PERCEPTION OF FINISH IN WHITE WINE By EMILY GOODSTEIN · mushroom acting as a suppressor of other flavors. Normalization of model wine time intensity analysis showed that data variation

PERCEPTION OF FINISH IN

WHITE WINE

By

EMILY GOODSTEIN

A thesis submitted in partial fulfillment of

the requirements for the degree of

MASTER OF SCIENCE IN FOOD SCIENCE

WASHINGTON STATE UNIVERSITY

Department of Food Science

DECEMBER 2011

ii

To the Faculty of Washington State University:

The members of the Committee appointed to examine the thesis of EMILY SIMPSON

GOODSTEIN find it satisfactory and recommend that it be accepted.

___________________________________

Carolyn Ross, Ph.D., Chair

___________________________________

Jeffri Bohlscheid, Ph.D.

___________________________________

Marc Evans, Ph.D.

iii

ACKNOWLEDGEMENTS

I would first and foremost like to acknowledge my father, who passed away on

February 7th

, 2010. I think his excitement over me going to graduate school would only be

surpassed by his excitement at my graduating. I appreciate and am reminded every day of

his love, humor, support, and understanding, all of which has led me to this point. His

passing during my time at graduate school has made me realize the importance of expressing

how much you appreciate the people around you and to never miss the opportunity to thank

someone for their contribution to your life.

In that spirit, there are so many people at Washington State University that I would

like to acknowledge. First, I would like to thank my advisor Carolyn Ross for the

opportunity to come to graduate school and pursue a project involving flavor chemistry. Her

support throughout this endeavor has made this project possible. I would like to thank my

committee members, Jeff Bohlscheid and Marc Evans for their continual assistance which

included answering numerous questions and giving excellent advice. I would also like to

acknowledge Barbara Rasco, who helped me with editing and was always available for

advice and discussion on many different topics.

I would like to thank Karen Weller who was extremely helpful throughout my time at

WSU and contributed so much to making this project successful. I would also like to thank

Jodi Anderson for kindly answering all the questions I had for her both before and after I

arrived at WSU. I would also like to acknowledge Scott Mattinson who contributed a great

deal of time answering the many questions I had regarding the gas chromatography analysis

performed in my study. I would like to thank all the members of the Ross Lab. We

iv

experienced all the ups and downs of research together, and from panels to in the lab, there

was always a question to be answered or a problem to be solved.

To my best friends, Sean, Jesse, and Steve—I could not have made through this time

without them. They have seen me through all the different emotions I have experienced and

all the changes in my life over the past two years. Thank you to Sean, who has been patient

and helpful beyond what I ever would have expected from anyone. Thank you to Jesse, who

has been available to both talk and listen, and always makes me laugh with her right-on

remarks. Thank you to Steve, who has always helped me reason out obstacles with practical

advice.

Finally, I would like to acknowledge my mother, who has always been my number

one cheerleader. She has helped me forge ahead and always keeps me looking towards the

future.

v

PERCEPTION OF FINISH IN

WHITE WINE

Abstract

by Emily Goodstein

Washington State University

December 2011

Chair: Carolyn F. Ross

Wine finish is a sensory attribute associated with wine quality. Wine quality

as it relates to finish is currently an understudied area of sensory research. This study aimed

to perform both a trained and consumer evaluation of finish in white wines, with additional

support provided by the analytical analysis of commercially available unoaked and oaked

Chardonnay wines. The hypothesis of this study was that certain flavors such as coconut and

mushroom would take longer to finish compared to fruit and floral flavors. Model and base

wines were prepared through the addition of four chemical compounds; ethyl hexanoate

(fruity), linalool (floral), oak lactone (coconut), and 1-octen-3-ol (mushroom). Trained

panelists executed time intensity analysis of these four compounds for the aroma attributes of

fruity, floral, coconut, and mushroom. Model wine evaluation of single compounds showed

fruity flavor finished significantly earlier than all other flavors. Model wine evaluation of

multiple flavors in solution indicated that fruity flavor displayed the shortest finish, with

mushroom acting as a suppressor of other flavors. Normalization of model wine time

intensity analysis showed that data variation was largely attributed to the contrasting

relationship between fruity and mushroom flavors. Base wine analysis did not yield

significant differences in flavor finish length. A trend of floral flavor suppression by coconut

vi

flavor emerged when analyzing time intensity parameters for intensity maximum (Imax) and

area under the curve (AUC).

Consumer panel analysis of finish of commercially available Washington State

Chardonnays indicated that consumers were able to discern differences in finish length

between unoaked and high oaked wines. Consumers significantly preferred unoaked wines

over high oak wines and were willing to pay higher dollar values per bottle for unoaked

wines than high oak wines. Solid phase microextraction-gas chromatography-mass

spectrometry was utilized to quantify cis- and trans-oak lactone concentration of unoaked,

medium, and high oak Chardonnay samples. Increases in oak treatment corresponded to

increases in cis- and trans-oak lactone concentrations. This study provides the first rigorous

examination of wine finish and its relationship to both the presence of different flavor

compounds and consumer acceptance.

vii

TABLE OF CONTENTS

Page

ACKNOWLEDGEMENTS .................................................................................................... iii

ABSTRACT ............................................................................................................................. v

LIST OF TABLES ................................................................................................................... x

LIST OF FIGURES ............................................................................................................... xv

CHAPTER

1. INTRODUCTION ................................................................................................. 1

2. LITERATURE REVIEW ...................................................................................... 3

Wine finish ................................................................................................. 3

Economic relevance ................................................................................... 3

Wine processing ......................................................................................... 4

Non-volatile components of wine .............................................................. 5

Volatile components of wine ..................................................................... 6

Volatiles to be analyzed by a trained panel ............................................... 7

Floral…………………………………………………………………..8

Fruity…………………………………………………………………10

Mushroom…………………………………………………………….10

Barrel aged/coconut…………………………………………………..11

Model wine .............................................................................................. 12

Flavor perception ..................................................................................... 13

Time intensity .......................................................................................... 14

Barrel aging of white wine ....................................................................... 17

viii

Gas chromatography-mass spectrometry analysis of oak lactone ........... 19

3. MATERIALS AND METHODS ......................................................................... 21

Model wine .............................................................................................. 21

Wines ....................................................................................................... 21

Trained panel ........................................................................................... 22

Trained panel demographics .............................................................. 22

Training sessions ................................................................................ 22

Trained panel formal evaluations....................................................... 24

Data analysis ...................................................................................... 25

Consumer panel ....................................................................................... 26

Consumer evaluation ......................................................................... 26

Data analysis ...................................................................................... 27

Gas chromatography-mass spectrometry (GC-MS) ................................. 27

4. RESULTS AND DISCUSSION .......................................................................... 29

Trained panel ........................................................................................... 29

Single flavor compound analysis in model wine ............................... 29

Multiple flavor compound analysis in model wine ............................ 31

Single flavor compound analysis in base wine .................................. 35

Multiple flavor compound analysis in base wine .............................. 36

Consumer panel evaluation of commercial Chardonnay ......................... 39

Gas chromatography-mass spectrometry ................................................. 40

5. FIGURES AND TABLES ................................................................................... 42

6. CONCLUSIONS AND FUTURE RESEARCH ............................................... 103

ix

7. REFERENCES .................................................................................................. 106

x

LIST OF TABLES

Page

1. Analysis of variance (ANOVA) of time to maximum

intensity (Tmax) for single compounds in model wine samples

analyzed by a trained panel (n=10) ............................................................................ 42

2. Analysis of variance (ANOVA) of maximum intensity

(Imax) for single compounds in model wine samples analyzed

by a trained panel (n=10) .......................................................................................... 43

3. Pairwise comparison of maximum intensity (Imax) evaluations

for single compounds in model wine samples analyzed by a

trained panel (n=10) .................................................................................................. 44

4. Analysis of variance (ANOVA) duration (Tend) for single

compounds in model wine samples analyzed by a trained panel (n=10) .................. 45

5. Pairwise comparison of duration (Tend) evaluations for single

compounds in model wine samples analyzed by a trained

panel (n=10) ............................................................................................................... 46

6. Analysis of variance (ANOVA) of area under the curve

(AUC) for single compounds in model wine samples analyzed

by a trained panel (n=10) .......................................................................................... 47

7. Pairwise comparison of area under the curve (AUC) evaluations

for single compounds in model wine samples analyzed by a

trained panel (n=10) .................................................................................................. 48

8. Mean Separation of Maximum Intensity (Imax), Duration (Tend),

and area under the curve (AUC) for single compounds in model

wine samples analyzed by the trained panel (n=10) .................................................. 49


intensity (Tmax) for two compounds in model wine samples


10. Analysis of variance (ANOVA) of maximum

intensity (Imax) for two compounds in model wine samples


11. Analysis of variance (ANOVA) of duration (Tend) for two


panel (n=10) ............................................................................................................... 58

xi


(AUC) for two compounds in model wine samples analyzed

by a trained panel (n=10) ........................................................................................... 59

13. Mean separation of contrasts of the four flavors analyzed in

two compound model wine solutions evaluated by a trained

panel (n=10) ............................................................................................................... 60


intensity (Tmax) for four compounds in model wine samples



(Imax) for four compounds in model wine samples analyzed


16. Analysis of variance (ANOVA) duration (Tend) for four

compounds in model wine samples analyzed by a trained panel (n=10) ................... 64

17. Pairwise comparison of duration (Tend) evaluations for four


panel (n=10) .............................................................................................................. 65


(AUC) for four compounds in model wine samples analyzed



for four compounds in model wine samples analyzed by a

trained panel (n=10) ................................................................................................... 67

20. Mean separation of duration (Tend) and area under the curve

(AUC) for four compounds in model wine samples analyzed by the trained panel

(n=10) ......................................................................................................................... 68


intensity (Tmax) for single compounds in base wine samples



(Imax) for single compounds in base wine samples analyzed


23. Analysis of variance (ANOVA) duration (Tend) for single

compounds in base wine samples analyzed by a trained panel (n=10)...................... 72

xii


(AUC) for single compounds in base wine samples analyzed



intensity (Tmax) for two compounds in base wine samples



(Imax) for two compounds in base wine samples analyzed


27. Analysis of variance (ANOVA) duration (Tend) for two



(AUC) for two compounds in base wine samples analyzed


29. Mean separation of contrasts of the four flavors analyzed in

two compound base wine solutions evaluated by a trained

panel (n=10) ............................................................................................................... 79


intensity (Tmax) for four compounds in base wine samples


31. Pairwise comparison of time to maximum intensity (Tmax)

evaluations for four compounds in base wine samples analyzed



(Imax) for four compounds in base wine samples analyzed


33. Pairwise comparison of maximum intensity (Imax) evaluations for

four compounds in base wine samples analyzed by a trained

panel (n=10) ............................................................................................................... 84

