and submitted to GC analysis (Bertrand,1981).

Oak wood volatile compounds (guaiacol, 4-methyl-guaiacol, o-cresol, m-cresol, p-cresol, phenol, 4-propyl-guaiacol, eugenol, iso-eugenol, syringol and 4-allyl-syringol) were measured by GC–MSanalysis of wine liquid extracts. Samples (10 ml of wine) wereadded with 1 mg/l internal standard (deuterated 4-ethylphenol)and extracted twice with 2 ml of dichloromethane. The organicfractions were collected and submitted to GC–MS analysis. Themass detector operated in the selected ion monitoring (SIM) modeand analytes were measured by comparing peak areas of specificions with that of the internal standard. (Bloem, Lonvaud-Funel, &de Revel, 2008).

2.6. Statistical analysis

All chemical analyses were performed at least in duplicate.Experimental data were compared by means of a student t testat a 95% confidence level unless otherwise noted. Data treatment

Table 3Volatile phenol, carboxylic acid and ethyl-ester contents of wines submitted totriangle test

Wine 1 Wine 2 Wine 3

N A N A N AP

Carboxylic acids C3–C10 (mg/l) 9.13 7.42 8.19 6.68 7.16 5.71P

Ethyl-esters C6–C10 (mg/l) 2.59 1.16 2.42 1.08 1.74 0.834-Vinylphenol (lg/l) 24 26 384-Vinylguaiacol (lg/l) 7 11 194-Ethylphenol (lg/l) 12 425 4914-Ethylguaiacol (lg/l) 4 35 59Correct answers 17/21a 17/21� 18/21�

Volatile phenol concentrations are referred to the amounts obtained after 30 days ofincubation (N) or after pure standard addition (A). Correct answers: number oftasters (on 21) who identified the sample that was different in the triangle test.

a Different at a 99.9% significance level. A = artificial, N = natural ‘‘Bretty” wine.

A. Romano et al. / Food Chemistry 114 (2009) 15–19 17

was performed employing Minitab statistical software (MinitabInc., State College, PA).

3. Results and discussion

3.1. ‘‘Bretty” notes and ethylphenol concentrations in commercialwines

Fifty-one Bordeaux red wines were profiled for commercial pur-poses. On the basis of the tasting notes samples were subdividedinto three classes. ‘‘Heavily tainted” wines presented ‘‘Bretty”descriptors for all judges, in ‘‘mildly tainted” wines such descrip-tors were found by at least one judge and ‘‘non-tainted” wineswere virtually faultless. Tainted wines with relative ‘‘Bretty”descriptors are reported in Table 1.

Wines were submitted to GC–MS analysis for volatile phenols.4-ethylphenol and 4-ethylguaiacol were present in all sampleswith a 10:1 average concentration ratio, typical of Bordeaux wines.Vinylphenols were instead not detectable.

The correlation between sensory data and ethylphenol contentsappeared to be very poor (Table 2). No significant differences werefound in the average ethylphenol content of each class. Many heav-ily tainted and mildly tainted wines contained below-thresholdamounts of ethylphenols and, more importantly, more than onethird of non-tainted wines contained more than 400 lg/l ofethylphenols.

3.2. Comparison between artificial and natural ‘‘Bretty” wines

In previous work (Romano et al., 2008) it was demonstratedthat in a dry wine the production of volatile phenols by B. bruxell-ensis is accompanied by the synthesis of carboxylic acids (from C3

to C10) and ethyl-esters (from C6 to C10). It could be inferred that

Table 1Tainted commercial wines and relative ‘‘Bretty” descriptors

Wine code Categorized as ‘‘Bretty” descriptors

05 Heavily tainted Leather, tobacco08 Leather13 Leather23 Leather, tobacco43 Ink, leather09 Mildly tainted Leather10 Leather33 Leather36 Leather38 Leather41 Ink, leather45 Mineral, leather48 Smoky54 Leather56 Leather

‘‘Heavily tainted” wines presented ‘‘Bretty” descriptors according to all judges, in‘‘mildly tainted” wines such descriptors were found by at least one judge.