34. Analysis of variance (ANOVA) duration (Tend) for four



(AUC) for four compounds in base wine samples analyzed


xiii


for four compounds in base wine samples analyzed by a trained

panel (n=10) ............................................................................................................... 87

37. Mean separation of time to maximum intensity (Tmax), maximum

intensity (Imax), and area under the curve (AUC) for four compounds

in base wine samples analyzed by the trained panel (n=10) ...................................... 88

38. Analysis of variance (ANOVA) table of time to maximum intensity

(Tmax) for three commercially produced Washington State

Chardonnay samples analyzed by a trained panel (n=10) ......................................... 90

39. Analysis of variance (ANOVA) table of maximum intensity

(Imax) for three commercially produced Washington State


40. Analysis of variance (ANOVA) table of duration (Tend) for three

commercially produced Washington State Chardonnay

samples analyzed by a trained panel (n=10) .............................................................. 92

41. Analysis of variance (ANOVA) table of area under the curve

(AUC) for three commercially produced Washington State


42. Analysis of variance (ANOVA) of consumer panel (n=60)

evaluation of Washington State Chardonnay finish with

timed finish perception .............................................................................................. 94

43. Pairwise comparison of finish time of Washington State

Chardonnays analyzed by a consumer panel (n=60) ................................................. 95

44. Mean separation of finish time (seconds) of Washington State

Chardonnays analyzed by a consumer panel (n=60) ................................................. 96

45. Analysis of variance (ANOVA) of consumer panel (n=60)

evaluation of Washington State Chardonnay finish with a

7-point hedonic scale ................................................................................................. 97

46. Mean separation of 7-point hedonic scale rating of Washington

State Chardonnays analyzed by a consumer panel (n=60) ........................................ 98

47. Coding of branched willingness to purchase responses determined

by the consumers (n=60) for the evaluation of the Washington

State Chardonnay ....................................................................................................... 99

xiv

48. Pairwise comparison of wines for branch willingness to purchase

response to Washington State Chardonnays by consumer panel (n=60) ................ 100

xv

LIST OF FIGURES

Page

1. Structure of coconut, floral, fruity, and mushroom compounds

utilized in the trained panel .......................................................................................... 9

2. Representative trained panelist time intensity evaluation of single

flavors in model wine................................................................................................. 50

3. Pearson‘s principle component (PC) analysis factor loadings and

factor scores from normalized time intensity data of single compounds

in model wine evaluation by a trained panel (n=10)................................................. 51

4. Comparison of standard data and normalized data from a representative

trained panelist‘s time intensity evaluation of the coconut flavor in

single compound model wine .................................................................................... 52


trained panelist‘s time intensity evaluation of the floral flavor in single

compound model wine .............................................................................................. 53


trained panelist‘s time intensity evaluation of the fruity flavor in single

compound model wine ............................................................................................... 54


trained panelist‘s time intensity evaluation of the mushroom flavor in

single compound model wine .................................................................................... 55

8. Pearson‘s principle component analysis factor loadings and factor

scores from normalized time intensity data of two compounds in model

wine evaluation by a trained panel (n=10) ................................................................. 61


scores from normalized time intensity data of four compounds in model

wine evaluation by a trained panel (n=10) ................................................................. 69


scores from normalized time intensity data of single compounds in

base wine evaluation by a trained panel (n=10) ........................................................ 74

xvi


scores from normalized time intensity data of two compounds in



scores from normalized time intensity data of four compounds in


13. Standard curve for cis-oak lactone in model wine (9% EtOH, 0.6%

fructose, pH 3.0) with three replicates ..................................................................... 101

14. Standard curve for trans-oak lactone in model wine (9% EtOH, 0.6%

fructose, pH 3.0) with three replicates ..................................................................... 102

1

CHAPTER 1

INTRODUCTION

Wine production represents a burgeoning industry in Washington State. Quality can be

influential in determining the economic success of a wine and the winery, many of which are

small businesses. Characteristic of high quality wine, wine finish refers to the lingering after-

flavor that occurs once wine has been swallowed or expectorated (Jackson 2000). Wine finish is

influenced by flavor, which is an important parameter of a wine sensory profile that can dictate

wine quality and/or consumer acceptance, thereby making wine flavor a topic of great interest to

researchers. Despite interest in wine flavor, wine finish has remained a largely unstudied topic.

The objectives of this study were to determine the impact of flavor compounds on white

wine finish. Evaluation of white wine was chosen over red wine. Although red wines typically

are associated with long finish, they are also associated with astringency. Because astringency

was not a sensory property of interest in the present study, the presence of this attribute could be

distracting to the panelist. This distraction could result in what is known as ―attribute dumping‖

(Lawless and Heymann 1998). ―Dumping‖ occurs when an obvious trait is not evaluated—

panelists attempt to evaluate the conspicuous trait by ―dumping‖ their thoughts and perceptions

onto another trait (Lawless and Heymann 1998). As this can cause misleading results, it was

preferable to utilize white wine for evaluation as it lacks the astringency of red wine that could

cause the ―dumping‖ phenomenon. In addition, astringency is known to be a challenging and

fatiguing sensory attribute to evaluate thus by excluding this from the study, the panelists were

able to focus on wine flavor.

In this study, a trained and consumer panel evaluated white wine finish. The objective of

the trained sensory panel was to perform a time intensity analysis of the finish of selected white

wine flavors in model wine and base wine to determine how different flavor compounds affected

2

and contributed to finish perception. These trained panelists evaluated ethyl hexanoate (fruity

flavor), linalool (floral flavor), oak lactone (coconut), and 1-octen-3-ol (mushroom). A

secondary objective of this panel was to determine the impact of presenting multiple flavor

compounds to panelists. It was hypothesized that compounds associated with fruit and floral

notes would finish earlier than compounds associated with coconut and mushroom notes. In

addition to a trained panel where panelists performed a time intensity analysis of four distinct

flavors, a consumer panel for the evaluation of overall finish was performed.

The consumer sensory panel evaluated commercial Chardonnay wines that were

subjected to different oaking regimes, corresponding to no oak, medium and high oak wines.

Consumers evaluated the length of finish, liking of finish, and their willingness to purchase the

wine. It was hypothesized that consumers would be willing to pay higher dollar amounts for

wines with a longer finish. Concurrent with the consumer panel, gas chromatography-mass

spectrometry (GC-MS) analysis of the wines was performed to quantify the concentration of oak

lactone in the Chardonnay wine. It was hypothesized that no oak, medium and high oak

Chardonnay wines would contain increasing concentrations of oak lactone.

This study of white wine finish yielded an idea of what flavors contribute to a

lengthening of wine finish perception. Since extended finish is an attribute that has been

associated with high quality wines, this study will allow wine producers to evaluate the presence

and methods to generate these flavors in their wines. By altering wine quality perception by

lengthening finish, economic success may be increased

3

CHAPTER 2

LITERATURE REVIEW

Wine Finish

Wine finish is defined by Jackson as the lingering taste following the swallowing of wine

(Jackson 2000). Lengthy finish is most notable in fine red wines but is also observed in

Sauvignon Blanc and Chardonnay white wine varieties (Amerine and Rossler 1983). Typically,

it is believed that certain flavors are associated with different lengths of finish. Fruity and floral

flavors are thought to have a shorter finish while oak, spice, and earthy flavors are thought to

have a longer finish. However, this is a relatively unexplored area and commonly accepted

theory is based largely upon conventional wisdom rather than scientific data.

Finish is an attribute closely related to quality. Longer finish is associated with higher

quality wines (Jackson 2002). High quality dictates the selling price and cache associated with a

particular wine. Studying the finish of different white wine flavors could indicate what flavors

are important determinants of perceived wine quality, and can therefore be related to economic

relevance.

Economic Relevance

The wine industry represents an important economic entity to many countries around the

world (Washington Wine Commission 2007). Historically, European countries such as France

and Germany have been major producers of white wine (Jackson 2000). But, in recent years,

there has been an increase in the number of countries attempting to create successful wine

industries. Wine production in Washington State has grown into an important economic

enterprise. The Washington Wine Commission reports that 12 million cases of wine are

produced annually in Washington State. This contributes approximately three billion dollars to

4

the Washington State economy and approximately 4.7 billion dollars to the national economy

(Washington Wine Commission 2007).

A significant portion of grapes grown in Washington State are white varietals, with

Riesling, Chardonnay, and Pinot Gris grapes ranking as the top three (Washington Wine

Commission 2007). Due to the economic importance of white wine production within

Washington State, it has become important to study factors that alter white wine quality. .

However, determining what factors affect white wine quality can be a difficult matter as wine is

a complex solution made up of non-volatile and volatile components (Jackson 2000). In order to

understand how these elements of physical composition arise, wine processing procedures must

be observed.

Wine Processing

The processing of grapes into a finished wine can have a distinct effect on white wine

flavor. The process begins by harvesting the grapes, either by hand or machine. Grapes are

prepared for crushing by removing the stems, which are a source of unwanted phenols. Grapes

are then crushed to form a must, a mixture of grape juice, skins, and seeds. Crushed grapes

inadvertently go through maceration. This is not always considered a desirable process for white

grapes, as it can lead to the collection of phenolic compounds in certain varietals (Jackson 2000).

In the context of grape must, phenols can lead to oxidative browning (Servili and others 2000).

However, maceration is thought to increase the amount of flavor compounds (Ramey and others

1986).

The must is then pressed and clarified, after which producers will make adjustments to

the juice composition to ensure optimal fermentation (Jackson 2000). Juice is either allowed to

ferment utilizing its naturally present microbial flora, or it is inoculated with a pure and/or

5

known strain of yeast to facilitate more successful fermentation results. At this point, certain

wine varieties may be stored in oak barrels. Barrel aging of wines is usually reserved for fine red

wines and certain white wine varieties such as Chardonnay and some Sauvignon Blanc. After

fermentation, modifications are made to the wines including sugar and acidity adjustment. Wine

is then stored in either stainless steel or oak barrels for aging before bottling (Jackson 2000).

Storage in oak barrels provides the wine exposure to flavor compounds that originate in the

wood and oxygen, both of which are thought to positively affect the wine (Wilker and Gallander

1988). Wine processing measures lead to the creation of the wine matrix which consists of a

combination of non-volatile and volatile components.

Non-volatile Components of Wine

Non-volatile wine components include sugars and acids which contribute to sweet and

sour tastes, respectively. The tastes created by the presence and proportion of these sugars and

acids have been described as the dominant contributors to white wine flavor and also a major

influence on the perception of wine quality (Zamora and others 2006). A study by Zamora and

others in 2006 discussed the importance of sweet and sour taste perception in white wine. In this

study, taste perception in solutions with tartaric acid to a pH of 3.0 and 3.8 and fructose at 11.1

mM and 38.9 mM were evaluated by panelists. The study found that tartaric acid acted as a

suppressor of fructose. Fructose also suppressed tartaric acid, but not as effectively. So, fructose

to tartaric acid ratio effects the perception of sweet and sour tastes (Zamora and others 2006).

Grape harvest is often timed in order to pick grapes when they have reached a desirable balance

between acid and sugar concentration (Lund and Bohlmann 2006). Acidity levels are higher

before grape maturation due to the development and storage of acids in storage vacuoles within

the grape mesocarp. As the grape matures, sugars such as glucose and fructose become more

6

abundant, while malic acids are utilized in metabolic processes (Combe 1992). Therefore,

appropriate timing of harvest can be highly influential in wine flavor and quality perception.

Polyphenolic compounds are also constituents of the non-volatile fraction of wine and

include compounds such as tannins and catechins, which contribute to astringent and bitter

sensations, respectively (Arnold and others 1980, Robichaud and Noble 1990). Research has

shown that perception of white wine quality is negatively influenced by increasing

concentrations of phenolic compounds due to significant increases in perceived astringency

(Arnold and Noble 1978). A 1978 study by Arnold and Noble showed that panelists found

significant differences when evaluating 25, 80 and 130 mg/L phenolic extract in model wine.

Perception of astringency increased significantly as phenolic concentration increased (Arnold

and Noble 1978). Although tannins and catechins are more typically characteristic of red wines,

flavonoid tannins have been noted to contribute to bitterness in Riesling wine varieties. A 2008

study by Etaio and others demonstrated that wine produced by cofermentation of Viura grapes (a

white variety) with Tempranillo grapes (a red variety) displayed decreased overall polyphenol

content of 6 abs280 and tannin concentration of 0.23 g/L when compared to Tempranillo wine

alone (Etaio and others 2008). Water, glycerol, and nonflavonoid phenols are also considered

non-volatile constituents of wine (Jackson 2000).

Volatile Components in Wine

Volatile compounds are numerous in white wine and other alcoholic beverages.

According to a 2001 review by Ebeler, over1300 compounds have been successfully dectected in

wine, beer, malt beverages, brandy, and distilled spirits. Major categories of flavor compounds

include terpenes, phenols, ethyl esters, acetate esters and lactones (Ebeler 2001). Different

compounds may be responsible for different perceived flavors. For example, monoterpenes are

7

associated with floral notes (Marais 1983) whereas oak lactones are associated with coconut

notes (Waterhouse and Towey 1994).

In a 1997 study by Guth, 44 odor active compounds were quantified in Scheurebe and

Gewürztraminer white wine varieties (Guth 1997). A 2003 study of Chardonnay wines utilizing

gas chromatography-olfactometry found 81 odor active volatile compounds and identified 61 of

these compounds (Lee and Noble 2003). Both studies examined odor active compounds,

meaning compounds that contribute to aroma. A compound that is not odor active can be present

but may not create noticeable impact on wine aroma.

Alcohols also represent a major volatile element present in wines. Ethanol is the main

alcohol present in wines, and is the product of yeast fermentation (Ebeler 2001). Ethanol plays a

number of important sensory roles in wine. It is responsible for burning sensations and also at

appropriate concentrations, sensations of sweetness. A 2006 study by Zamora and others

showed that increasing ethanol percentage from 4% to 12% led to larger sweetness intensity

ratings when panelists were evaluating model wine solutions (Zamora and others 2006). Other

alcohols are also present in wines including fusel alcohols (i.e. hexanol and 1-octen-3-ol) and

methanol (Jackson 2000). Although there are numerous volatile compound found within the

wine matrix, four compounds were chosen for analysis by trained panelists in the present study.

Volatiles to be Analyzed by a Trained Panel

Compounds were chosen for this study based on a variety of factors. Choices were based

on the Wine Aroma Wheel published by Noble and others in 1987. The Wine Aroma Wheel

categorizes aroma notes from general to specific terms, in order to provide a lexicon of

terminology with which wine aroma can be described (Noble and others 1987). The objective

was to utilize four compounds that represented four distinct areas of the Aroma Wheel. With

8

this in mind, a representative compound was found for floral, fruity, mushroom, and oak.

Compounds were also determined based on cited odor activity or presence in at least one white

wine variety. Molecular weight was taken into account to avoid the usage of extremely volatile

compounds that would be inappropriate for sensory evaluation.

Floral

Linalool (Figure 1a) is a volatile terpene (Clarke and Bakker 2004) associated with

green, floral and citrus notes (Burdock 20009). With a reported taste threshold of 5 ppm

(Burdock 2010), linalool was chosen due to notable floral qualities. The compound is found in

numerous white grape varieties, but present at highest quantities in Gewürztraminer (0.006-0.19

ppm), Sheurebe (0.007-0.370 ppm), and Morio-Muscat (0.16-0.28 ppm) cultivars (Schreier and

others 1977). Linalool has also been reported in Chardonnay, but at a lower concentration (0.1

ppm respectively) (Aldave and other 1993).

Linalool can be available in different forms within a grape, some of which are non

volatile (Ebeler 2001). These forms include the free, or volatile, linalool, oxidized linalool, and

glycosidically bound linalool (Wilson and others 1986). A 1986 study by Wilson and others

found that free linalool was present in equivalent amounts in the skin and juice of Muscat grape

varieties, while a lower concentration of free linalool was found in the grape pulp. These

findings were contrary to the popular theory that terpenes were generally a constituent of the

grape skin (Wilson and others 1986).

Research has also been performed to determine the impact of processing conditions and

different levels of grape maturation on linalool concentration. Marais and van Wyk

demonstrated that maturation caused significant increases in linalool concentration in Riesling

juice with concentrations increasing from 0.49 to 5.35 µg/L. While studying four maturation

9

Figure 1. Compound structures for a) linalool, b) ethyl hexanoate c) 1-octen-3-ol and d) oak

lactone

a)

b)

c)

d)

10

levels, which corresponded to 16.1, 18.8, 18.7, and 19.7 Brix, continuous increases in linalool

were seen in grapes in the first three levels (16.1-18.7 Brix), after which decline to 0.35 µg/L

was noted. Since continuous increases in linalool concentration were observed during grape

maturation, the researchers postulated that results were caused by the occurrence of hydrolytic or

enzymatic reactions. These reactions may have either released bound linalool or converted

precursor compounds into linalool during processing (Marais and van Wyk 1986).

Fruity

Ethyl hexanoate (Figure 1b) is an ethyl ester that is associated with fruity (Burdock

2009) and apple (Genovese and others 2007) notes. With a taste threshold of 10 ppm, ethyl

hexanoate was chosen due to its high odor activity in numerous white wine varieties. Odor

activity has been documented in Gewürztraminer and Scheurebe (Guth 1997), Fiano (Genovese

and other 2007), and Chardonnay wines (Lee and Noble 2003).

Ethyl esters are formed as a result of fermentation (Bardi and others 1998). During

fermentation, fatty acids are released and are esterified with alcohols (Nordstom 1964).

Synthesis of ethyl hexanoate is influenced by fermentation conditions (Nordstom 1964).

Anaerobic conditions resulted in the formation of 5 fold higher concentrations of ethyl hexanoate

than semi-anaerobic conditions (Bertrand 1968). These results were strain dependent to S.

oviformis and S. ellipsoideus. Also, utilizing strains of S. cerevisiae resulted in higher levels of

ethyl hexanoate than in fermentations utilizing S.uvarum yeast strains (Bertrand 1968).

Mushroom

1-octen-3-ol (Figure 1c) is an alcohol associated with mushroom notes with a recognition

threshold of 10 ppm (Burdock 2009). Linked with fungal growth on grapes, this compound was

chosen due to its distinct mushroom aroma and flavor (La Guerche and others 2006). A 2006

11

study by La Guerche and others examined the mushroom and earthy compounds found in grapes

with fungal growth. Although typically associated with Botrytis cinerea growth, 1-octen-3-ol is

also produced by four other types of fungi. Quantification by gas chromatography-mass

spectrometry (GC-MS) revealed the presence of 1-octen-3-ol in Gamay and Semillion grape

varieties, while gas chromatography-olfactometry (GC-O) analysis indicated that 1-octen-3-ol

was an odor active compound with mushroom aroma. Sensory testing determined that 1-octen-

3-ol had a low sensory odor perception threshold in three different media (water: 2 µg/L, model

wine 20 µg/L, red wine 40 µg/L). Researchers noted this as an important factor, as low

concentrations could lead to noticeable impact on wine. The study found that 1-octen-3-ol is

also able to withstand fermentation with minimal degradation and/or decreases in concentration

(La Guerche and other 2006).

In additional to being a result of fungal growth on grapes, presence of 1-octen-3-ol can

also be attributed to cork taint (Ezquerro and Tena 2005). Cork taint is a wine fault that

negatively impacts wine odor (Jackson 2000). While cork taint is typically associated with 2,4,6

trichloroanisole, a 2005 study by Ezquerro and Tena determined that 1-octen-3-ol is also a

contributing compound to this wine fault (Ezquerro and Tena 2005).

Barrel aged/Coconut

Oak lactone is an important volatile compound present in wines aged in oak barrels, with

characteristics of vanilla, wood, and coconut (Burdock 2009). Lactones are classified as cyclized

esters, and have different flavor qualities based on isomerization and carbon chain length (de

Mann 1999). Also known by the names whiskey lactone, 4-hydroxy-3-methyloctanoic acid

lactone, and 2(3H)-furanone, oak lactone has a reported taste threshold of 0.5 ppm (Burdock

2009).

12

There are four isomers of oak lactone (Figure 1d). These forms include naturally

occurring (4S, 5S) cis-oak lactone and (4S, 5R) trans-oak lactone, and non-naturally occurring

(4R, 5R) cis-oak lactone and (4R, 5S) trans-oak lactone (Brown and others 2006). A sensory

analysis of all four oak lactone isomers was performed by Brown and others in 2006 to

determine if sensory differences in aroma detection exist between the four isomers of oak-

lactone. Aroma detection threshold tests were performed by untrained panelists (n=33 to 43)

using red and white wine as the base for presenting the four isomers of oak lactone. In both red

and white wine, aroma detection thresholds were lower in naturally occurring cis-oak lactone

than in naturally occurring trans-oak lactone. Duo-trio tests were performed by untrained

panelists (n=36) to evaluate if there was a perceivable difference in aroma between oak lactone

enantiomers in wine. There was not a significant difference between naturally occurring and

non-naturally occurring trans-oak lactone enantiomers, but significant differences were observed

between naturally occurring and non-naturally occurring cis-oak lactone enantiomers. Duo-trio

testing comparing wines with naturally occurring cis-oak lactone to wines with naturally

occurring cis and trans- oak lactone (150 µg/L in white wine, 300 µg/L in red wine for all

isomers) did not yield significant results. The study concluded that cis-oak lactone has a greater

impact on wine aroma than the trans-oak lactone in red and white wine (Brown and other 2006).