Table 2Sensory categorisation and ethylphenol contents of 51 Bordeaux wines

Categorized as Number of winesP

Ethylphenols (lg/l)

Min Max Average

Heavily tainted 5 196 746 375Mildly tainted 10 8 563 272Non-tainted 36 5 1370 403

‘‘Heavily tainted” wines presented ‘‘Bretty” descriptors according to all judges, in‘‘mildly tainted” wines such descriptors were found by at least one judge and ‘‘non-tainted” wines were virtually faultless.

this activity has implications on the perception of Brett character.In order to verify this assumption, spoiled samples were preparedby inoculation of three red wines with a B. bruxellensis strain. Thewines presented no sensory defects and had similar properties interms of the most important parameters of oenological interest(Section 2).

The production of variable amounts of volatile phenols was ob-served after one month of incubation (Table 3). B. bruxellensis al-ways proliferated attaining populations of 1.2–1.7 106 viablecells/ml. The comparison of the volatile compound profile of inoc-ulated samples and non-inoculated blanks confirmed the synthesisof carboxylic acids (up to 0.8 mg/l isobutyric acid and 0.7 mg/l iso-valeric acid) and ethyl-esters (up to 0.4 mg/l ethyl-octanoate and0.8 mg/l ethyl-decanoate).

Natural ‘‘Bretty” wines (i.e. inoculated samples) and artificial‘‘Bretty” wines (i.e. non-inoculated wines supplemented with vol-atile phenol pure standards) were compared for the three sets ofwines by means of a triangle test. The significance of the resultwas evaluated on the basis of the binomial law and significant dif-ferences (Table 3) were obtained for each set of samples. This dem-onstrated that volatile compounds other than volatile phenols havea detectable sensory impact on wines spoiled by B. bruxellensis.

3.3. Wine analysis for carboxylic acids, ethyl-esters and volatile oakcompounds

The complexity of ‘‘Brett character” was evaluated under wine-making conditions. Twenty-six commercial wines (5 heavilytainted, 6 mildly tainted and 15 non-tainted) were selected amongthe previously mentioned 51 and submitted to GC–FID analysis forcarboxylic acids and ethyl-esters. A statistically significant correla-tion (1 � a = 0.95) could be evidenced between ethylphenols andisobutyric (iC4) and isovaleric (iC5) acids. All other carboxylic acids(from C3 to C10) and ethyl-esters (from C6 to C10) showed insteadvery poor correlations (Table 4). B. bruxellensis is presumably the

Table 4Carboxylic acid and ethyl-ester contents of 26 Bordeaux wines

Concentration (mg/l)

C3 + C4 iC4 iC5 C6 C8 C10 C2C6 C2C8 C2C10

Min <0.1 1.34 1.87 0.63 0.51 0.21 0.07 0.10 0.18Max 4.67 6.73 1.92 2.22 1.36 0.48 0.50 0.67Average 2.62 3.49 1.02 0.82 0.45 0.31 0.27 0.42Correlation

(r) withethylphenols

n.a. 0.66a 0.78a 0.05 0.14 0.07 0.19 0.13 0.29

Correlation coefficients (r) with ethylphenols are also reported.a Significant at a 95% confidence level.

Fig. 1. Probabilities of success and orthonasal detection thresholds of ethylphenolsin unmodified wine and wine added with 1 mg/l isobutyric and isovaleric acid. Errorbars represent 95% confidence limits. Threshold values and respective low and highconfidence limits are reported below the graph.

18 A. Romano et al. / Food Chemistry 114 (2009) 15–19

only known species that produces ethylphenols under winemakingconditions. Our results indicate therefore that isobutyric acid andisovaleric acid are to be considered as important markers of Brettcharacter. The contribution of B. bruxellensis to the synthesis ofethyl-esters and other carboxylic acids does not seem to berelevant.

The sensory profiles of the wines did not contain any mention tothe ‘‘rancid” and ‘‘sweaty” descriptors typical of these carboxylicacids, it remained to explain whether these compounds had an im-pact on the perception of ‘‘Bretty” sensory notes.

The poor correlation between ethylphenol contents and Brettcharacter perception could also be due to the presence of othersimple phenols (typically phenol, cresols, guaiacols and syringols)derived from barrel aging and whose descriptors (smoky, hospital,leather, smoky, spicy, bitumen) are similar to those of ethylphenols(Escudero, Campo, Fariña, Cacho, & Ferreira, 2007). Ten selectedwines (5 non-tainted and 5 heavily tainted) were submitted toGC–MS analysis for these oak wood volatile compounds. The winescontained up to 20 lg/l guaiacol, 20 lg/l phenol, 60 lg/l eugenol,60 lg/l syringol and 50 lg/l 4-allyl-syringol but no significant dif-ference was found in the profiles of tainted and non-tainted wines.Moreover none of the quantified compounds was present inamounts higher than the corresponding detection thresholds (Boi-dron et al., 1988; Lopez, Aznar, Cacho, & Ferreira, 2002; van Gemert& Nettenbreijer, 1977).