Model Wine

Model wines act as a simplified wine matrix, containing only a limited number of

components. In the present study, trained panelists evaluated a model wine consisting of water,

ethanol, tartaric acid, and fructose. Tartaric acid and fructose were chosen based on their natural

presence in wine (Jackson 2000). Examples of model wine usage include the previously cited

study by Zamora and others in 2006, where sweet and sour perception was evaluated in the

13

presence of differing levels of ethanol. Model solutions contain 2, 4, or 12% ethanol combined

with 11.1 mM or 38.9 mM fructose, 3.0 or 3.8 pH (modified by tartaric acid) (Zamora and others

2006). A 1999 study by Kadim and Mannheim utilized model wine with 12% ethanol,

potassium hydrogen tartrate, and tartaric acid to 3.2 pH to observe the kinetics of phenolic

extraction from oak during the aging process. For the purpose of the current study, model wine

enabled panelists to isolate specific volatile compounds and determine their sensory effects.

Flavor Perception

Perception of flavor (i.e. strawberry, peach, mushroom) is often confused with taste.

Taste includes perception of bitter, salty, sweet, sour, and umami sensations. These perceptions

occur due to the presence of buds which line the tongue, epiglottis, and soft palate (Jackson

2002). Taste buds are receptors and their stimulation leads to signaling which travels to the

thalamus and cortex resulting in the sensation of taste (Jackson 2002). Flavor is considered the

result of retronasal stimulation (Meilgaard and others 2007). This occurs when volatile

compounds in the mouth and/or upper digestive tract are able to stimulate receptors in the nasal

passage. The stimulation of nasal receptors elicits the transmission of a signal to the brain,

specifically thought to be the orbitofrontal cortex, resulting in the perception of flavor (Jackson

2002). Some researchers feel that flavor should be defined as any perception originating in the

mouth (Meilgaard and others 2007). Meilgaard and others state that this would mean that taste

and flavor are combined under the same descriptor. Still, it is important to recognize that

different pathways are involved in eliciting these sensations. The focus of the current study will

be on flavor finish and not taste or mouthfeel finish. Panelists will be required to separate flavor

from taste and mouthfeel sensation when performing time intensity analysis.

14

Time Intensity

Time intensity is a method of sensory evaluation that allows the panelist to evaluate the

intensity of an attribute over a period of time. This method has been used to measure flavor

changes in chewing gum (Ovejero-Lopez and others 2005), evaluate astringency of alcoholic

beverages (Valentova and others 2002), and analyze differences in sweeteners (Bonnans and

Noble 1993). In wine, studies have been performed to analyze bitter and astringent sensations

caused by phenols (Robichaud and Noble 1990), astringency and sweetness

perception/interaction (Ishikawa and Noble 1995) and effectiveness of different palate cleansers

(Ross and others 2007).

The time period can be either fixed or unlimited based on the purpose of the study and/or

the discretion of the researcher. The data collected over a time period can then be evaluated

based on summary statistics which include Tmax, Imax, duration (Tend), and area under the curve

(AUC). Tmax indicates the time it takes to reach maximum intensity, whereas Imax indicates the

maximum intensity based on a scale of 0-100% low to high intensity. Duration indicates the

length of time a panelists takes to return to 0% intensity. Area under the curve (AUC) is an

integration which has a value that‘s interpretation can differ depending upon study objectives

(Meilgaard and others 2007). Values for increasing and decreasing inclines and their

corresponding areas are also generated. This is typically used as a computer based method and

the test requires that panelists be focused, understand the attribute in question, and understand

how to utilize the computer program. As a result of these requirements and the relative

complexity of time intensity tests, it is utilized in trained panels.

Due to the multitude of data points generated by the panelist, there is variation even

among trained panelists. Therefore, much discussion and research has been devoted to

15

determining the best method of analysis. The most basic analysis requires that the summary

statistics listed above be compared using analysis of variance (ANOVA) to determine significant

differences (MacFie and Liu 1992). However, in a 1992 paper, researchers MacFie and Liu

asserted that this method is insufficient as it does not take information from the entire curve.

MacFie and Liu proposed a normalization method that averages curves generated by all panelists

for a particular sample. Normalization occurs for both intensity and time as follows:

Intensity normalization:

Time normalization:

(MacFie and Liu 1992)

The intensity normalization equation compares the mean Imax value (Imax) to the

individual Imax value (Imax i). The time normalization also compares mean time values (tmax, tend,

tdec) to individual time values (tmax i, tend i, tdec i), but is more complicated as there are separate

equations to analyze the ascending (to Tmax) and descending (from Tmax) portions of the curve.

MacFie and Liu state that this method is most appropriate because it normalizes both variables

and preserves qualities of the curve created by the original data (MacFie and Liu 1992).

Principle Component Analysis (PCA) is also suggested by MacFie and Liu as a potential method

for analyzing data, particularly if great differences are seen in the data generated by the panelists.

PCA focuses on grouping of similar attributes and thus may aid in identifying similarities when

results seem quite dissimilar (MacFie and Liu 1992).

16

Time intensity studies have previously been performed with different types of alcoholic

beverages. A 2000 study by Piggott and others demonstrated the utilization of time intensity for

Scotch malt whisky. The objective of this study was to determine if time intensity could be used

as a viable testing method and determine the effect of barrel aging on a Scotch malt whisky

attribute. Five different barrel types were utilized and whisky was aged for 24, 30, 42, and 60

months. Panelists (n=13) evaluated the time intensity of sweetness over a 60 second time period

that consisted of a ten second in mouth holding time followed by swallowing and aftertaste

perception. Based on the data collected in this experiment, the researchers performed a

comparison of different data analysis methods. The researchers determined that no method of

analysis produced statistically significant results, which they felt was due to the extreme

variation observed between panelists. Utilization of the 1992 MacFie and Liu method previously

described provided what Piggott and others considered the most reliable results (Piggott and

others 2000).

Several studies have evaluated flavor perception over time with time intensity testing. A

1997 study by Mialon and Ebeler examined perception of vanillin and limonene compounds in

emulsions of varied oil to water ratios. Panelists evaluated flavor intensity for both in mouth

intensity and post expectoration intensity. Salivary flow was measured for each panelist, as it is

thought that factors that alter environment of the mouth may influence retronasal perception

(Mialon and Ebeler 1997). The two compounds did not display similar patterns of change as

the oil to water ratio in which the compounds were presented in was changed. Most notable,

duration of vanillin perception was heightened as oil content was raised. Significant differences

were observed when comparing duration of vanillin perception in the presence of 0% oil and

50% oil. Researchers felt this change may have been influenced by the salivary flow of

17

individual panelists. Limonene was perceived as less intense by panelists when oil content was

increased in the emulsion. The researchers attributed this to the nonpolar nature of the

compound (Mialon and Ebeler 1997). Time intensity studies provide unique methodology to

evaluate flavor perception. In the current study, flavor perception based on effects of barrel

aging procedures was be analyzed.

Barrel Aging of White Wines

Numerous studies have been completed regarding barrel aging of both red and white

wine. This is a topic of interest due to the unique flavor and aroma qualities oak barrels can

impart to finished wines. Barrel aging is related to the introduction/production of at least ten

volatile compounds. These compounds include furfural, eugenol, guaiacol, and oak lactone

(Towey and Waterhouse 19962). Additionally, profiling of flavor compounds derived from oak

barrels can give some indication of the origin and seasoning/treatment of the wood utilized

during aging.

Since oak lactone has been determined to have sensory impact on white wine, it is

important to determine factors that can lead to the increase or decrease of oak lactone throughout

the barrel aging process. A 1996 Towey and Waterhouse study examined the effect of the age of

the barrel on the volatile extraction and subsequent volatile composition of Chardonnay wines

aged in American, French, and Hungarian oak barrels. The researchers examined numerous

volatile compounds, including oak lactone. Oak lactone levels were highest in wines aged in one

year old barrels. Also, oak lactone were found at higher concentration in wines aged in new

barrels than in wines aged in barrels that were two years of age. Origin of oak also affected the

levels of oak lactone. New American and Hungarian oak barrels were found to lead to the

18

extraction of a lower concentration of volatiles than new French oak barrels (Towey and

Waterhouse 1996).

In an earlier Waterhouse and Towey study performed in 1994, GC was utilized to analyze

Chardonnay aged in American oak barrels and Chardonnay aged in French oak barrels. The

objective of this study was to determine whether or not ratios of cis and trans- oak lactone could

be utilized as markers of wood origin. The researchers stated that the ability to distinguish

between American and French oak had economic significance, as French oak was a more

expensive product. Measurements of alcohol content, pH, and titratable acidity did not differ

significantly between Chardonnays aged in American oak versus Chardonnays aged in French

oak. However, differences were observed when comparing levels of oak lactone found in

American and French oak aged wines. American oak aged Chardonnay typically had greater

levels of oak lactone, but researchers saw discrepancies among the data, with oak originating in

Oregon producing the lowest oak lactone levels of all oaks. Discrepancies in oak lactone levels

observed among wines exposed to American and French oak were not consistently observed.

Therefore, researchers deemed that this was not an appropriate parameter by which to distinguish

American and French oaks. Significant and consistent discrepancies were noted between

American and French ratios of cis/trans oak lactone. American oak had significantly higher

cis/trans oak lactone ratios (mean=6) than French oaks (mean=1.3). It was concluded that

analysis of cis/trans oak lactone ratios via GC was a viable method for determining wood origin

of barrels (Waterhouse and Towey 1994).

The oak barrel aging of wines impacts sensory characteristics by intensifying aroma notes

such as vanilla and oak. Oak extracts are thought to play a role in the intensification of wine

aromas. The objective of a 1992 study by Francis and others was to determine if wood origin

19

and processing contributed to the sensory qualities of oak wood extracts presented in a model

wine. Oak wood was collected from three French forests and one American forest. Three

seasoning treatments were evaluated: green, seasoned in the country of origin, and seasoned in

Australia. The oak was then either heat treated or not heat treated. Quantitative difference

testing was performed to analyze the aroma of the oak extracts. Country of origin was a

significant factor, with oak aromas in French oak being comparable or significantly more intense

than oak aromas in American oak. Green versus seasoned treatments were significantly different

for the aroma intensity of cedar, nutty, and raisin notes. Significant differences were also found

when comparing country of seasoning. There was a significant difference between the unheated

and heated treatments for the intensity of caramel, cedar, nutty, and raisin aromas. Oak origin,

seasoning, and heat treating created significant differences in aroma perception of oak extracts.

These differences can be observed by utilizing GC-MS.

Gas Chromatography-Mass Spectrometry Analysis of Oak Lactone

A method optimization for quantification of barrel aging compounds in wine using solid-

phase microextraction gas chromatography mass spectrometry (HS-SPME-GC-MS) was

performed in a 2006 study by Carrillo and others. In this optimization, researchers altered

numerous factors and utilized model systems and red wine. The researchers explored the use of

four different types of SPME fibers, with divinylbenzene–carboxen–polydimethylsiloxane

(DVB–CAR–PDMS) determined to be of greatest use in the procedure. Optimal sodium

chloride content in headspace SPME vials was determine to be 30% (w/v) of the solution.

Finally extraction temperature and time were found to be significant factors when optimizing the

quantification of barrel aging compounds. Long time and high temperature were necessary to

elicit well-resolved peaks. The researchers recommended a 60 minute extraction carried out at

20

70° Celsius. With these parameters, the detection range for cis-oak lactone was 4-222 µg/L-1

(Carrillo and others 2006). Quantification of oak lactone by in a Towey and Waterhouse study

examining barrel aged Chardonnay resulted in a range of mean cis-oak lactone 89.2 – 188.9 µg/L

(Towey and Waterhouse 19962). This indicated that the method established by Carillo and other

displayed appropriate sensitivity for quantification of oak lactone in white wines.

Conclusions

Wine is a complex matrix and can thereby be influenced by numerous factors. Harvest

time, fermentation, and barrel aging are all factors that can significantly alter the composition of

either the non-volatile or the volatile fraction of white wines. By altering composition of the

wine, it is possible that wine finish is altered. Barrel aging provides an excellent example of this.

The barrel aging process is typically utilized on wine varieties such as Chardonnay which is

generally noted for having a comparably long finish for a white wine. Research shows that

barrel aging leads to the extraction of numerous volatile compounds, including oak lactone, a

compound associated with wood, vanilla, and coconut sensory attributes. Although research

correlating the production of oak lactone to wine finish has not previously been pursued, logic

dictates a potential connection between the enhanced finish of barrel aged white wines and

extracted compounds like oak lactone. Wine flavor and finish are important measures of quality.

Wine quality can determine economic success and therefore, the study of white wine finish is an

economically relevant issue.

21

CHAPTER 3

MATERIALS AND METHODS

Model wine

Panelists were presented with 9% EtOH, 0.6% fructose, 3.0 pH model wine. Model wine

was prepared with milliQ water, 200 proof ethanol (Decagon, Pullman, WA), D – (-)-fructose

(Sigma Aldrich, St. Louis, MO), and DL-tartaric acid (Sigma Aldrich, St. Louis, MO). Fruity,

floral, coconut, and mushroom were the descriptors used for 65.175 mg/L ethyl hexanoate, 43.1

mg/L linalool, 47.6 mg/L whiskey lactone, and 41.5 mg/L 1-octen-3-ol, respectively (Sigma-

Aldrich, St. Louis, MO). Concentrations were chosen based on reported flavor compound

threshold levels and bench testing. Flavor compound concentrations remained constant

throughout the panel.

Wines

Trained panelists and consumer panelists were given three different Washington State

Chardonnays. The wines evaluated were Columbia Crest unoaked Chardonnay (13.5% alcohol,

Paterson, WA 2008), Columbia Crest Horse Heaven Hills Chardonnay (13.5% alcohol, Paterson,

WA, 2008), and Columbia Crest Reserve Chardonnay (14.4% alcohol, Paterson, WA, 2008).

These wines corresponded to unoaked (not barrel aged), medium oak (70% oak barrel /30% steel

barrel aged), and high oak (100% oak barrel aged) treatments, respectively.

For the trained panel a base white wine was utilized. Carlo Rossi Chabli (9.5% alcohol,

Modesto, CA) was chosen as the base white wine. All wines, including the base wine and

Chardonnay wines, were purchased at the Pullman, Washington Safeway.

22

Trained Panel

Trained Panel Demographics

Ten panelists from the Washington State University School of Food Science community

participated in the study. Panelists were respondents to electronic advertisements. The panel

included 8 females and 2 males. Ages ranged from 24-62 with a mean age of 39.3 and a median

age of 31.5. All panelists reported drinking wine at least once a month with Chardonnay and

Riesling varieties being the most popular varieties. Panelists participated in a 24 sessions held at

the WSU Sensory Evaluation Facility. Non-monetary incentives were awarded at the end of

each completed session, with a final non-monetary incentive awarded for completion of the

trained panel.

Training Sessions

Training consisted of six one-hour sessions. On the first day of training, panelists were

introduced to the concept of wine finish with finish being defined as the lingering taste following

the swallowing of wine (Jackson 2000). Panelists sampled three commercial Chardonnays in

order to evaluate finish length, flavor, and intensity characteristics. The commercial wines

included two barrel aged Chardonnays and one unoaked Chardonnay. Wines were purchased

from Safeway (Pullman, WA) and included Toad Hollow Vineyards Unoaked Chardonnay

(13.9% alcohol, Healdsburg, CA 2009), Columbia Crest Horseheaven Hills Chardonnay (13.5%

alcohol, Paterson, WA 2008), and Toasted Head Chardonnay (14.5% alcohol, Woodbridge, CA

2009). Panelists had a group discussion about perceived finish differences between the wines,

and came to the consensus that barrel aged Chardonnays had longer finish than unoaked

Chardonnay.

23

The second training day included an introduction of the Wine Aroma Wheel (Noble and

others 1987). Panelists then evaluated four samples of EtOH, 0.6% fructose, 3.0 pH model wine.

Each sample contained one flavor compound (65.175 mg/L ethyl hexanoate, 43.1 mg/L linalool,

47.6 mg/L whiskey lactone, or 41.5 mg/L 1-octen-3-ol). Panelists were given the descriptors

fruity, floral, coconut, and mushroom to identify each compound. Tasting protocol for model

wine samples was discussed. Panelists were told to hold each sample in their mouth for 10

seconds before swallowing or expectorating the sample.

Panelists were given model wine without added flavor compounds at the beginning of the

third day. Panelists were instructed to taste the model wine so that they could differentiate

between model wine attributes/sensations and flavor in later evaluation sessions. Ethanol burn

and ethanol sweetness were discussed in order to aid panelists in overlooking these perceptions

during formal evaluation and concentrate on flavor perception. Familiarization with the four

flavor compounds continued. Panelists tasted four model wine solutions with single flavor

compounds. Once panelists felt they could comfortably identify each flavor in model wine, a

base wine sample (Carlo Rossi Chabli, 9.5% alcohol, Modesto, CA) was served as a reference.

Then, two base wine samples were served. Each base wine sample contained a single flavor

compound. Floral and coconut flavors were introduced in the base wine. Flavor identification

with descriptors (fruity, floral, coconut, and mushroom) was performed on the model and base

wine. The panel discussed flavor intensity, finish, and overall perception of samples.

The fourth training day began with the introduction of mushroom and fruity flavors in

base wine. Panelists were then given a 15-cm line scale marked with ten second intervals. High

and low intensity were represented by a vertical scale. Panelists were instructed to taste samples

of model wine with single flavor compounds as a group and record perceived finish at ten second

24

intervals that were verbally announced by the trainer. This exercise helped prepare panelists for

the computer time intensity evaluation they would perform in formal evaluations.

In-booth computer training began on the fifth day. Panelists evaluated Listerine strips

(Johnson and Johnson, New Brunswick, NJ) and cinnamon candies (Brachs, Round Lake, MN)

utilizing the Compusense (Compusense 5.2, Guelph, ON) time intensity program. A five minute

waiting period took place followed by evaluation of fruity, floral, coconut, and mushroom

flavors in model wine.

The final training session included a group discussion of a flavored model wine sample.

Topics including ethanol perception, flavor perception, and computer software usage were

discussed. Panelists repeated in-booth evaluation of fruity, floral, coconut, and mushroom

flavors in model wine with the time intensity program.

Trained Panel Formal Evaluation

Panelists performed a time intensity evaluation of a single flavor in model wine with

flavor compound mixtures of increasing complexity. Analysis began with model wines

containing one compound, followed by model wines with two compounds and then four

compounds. Panelists were given 2 minutes and 30 seconds to analyze each sample with two

minutes between each sample. Unsalted saltine crackers and milliQ water were provided for

palate cleansing between samples.

Two compound analysis included 6 different flavor solutions with two evaluations per

solution (i.e. evaluating floral flavor in a fruity and floral mix and then evaluating fruity flavor in

the same mixture). This led to twelve overall evaluations which, for clarity in later discussion,

25

are coded with the flavor being evaluated listed first:

1: Coconut and floral

2:Coconut and fruity

3: Coconut and mushroom

4: Floral and coconut

5:Floral and fruity

6:Floral and mushroom

7:Fruity and coconut

8:Fruity and floral

9:Fruity and mushroom

10:Mushroom and coconut

11:Mushroom and floral

12:Mushroom and fruity

Time intensity analysis of one compound, two compound, and four compound solutions was

repeated in base wine. Carlo Rossi Chablis (9.5% alcohol, Modesto, CA) was used as the base

wine.

Trained panelists performed a time intensity analysis of three commercially produced

Washington State Chardonnays. Time intensity parameters collected included time of maximum

intensity (Tmax), intensity maximum (Imax), duration (Tend), and area under the curve (AUC).

Data Analysis

Time intensity parameters including Tmax, Imax, Tend, and area were analyzed by XLStat

2011 (Addinsoft, Paris France) and Stata (StataCorp LP, College Station, Texas). Three-way

analysis of variance (ANOVA) with panelist, wine, and replicate main effects was performed

followed by Tukey‘s Honestly Significant Difference (HSD) test. Data were then normalized

following the Liu and MacFie method (1992). After normalization, principle component

analysis (PCA) was performed using mean time intensity parameters from normalized data.