3.4. Calculation of ethylphenol detection thresholds

In the attempt to verify whether isobutyric acid and isovalericacid had an effect on the perception of ‘‘Bretty” notes we per-formed two series of three-alternate forced-choice (3-AFC) tests.During the first tasting session an aggregate detection thresholdwas measured in wine for the mixture 4-ethylphenol:4-ethyl-guaiacol in a 10:1 concentration ratio. In the second experimentthe detection threshold was calculated in a wine supplementedwith 1 mg/l isobutyric acid and 1 mg/l isovaleric acid. In a preli-minary trial we verified that ‘‘rancid” and ‘‘sweaty” olfactory notestypical of these compounds were not detectable at theseconcentrations.

In the present experiments, wine samples were presentedemploying screw capped bottles instead of more commonly usedtulip-shaped tasting glasses. This was to allow for a rapid equili-bration between wine and its headspace. It was verified thatheadspaces were equilibrated after 1 h (data not shown). Theneed to standardise headspace composition for detection thresh-old calculation has already been emphasised by Atanasova et al.(2005).

The detection thresholds (Fig. 1) correspond to a 0.5 probabilityof detection after correction for chance guessing. The methodologyemployed for threshold calculation (ISO 13301:2002) uses a ‘‘logit”transformation of the corrected probabilities followed by a linearregression.

The comparison between the two detection thresholds of theethylphenols allowed us to ascertain that isobutyric acid and iso-valeric acid do possess a masking effect with respect to these eth-ylphenols: in fact the detection threshold value was three timeshigher when wine was supplemented with the carboxylic acids(Fig. 1).

The calculated detection threshold for ethylphenols in wine wasabout four times lower than that reported in literature (Chatonnetet al., 1992), but it must be remembered that this is easily influ-enced by wine complexity. The wine employed for detectionthreshold calculation, unlike the ones submitted to descriptiveanalysis, was not barrel aged and its phenolic complexity was sig-nificantly below average, as indicated by the values of absorbanceat 280 nm (Section 2).

4. Conclusions

In this work the role of ethylphenols as sole markers of Brettcharacter was re-examined. We demonstrated that some second-ary metabolites other than volatile phenols have a sensory impacton the spoiled wine. Moreover, the addition of isobutyric acid andisovaleric acid to wine resulted in a remarkable reduction of thedetection threshold of ethylphenols. Indeed, Atanasova et al.(2004) demonstrated that the interaction between wine aromaticcompounds belonging to different odour families (e.g. woody/fru-ity) can result in the appearance of masking effects. This partly ex-plains the poor correlation between sensory data and ethlylphenolcontents. It must be noted that ‘‘rancid” and ‘‘sweaty” notes typicalisobutyric acid and isovaleric acid are quite close to some animal‘‘Bretty” descriptors and the importance of semantic factors inwine tasting has already been discussed by various authors (Bro-chet & Dubourdieu, 2001; Parr, White, & Heatherbell, 2004). Fur-ther work is required in order to ascertain whether theincoherencies regarding the sensory evaluation of ‘‘Bretty” winescould also be related to semantic confusion.

Acknowledgments

During his research work Dr Andrea Romano benefited from anISVV travel grant, issued by the Institut des Sciences de la Vigne et duVin. The authors wish to thank Mr Guy Guimberteau for his assis-tance in the profiling of commercial wines.

References

Atanasova, B., Thomas-Danguin, T., Langlois, D., Nicklaus, S., Chabanet, C., &Etiévant, P. (2004). Perceptual interactions between fruity and woody notes ofwine. Flavour and Fragrance Journal, 19, 476–482.

Atanasova, B., Thomas-Danguin, T., Langlois, D., Nicklaus, S., Chabanet, C., &Etiévant, P. (2005). Perception of wine fruity and woody notes: influence ofperi-threshold odorants. Food Quality and Preference, 16, 504–510.

A. Romano et al. / Food Chemistry 114 (2009) 15–19 19

Barata, A., Nobre, A., Correia, P., Malfeito-Ferreira, M., & Loureiro, V. (2006). Growthand 4-ethylphenol production by the yeast Pichia guillermondii in grape juices.American Journal of Enology and Viticulture, 57, 133–138.