26

Consumer Panel

Consumers (n=60) participated in a study to evaluate white wine finish. Consumers were

respondents to electronic and posted advertising. Demographic data indicated that the panel was

composed of 33 males and 26 females (1 panelist preferred not to answer) with 31 panelists age

21-29, 8 panelists age 30-39, 5 panelists age 40-49, 11panelists age 50-59, and 4 panelists age

60 and over (1 panelist preferred not to answer). Forty-six panelists were occasional white wine

drinkers (a few times per month), 8 panelists identified themselves as frequent white wine

drinkers (a few times per week), and 1 panelists identified themselves as a daily drinker of white

wine. Five panelists reported that they never drink white wine. Panelists reported drinking the

following white wine varieties: Riesling (51.7%), Chardonnay (50%), Sauvignon Blanc (31.7%),

Gewürtraminer (30%), Pinot Grigio (28.3%), and other (36.7%). The panel took place at the

WSU Sensory Evaluation Facility and consumers were given non-monetary incentives for their

participation. Compusense 5.2 (Guelph, ON) was utilized to perform sensory evaluation.

Consumer Evaluation

Panelists were asked to time how long they perceived wine finish and express this time in

seconds. To do this, panelists were instructed to sample the wine and then use the digital timer

(Fisher Scientific, Santa Clara, CA) provided to record the length of the wine after taste in

seconds. The experimental design was a complete balanced block with monadic presentation.

Panelists were given unsalted crackers, a cuspidor, and milliQ water. Panelists were instructed

to wait for thirty seconds in between samples and cleanse their palates with water and crackers.

Panelists were then asked to rate on a 7-point hedonic scale how much they liked the

wine finish with 1=dislike very much and 7 = like very much. Finally, panelists rated their

willingness to purchase each wine using a branched question format. Each panelist was first

27

asked if they would be willing to pay ten dollars for a bottle of the wine they had sampled. If the

response was positive, the dollar amount increased by two dollars (up to 14 dollars). If the

response was negative, the amount decreased by two dollars (down to 6 dollars).

Data Analysis

Data were analyzed by XLStat 2011 (Addinsoft, Paris France). Two-way analysis of

variance with panelist and wine main effects was performed followed by Tukey‘s HSD. Branch

willingness to purchase data was converted to numerical data that represented dollar ranges to

allow for analysis by an ordered probit model (Version 10, Stata Corp, College Station, TX).

Gas Chromatography Mass Spectrometry (GC-MS)

Methods were chosen based on a 2006 optimization performed by Carillo and others.

GC-MS was performed to analyze the wines utilized in the consumer panel. A standard curve

was made utilizing model wine with increasing levels of whiskey lactone (Sigma Aldrich).

Model wine was 9% ethanol, 0.6% D – (-)-fructose, DL-tartaric acid, adjusted to pH 3.0. The

standard curve was prepared by diluting a 10 mg/L whiskey lactone (Sigma Aldrich) stock

solution to concentrations of 0.1, 0.5, 1.0, 1.5, 2.0, and 3.0 mg/L whiskey lactone in model wine.

Samples were prepared by placing 5 mL of sample in 10 ml SPME vials. Sodium

chloride (J.T. Baker, Austin, TX) was added to achieve 30% (w/v) salt concentration. 1-

dodecanol was added as an internal standard by adding 0.5 µL of 250 mg/L 1-dodecanol solution

for a final concentration of 0.025 mg/L to samples of model wine with whiskey lactone or

Chardonnay. Samples were stirred for 5 minutes after the addition of the internal standard.

A divinylbenzene–carboxen–polydimethylsiloxane (DVB–CAR–PDMS) fiber was

utilized in combination with a CTC Combi-PaL autosampler (Zwingen, Switzerland) to extract

28

sample volatiles. Pre-incubation occurred for 10 minutes at 70° C. Extraction time was 1 hour

at 70° C without agitation.

An Agilent GC 6890N chromatograph with MS 5975 mass spectrometer was utilized to

analyze wine samples. A HP-5MS (30.0 m x 250 µm x 0.25 µm) column was used and splitless

injection (using helium carrier gas) was performed at 260°C with a 7 minute desorption time.

Temperature program began at 40° C with a hold of 3 minutes. After the initial 3 minute at 40°

C, temperature ramped at 7° C per minute to a maximum temperature of 230° C followed by a

hold time of 10 minutes, for a total program time of 40.14 minutes. The mass spectrometer

detector collected data using SCAN mode (33 m/z to 300m/z). The NIST mass spectra library

(version 2.0d) was used to confirm compound identities.

29

CHAPTER 4

RESULTS AND DISCUSSION

Trained Panel

Single Flavor Compound Analysis in Model Wine

Analysis of a variance (ANOVA) evaluation of solutions containing a single flavor

compound in model wine (Table 1) determined that there were no significant differences in time

to maximum intensity (Tmax) for flavor (coconut, fruity, floral, and mushroom) or replicate

evaluation. No significant interactions for panelist*flavor was found. However, a significant

panelist effect (p<0.0001) was observed which was not unexpected. Previous research has

indicated that panelists create characteristic time intensity curves which, while reproducible for

each panelist, often result in dissonance amongst panelists (Noble 1995, van Buuren 1992). In

the statistical model used in this study, panelists were treated as fixed effect. O‘Mahony states

that this is typical in trained panel evaluations, as the results apply only to the panel, not to the

population at large, since trained panelists are purposefully selected and trained and therefore

may not represent the entire population (O‘Mahony 1986).

Lack of significant differences in Tmax between flavors may be due to the super-threshold

concentrations of the individual flavor compounds. Super-threshold concentrations could cause

flavors to be perceived at saturation levels very quickly, leading to limited Tmax variation. This

would be consistent with the principle of Rmax from the Beidler Model of stimulus response.

Rmax is the value corresponding to maximal response to a stimulus (Meilgaard and others 2007).

It is possible that all flavors were presented at, or above, Rmax and therefore panelists perceived

Tmax similarly.

30

Panelist (p<0.0001) and flavor (p=0.003) significantly impacted the perceived maximum

intensity (Imax) in solutions containing a single flavor compound in model wine (Table 2). A

significant panelist*flavor interaction (p=0.016) indicated that variation among panelists

influenced the variation among flavors. Comparison with Tukey‘s HSD (Table 3) showed a

significant difference between fruity and mushroom flavors (p=0.002). Mean separation (Table

8) indicated that panelists perceived mushroom to have a higher mean Imax than fruity. For these

studies, concentrations of flavor compounds were selected to be similar in intensity. Significant

intensity differences in only one of six possible comparisons of flavor compound intensity

indicated that the selected concentrations were reasonably similar.

Significant differences in duration (Tend) (Table 4) were found for panelist and flavor

(p<0.0001) for single flavor compounds in model wine. There was significant panelist*flavor

interaction (p=0.033) indicating that variation between panelists caused some of the variation

between flavors. However, the strength of the significance for the flavor main effect indicated

that not all of the observed variation was a result of panelist interaction. Pairwise comparison

(Table 5) followed by mean separation (Table 8) showed that fruity flavor finished significantly

earlier than coconut (p<0.0001), floral (p=0.001), and mushroom (p<0.0001) flavors. Fruity

flavor finished 29.9 to 47.1 seconds earlier than the other flavors. This result supports the

hypothesis that fruity flavor would finish earlier than coconut and mushroom flavors. The

hypothesis that floral flavor would finish earlier than coconut and mushroom flavors is supported

numerically (second shortest mean finish time), but it is not supported through statistically

significant results. Mean Tend values show that flavors finish in the following order: fruity, floral,

coconut, and mushroom. This order is reflected in Figure 2 which shows the time intensity

curves from a representative panelist for each flavor.

31

Area under the curve (AUC) data from single flavor compounds in model wine (Table 6)

showed significant panelist (p<0.0001) and flavor (p=0.000) main effects. No significant

interactions were observed. Pairwise comparison (Table 7), followed by mean separation

(Table 8) showed that fruity flavor had a significantly smaller AUC than floral (p=0.023) and

mushroom (p=0.000).

PCA analysis of normalized single flavor compounds in model wine was also performed

(Figure 3). The PCA analysis indicated that variation in the coconut flavor was best described

the Tmax parameter. Variation in mushroom and floral flavors was best described by the

remaining parameters of Imax, Tend, and AUC. Additionally, with PC1 accounting for 71.6% of

variation, the majority of the variation observed among the flavors is a result of the relationship

of fruity and mushroom flavors. This reflects patterns seen in the standard data analysis, which

showed fruity flavor having significantly lower values for Imax, Tend, and AUC than mushroom

flavor.

The effect of data normalization is demonstrated in Figures 4, 5, 6, and 7. Normalization

allowed averaged curves evaluated by a standard ANOVA procedure to be compared to

mathematically altered (normalized) curves. Normalized curves took into account the behavior

of all the trained panelists, but unlike averaging, still allowed the characteristics of an

individual‘s evaluation to be represented (MacFie and Liu 1992).

Multiple Flavor Compound Analysis in Model Wine

Panelists evaluated solutions containing two flavor compounds in model wines. Panelists

were asked to evaluate a specific flavor of the two flavors presented. For the parameter Tmax

(Table 9), there was a significant panelist effect (p<0.0001), which then translated to a

32

significant panelist*flavor interaction (p=0.001). However, flavor and replicate were not

significant.

There were significant panelist (p<0.0001) and flavor (p<0.0001) effects for the

parameter Imax (Table 10), without any significant interactions. Pairwise comparison of the

flavor solutions indicated that the Imax of solution 12 (see p. 26 for code) was significantly larger

than solution 1, 3, and 9. Solution 11 was significantly larger than solution 9. This indicated

that maximum mushroom intensity (in the presence of fruity and floral flavors) was perceived

are more intense than fruity and coconut flavor in two compound evaluations. The Imax of

solution 4 was significantly larger than solutions 1 and 9, indicating that floral flavor in the

presence of coconut flavor had a greater maximum intensity than coconut flavor evaluation in the

presence of floral flavor and fruity flavor evaluation in the presence of mushroom flavor.

Mushroom flavor may have acted as a suppressor of fruity flavor in this instance. This may be

similar to suppression effects caused by Brettanomyces aroma compounds 4-ethylphenol and 4-

ethylguaiacol (Bramley and others 2008). In a 2008 study by Bramley and others, a base wine

without added Brettanomyces aroma compounds was perceived to have higher fruit aroma than

wines with added 4-ethylphenol and 4-ethylguaiacol (Bramley and others 2008).

Significant panelist (p<0.0001) and flavor (p=0.001) effects were seen for Tend evaluation

in two compound model wine solutions (Table 11). A significant panelist*flavor interaction

(p=0.003) was observed, indicating that panelist contributed to some of the variation seen among

flavors. Pairwise comparison showed solutions 2, 4, 11, and 12 as significantly larger than

solution 8. Solution 8 was had the smallest mean Tend of all 12 evaluations, indicating that fruity

(which also had the smallest mean Tend in single compound evaluations) may be further

suppressed by the presence of floral flavor.

33

Finally, area under the curve for two compounds in model wine was analyzed (Table 12).

Significant panelist (p<0.0001) and flavor (p<0.0001) were observed with no significant

replicate effect or interactions. Pairwise comparison shows that solution 12 area was

significantly larger than area for solutions 7, 8, and 9. Solution 2 area was significantly larger

than solution 8 area. Solution 2 and 11 areas were significantly larger than solution 9 area. This

indicates that fruity flavor perception may be suppressed by the presence of addition flavors,

particularly mushroom. However, it may also be seen (as described above) as a result of having

lower mean Tend values without significantly higher Imax values.

Contrasts of the four overall flavors (coconut, floral, fruity, and mushroom) were

performed. No significant differences were observed for the parameter Tmax. Mushroom and

floral flavors had significantly higher Imax values than coconut and fruity flavors. Fruity flavor

had a significantly shorter Tend and area under the curve than coconut, floral, and mushroom.

Principle component analysis was performed on the normalized data for two compounds

in model wine solutions (Figure 8). Solutions 3 and 6 were best described by the parameter

Tmax, while Imax described solution 7 and 8. Tend described solution 2. Area describes variation

for solution 4, 11, and 12. PC1 represented 64.7% of observed variation. The opposing

relationship of fruity perception and mushroom perception accounted for part of the PC1

variation, which is a similar outcome to the observations from the standard data analysis

Solutions with four flavor compounds in model wines were evaluated by the trained

panelists. For the parameter Tmax (Table 14), a significant panelist effect (p<0.0001) was

observed. No significant main effects were found for the parameter of Imax when trained

panelists evaluated four flavor compounds in model wine (Table 15). Significant panelist

(p<0.0001) and flavor (p=0.022) effects occurred for Tend (Table 16). Pairwise comparison

34

(Table 17) followed by mean separation (Table 20) indicated that floral flavor Tend was

significantly (p=0.017) longer than fruity flavor. Floral flavor had the longest Tend of all four

flavors. This result differs from the single compound flavor solutions, in which floral flavor

finished before coconut and mushroom flavors. These differences may be explained by effects

caused by the presence of the additional flavors in solution. For instance, flavor enhancement, or

the increase in perception of one flavor cause by the presence of other flavors (Meilgaard and

others 2007), could cause elongated perception of floral flavor not observed in single compound

model wine solutions.

The concept of flavor enhancement increasing floral flavor perception is further

supported by results for the parameter AUC for the four compound model wine solutions (Table

18). There were significant panelist (p<0.0001) and flavor (p=0.032) effects. Pairwise

comparison (Table 19) followed by mean separation (Table 20) showed that floral flavor had a

significantly larger (p=0.021) AUC than fruity flavor, and the largest mean AUC of all four

flavors. The large AUC for floral flavor indicated a larger percentage of flavor perception. As

with differences in Tend, enhancement effects caused by the presence of the additional flavors

could create significant differences between fruity and floral that did not exist when the flavors

were evaluated alone.

PCA of normalized data for four compounds in model wine (Figure 9) further explained

variation among flavors. Coconut and floral flavor variation were best described by the

parameters Imax, Tend, and Area. Mushroom flavor variation was best described by Tmax.

Coconut and floral flavor have similar characteristics, but variation in the data can be attributed

to the relationship of coconut and floral flavor to fruity flavor. The antagonist relationship

between floral and fruity flavor reflects differences observed in standard data analysis.

35

Single Flavor Compound Analysis in Base Wine

Base wine solutions with single flavor compounds were evaluated by the trained

panelists. With the exception of significant panelist effect (p=0.039) for Tmax (Table 21) and

significant panelist effect (p<0.0001) for Imax (Table 22), Tend (Table 23), and AUC (Table 24),

no significant differences were found among flavors for any of the time intensity parameters.

This could be the result of the panelists adjusting to the complexity of the base wine (compared

to the model wine solutions). Non-significant results could also be the result of flavor

suppression caused by flavor compounds already present in the base wine. Suppression occurs

when the presence of one or more flavor compounds lessens the perception of another flavor

compound (Meilgaard and others 2007). A time-intensity evaluation of suppression and

enhancement effects on flavor was performed in a 1993 study by Bonnans and Noble. In the

study the effect of sweet and sour tastes on fruit flavor was observed. Bonnans and Noble found

changes in fruity flavor maximum intensity and duration, both of which were perceived to be

higher in sweet and sour taste solutions (Bonnans and Noble 1993). Similar work observing

changes in time intensity parameters when flavors are compared has not been completed.

PCA analysis of normalized data from single compound in base wine (Figure 10)

evaluation further described the relationship of flavors and time-intensity parameters. Variation

was explained by the relationship between fruity and floral, as was seen in both standard and

normalized four compound in model wine evaluations. Variability in coconut flavor perception

could be attributed to Tend. Tmax and AUC described variation in floral flavor. Mushroom flavor

variation was best described by Imax.

36

Multiple Flavor Compound Analysis in Base Wine

Trained panelists evaluated two compounds in base wine. Aside from a significant

panelist effect (p<0.0001), no significant effects or interactions were observed for Tmax (Table

25). Significant panelist (p<0.0001) and flavor (p<0.0001) effects and significant panelist*flavor

(p=0.006) interactions were observed for Imax (Table 26). Pairwise comparison showed that

solution 4, 5, 6, 8, 10, 11, 12 had significantly higher Imax values than solution 1. This indicated

that coconut flavor intensity may have been suppressed by floral flavor. Solution 11 had a

significantly higher Imax value than solution 9.

Analysis of the parameter Tend in two compounds in base wine evaluations (Table 27)

showed a significantly panelist effect (p<0.0001) and a significant panelist*flavor interaction

(p=0.041). However, there was not a significant flavor effect. Area under the curve (Table 28)

analysis showed a significant panelist (p<0.0001) and flavor (p=0.001) effect with a significant

panelist*flavor interaction (p<0.0001). This indicated that variation among panelists contributed

to the variation observed in flavors. Pairwise comparison of solutions indicated that solutions 4,

10, and 11 had significantly larger area under the curve than solution 1. Coconut perception may

be masked or suppressed by floral flavor perception.

Contrasts were performed for overall coconut, floral, fruity, and mushroom flavors

(Table 29). No significant differences were observed between flavors for Tmax or Tend.

Significant differences were observed for the parameter Imax, with mushroom and floral flavors

having significantly higher mean Imax values than coconut flavor. Mushroom flavor had a

significantly higher Imax value than fruity flavor. Mushroom flavor also had significantly larger

area under the curve than coconut and fruity flavor.

37

Principle component analysis was performed on normalized data for two compounds in

base wine solutions (Figure 11). The parameter Tmax best described solutions 2 and 3.

Parameters Imax, Tend, and area best described solutions 4, 5, 6, 10, 11, and 12.

Four compounds in base wine were analyzed by trained panelists. Significant panelist

(p<0.0001) and flavor (p=0.023) effects were observed for the parameter Tmax (Table 30).

Pairwise comparison (Table 31) followed by mean separation (Table 37) showed significant

differences between the evaluation of coconut and fruity flavors (p=0.035) and coconut and

floral flavors (p=0.038). Suppression and/or enhancement may have caused significant

differences among flavors for Tmax that were not observed in single compound base wine

solutions.

A similar suppression/enhancement issue may have also occurred for the parameter Imax

(Table 32). Significant main effects for panelist (p<0.0001) and flavor (p=0.052) were

observed. A significant interaction for panelist*flavor (p=0.033) was also observed, indicating

that variability amongst panelist contributed to some of the variability observed between flavors

Pairwise comparison (Table 33) followed by mean separation (Table 37) indicated that floral

had a significantly higher mean Imax than coconut flavor (p=0.033). Floral flavor may be

enhanced by the flavor compounds creating the fruity, coconut, and mushroom flavors or other

flavor compounds that were preexisting in the base wine. Alternatively, coconut flavor may

have been suppressed by the presence of additional flavor compounds. This explanation would

also explain the significantly higher Tmax observed in coconut evaluations. Panelists may have

had to wait for the effects of other flavor compounds that were suppressing coconut to lessen

before being able to observe peak coconut flavor.

38

Significant differences were not observed for Tend in the four compounds in the base wine

(Table 34) with the exception of a significant panelist effect (p<0.0001). Panelists may not have

found differences due to suppression from pre-existing flavors in the base wine. Significant

differences for the main effects of panelist (p<0.0001) and flavor (p=0.008) for the parameter

area were observed (Table 35). Pairwise comparison (Table 36) followed by mean separation

(Table 37) showed significant differences (p=0.004) between floral and coconut evaluation.

Coconut AUC was 1679.8 as compared to floral area AUC which was 2587.05. Differences in

AUC between coconut and floral flavor further supported the explanation that coconut was

suppressed in the four compound solutions. Coconut‘s decreased AUC indicated that a

suppressed perception occurred for the panelists.

PCA analysis of normalized data for four compound solutions in base wine (Figure 12)

indicated that the parameters Imax, Tend, and AUC best characterized floral flavor variation.

Variation in the mushroom flavor was best explained by the parameter Tmax. The relationship

between floral and coconut explained part of the PC1 variation and reflected differences

observed in the standard data analysis.

Trained panelists evaluated three commercially available Chardonnays. Aside from a

significant panelist effect, panelists did not find any significant differences for Tmax (Table 38),

Imax (Table 39), Tend (Tabel 40), or AUC (Table 41). Of these parameters, Tend was of greatest

interest. Mean Tend values were 91.611, 96.056, and 110.389 for unoaked, medium oak, and high

oak samples, respectively. These results indicated that the data followed the hypothesized

numerical trend for the data, but it failed to yield statistically significant results.

39

Consumer Panel Evaluation of Commercial Chardonnay:

Consumers found significant differences (p=0.051) in the length of finish when

evaluating three commercially produced Washington State Chardonnays (Table 42). A

significant panelist effect (p<0.0001) was observed, but this can be accounted for by several

factors including physiological differences in taste perception among panelists and the

complexity of the testing procedure for untrained panelists. Pairwise comparison with Tukey‘s

HSD determined there was a significant difference (p=0.059) in the perception of finish between

Unoaked Chardonnay and high oak Chardonnay (Table 43). The mean separation of the three

Washington State Chardonnay wines (Table 44) indicated that Unoaked Chardonnay‘s mean

finish of 45.8 seconds was significantly shorter than the high oak Chardonnay with a mean finish

of 53 seconds. The medium oak Chardonnay had a mean finish length of 51.817 seconds and

was not significantly different from either the high or no oak Chardonnay samples. The increase

in finish length of the wines with increased oak exposure and significantly longer finish time of

the high oak Chardonnay wine supports the study‘s hypothesis. These differences may have

resulted from the addition of non-volatile and volatile components attained from barrel exposure

(Towey and Waterhouse 1996).