Bertrand, A. (1981). Formation des substances volatiles au cours de la fermentationalcoolique. Incidence sur la qualité du vin. In Colloque Société Française deMicrobiologie (pp. 251-267). Reims, France.

Bloem, A., Lonvaud-Funel, A., & de Revel, G. (2008). Hydrolysis of glycosidicallybound flavour compounds from oak wood by Oenococcus oeni. FoodMicrobiology, 25, 99–104.

Boidron, J. N., Chatonnet, P., & Pons, M. (1988). Influence du bois sur certainessubstances odorantes des vins. Connaissance de la Vigne et du Vin, 22, 275–294.

Boulton, R., Singleton, V., Bisson, L., & Kunkee, R. (1996). Microbiological spoilage ofwine and its control. In Principles and practice of winemaking (pp. 364–381). NewYork: Chapman & Hall.

Brochet, F., & Dubourdieu, D. (2001). Wine descriptive language supports congnitivespecificity of the chemical senses. Brain and Language, 77, 187–196.

Chatonnet, P., Dubourdieu, D., Boidron, J. N., & Pons, M. (1992). The originof ethylphenols in wine. Journal of the Science of Food and Agriculture, 60,165–178.

Couto, J. A., Campos, F. M., Figueiredo, A. R., & Hogg, T. A. (2006). Ability of lactic acidbacteria to produce volatile phenols. American Journal of Enology and Viticulture,57, 166–171.

Divol, B., & Lonvaud-Funel, A. (2005). Evidence for viable but nonculturable yeastsin botrytis-affected wine. Journal of Applied Microbiology, 99, 85–93.

Escudero, A., Campo, E., Fariña, L., Cacho, J., & Ferreira, V. (2007). Analyticalcharacterization of the aroma of five premium red wines. Insights into the roleof odor families and the concept of fruitiness of wines. Journal of Agricultural andFood Chemistry, 55, 4501–4510.

Etievant, P. (1991). Wine. In H. Maarse (Ed.), Volatile Compounds in Foods andBeverages (pp. 483–546). New York: Marcel Dekker.

Gerbaux, V., & Vincent, B. (2001). Effects of volatile phenols on sensory quality ofred wines from different cultivars. Revue des �nologues, 99, 15–18.

Goode, J. (2005). Brettanomyces. In M. Beazley (Ed.), Wine Science (pp. 136–143).London: Octopus Publishing.

Heresztyn, T. (1986). Metabolism of phenolic compounds from hydroxycinnamicacids by Brettanomyces yeasts. Archives of Microbiology, 146, 96–98.

ISO 13301:2002. Sensory analysis – Methodology – General guidance for measuringodour, flavour and taste detection thresholds by a three-alternative forced-choice (3-AFC) procedure.

ISO 3591:1977. Sensory analysis – Apparatus – Wine tasting glass.ISO 4120:2004. Sensory analysis – Methodology – Triangle test.ISO 8589:2007. Sensory analysis – General guidance for the design of test rooms.Lopez, R., Aznar, M., Cacho, J., & Ferreira, V. (2002). Quantitative determination of

minor and trace volatile compounds in wine by solid-phase extraction and gaschromatography with mass spectrometric detection. Journal of ChromatographyA, 966, 167–177.

Loureiro, V., & Malfeito-Ferreira, M. (2006). Dekkera/Brettanomyces spp.. In C. deBlackburn (Ed.), Food spoilage microorganisms (pp. 354–398). Cambridge:Woodhead Publishing.

Parr, W. V., White, K. G., & Heatherbell, D. A. (2004). Exploring the nature of wineexpertise: What underlies wine experts’ olfactory recognition memoryadvantage? Food Quality and Preference, 15, 411–420.

Renouf, V., Lonvaud-Funel, A., & Coulon, J. (2007). The origin of Brettanomycesbruxellensis in wines: A review. Journal International des Sciences de la Vigne et duVin, 41, 161–173.

Romano, A., Perello, M. C., de Revel, G., & Lonvaud-Funel, A. (2008). Growth andvolatile compound production by Brettanomyces bruxellensis in red wine. Journalof Applied Microbiology, 104, 1577–1585.

van Gemert, L. J., & Nettenbreijer, A. H. (1977). Compilation of odour threshold valuesin air and water. Zeist: TNO.