The consumer panel also evaluated how much they liked the aftertaste/finish of the

Washington State Chardonnay on a seven-point hedonic scale (Table 45). Significant

differences were found between the wines (p=0.059). Pairwise comparison with Tukey‘s HSD

of the three Washington State Chardonnay wines determined a significant difference (p=0.046)

between unoaked and high oak wines. Mean separation data (Table 46) showed that unoaked

Chardonnay had a higher mean score than high oak Chardonnay, indicating that the aftertaste of

the unoaked Chardonnay was preferred. These results support those found in a 2011 study

40

(Stump and others) which also found that consumers preferred unoaked Chardonnay

significantly more than high oak Chardonnay. Significant panelist effect was also observed

(p<0.0001).

Willingness to purchase response data were coded so that dollar value ranges

corresponded to a value (Table 47). Significant differences (p=0.036) in willingness to purchase

were found between the unoaked and high oak Chardonnay samples (Table 48). Panelists were

willing to pay higher dollar amounts for a bottle of unoaked Chardonnay than high oak

Chardonnay. Willingness to pay higher prices for unoaked wine was counterintuitive since

barrel aged wines tend to be more expensive and regarded as higher quality than unoaked wines

(Stump and others 2011). However, Stump and others found similar results, with consumer

panelists willing to pay $0.67 more per bottle for unoaked Chardonnay than highly oaked

Chardonnay. These results indicated that quality does not always positively influence consumer

preference or hedonic evaluation of wine.

Gas Chromatography – Mass Spectrometry (GC/MS):

MS library identification of the two oak lactone isomers was not able to distinguish

between the two compounds. Although oak lactone was recognized, the library classified both

peaks as the cis isomer. Thus, in order to identify and quantify both isomers, further

investigation would be necessary. Although definitive ion based identification was not

accomplished, it was assumed that the later eluting oak lactone was the cis isomer. In the

Chardonnay wines analyzed, the later eluting lactone displayed a much larger peak area than the

earlier eluting lactone, translating to a higher overall concentration. This is consistent with a

study by Towey and Waterhouse (1996) which examined variation within lots of barrel aged

Chardonnay and between American and French barrel aged Chardonnay. This study found that

41

the cis isomer at a consistently higher concentration than the trans isomer. Additionally, a study

by Perez-Coello and others (1997) presented chromatographic data that displayed the trans

isomer as having a smaller peak area and eluting earlier than the cis isomer (Perez-Coello and

others 1997).

Unoaked Chardonnay wines did not contain cis or trans oak lactone. This is expected

since the isomers originate from exposure to oak (Brown and others, 2006). High oak

Chardonnay samples had higher concentrations of cis and trans lactone than medium oak

Chardonnay. For cis-oak lactone, high oak Chardonnay samples had a mean concentration of

0.358 ug/L, compared to a mean concentration of 0.289 ug/L in medium oak Chardonnay

samples. For trans-oak lactone, high oak Chardonnay samples had a mean concentration of

0.120 ug/L, compared to a mean concentration of 0.0757 ug/L in medium oak Chardonnay

samples. Ratios of cis/trans oak lactone were 3.82 and 2.98 for medium and high oak

Chardonnay respectively. These values were within the 1.31-7.35 cis/trans oak lactone ratio

range found 2Towey and Waterhouse (1996).

42

Table 1. Degrees of freedom and F values from analysis of variance (ANOVA) of time to maximum intensity (Tmax) for single

compounds (floral, fruity, mushroom and coconut) in model wine samples as analyzed by a trained panel (n=10).

Source DF Sum of squares Mean squares F Pr > F

Panelist 9 4864.00 540.44 9.27 < 0.0001

Flavor 3 415.75 138.58 2.38 0.085

Replicate 1 5.00 5.00 0.09 0.771

Panelist*Flavor 27 1270.00 47.04 0.81 0.718

43

Table 2. Degrees of freedom and F values from analysis of variance (ANOVA) table of maximum intensity (Imax) for single



Panelist 9 13452.80 1494.76 12.81 < 0.0001

Flavor 3 1935.25 645.08 5.53 0.003

Replicate 1 92.45 92.45 0.79 0.379

Panelist*Flavor 27 6668.50 246.98 2.12 0.016

44

Table 3. Pairwise comparison of maximum intensity (Imax) evaluations for single compounds in model wine samples analyzed by a

trained panel (n=10) using Tukey‘s HSD (p≤0.05).

Contrast Difference Standardized difference Critical value Pr > Diff

Mushroom vs Fruity 13.45 3.94 2.68 0.002

Mushroom vs Coconut 8.60 2.52 2.68 0.073

Mushroom vs Floral 5.05 1.48 2.68 0.460

Floral vs Fruity 8.40 2.46 2.68 0.083

Floral vs Coconut 3.55 1.04 2.68 0.728

Coconut vs Fruity 4.85 1.42 2.68 0.495

Tukey's d critical value:

3.79

45

Table 4. Degrees of freedom and F values from analysis of variance (ANOVA) table of duration (Tend) for single compounds (floral,

fruity, mushroom and coconut) in model wine samples as analyzed by a trained panel (n=10).


Panelist 9 59994.30 6666.03 11.88 < 0.0001

Flavor 3 24673.45 8224.48 14.66 < 0.0001

Replicate 1 162.45 162.45 0.29 0.594

Panelist*Flavor 27 28806.80 1066.92 1.90 0.033

46

Table 5. Pairwise comparison of duration (Tend) evaluationsfor single compounds in model wine samples analyzed by a trained panel

(n=10) using Tukey‘s HSD, p≤0.05.

Contrast Difference

Standardized

difference Critical value Pr > Diff

Mushroom vs Fruity 47.05 6.28 2.68 < 0.0001



Coconut vs Fruity 37.10 4.95 2.68 < 0.0001

Coconut vs Floral 6.75 0.90 2.68 0.804

Floral vs Fruity 30.350 4.052 2.68 0.001


3.79

47

Table 6. Degrees of freedom and F values from analysis of variance (ANOVA) table of area under the curve (AUC) for single



Panelist 9 66332278.63 7370253.18 6.47 < 0.0001

Flavor 3 27451590.91 9150530.30 8.03 0.000

Replicate 1 31660.90 31660.90 0.03 0.869

Panelist*Flavor 27 34231621.43 1267837.83 1.11 0.374

48

Table 7. Pairwise comparison of area under the curve (AUC) evaluations for single compounds in model wine samples analyzed by a

trained panel (n=10) using Tukey‘s HSD, p≤0.05.

Contrast Difference

Standardized





Floral vs Fruity 1014.13 3.00 2.68 0.023




3.79

49

Table 8. Mean Separation of Maximum Intensity (Imax), Duration (Tend), and area under the curve (AUC) for single compounds in

model wine samples analyzed by the trained panel (n=10) using Tukey‘s HSD. Different letters indicate significant differences

between the flavors (p≤0.05).

Flavor Mushroom Floral Coconut Fruity

Imax LS means 67.20a 62.15ab 58.60ab 53.75b

Tend LS means 137.10a 120.40a 127.15a 90.05b

Area LS means 4052.80a 3427.60a 3227.95ab 2413.48b

50

Figure 2. Representative trained panelist time intensity evaluation of single flavors in model wine

0

10

20

30

40

50

60

70

0 50 100 150 200

Inte

nsi

ty

Time (sec)

Floral

Coconut

Mushroom

Fruity

51

Figure 3. Pearson‘s principle component (PC) analysis factor loadings and factor scores from normalized time intensity data of single

compounds in model wine evaluation by a trained panel (n=10). The plot illustrates the PC space for each compound and four time-

intensity parameters, Imax (maximum intensity), AUC (area under the curve), Tend (duration) and Tmax (time to maximum

intensity).

Coconut

Floral

Fruity

Mushroom

Tmax

Imax

Tend

AUC

-2

0

2

-5 -3 -1 1 3 5

F2 (

26.4

%)

F1 (71.6 %)

52

Figure 4. Comparison of standard data and normalized data from a representative trained panelist‘s time intensity evaluation of the

coconut flavor in single compound model wine

0

10

20

30

40

50

60

70

0 50 100 150 200 250

Inte

nsi

ty

Time (sec)

Standard Data

Normalized Data

53


floral flavor in single compound model wine

0

10

20

30

40

50

60

70

0 50 100 150 200 250

Inte

nsi

ty

Time (sec)

Standard Data

Normalized Data

54


fruity flavor in single compound model wine

0

10

20

30

40

50

60

0 20 40 60 80 100 120 140 160

Inte

nsi

ty

Time (sec)

Standard Data

Normalized Data

55


mushroom flavor in single compound model wine

0

10

20

30

40

50

60

70

80

0 50 100 150 200 250

Inte

nsi

ty

Time (sec)

Standard Data

Normalized Data

56

Table 9. Degrees of freedom and F values from analysis of variance (ANOVA) of time to maximum intensity (Tmax) for two

compounds (all possible combinations of floral, fruity, mushroom and coconut) in model wine samples as analyzed by a trained panel

(n=10).


Panelist 9 22098.87 2455.43 38.10 < 0.0001

Flavor 11 1260.17 114.56 1.78 0.064

Replicate 1 223.99 223.99 3.33 0.070

Panelist*Flavor 99 11722.04 118.40 1.83 0.001

57

Table 10. Degrees of freedom and F values from analysis of variance (ANOVA) of maximum intensity (Imax) for two compounds (all

possible combinations of floral, fruity, mushroom and coconut) in model wine samples as analyzed by a trained panel (n=10).


Panelist 9 54284.37 6031.60 32.72 < 0.0001

Flavor 11 8082.51 734.77 3.99 < 0.0001

Replicate 1 324.34 324.34 1.76 0.187

Panelist*Flavor 99 23167.28 234.01 1.27 0.106

58

Table 11. Degrees of freedom and F values from analysis of variance (ANOVA) of duration (Tend) for two compounds (all possible

combinations of floral, fruity, mushroom and coconut) in model wine samples as analyzed by a trained panel (n=10).


Panelist 9 284953.52 31661.50 44.95 < 0.0001

Flavor 11 24423.11 2220.28 3.15 0.001

Replicate 1 182.00 182.00 0.26 0.612

Panelist*Flavor 99 117580.43 1187.68 1.69 0.003

59

Table 12. Degrees of freedom and F values from analysis of variance (ANOVA) of area under the curve (AUC) for two compounds

(all possible combinations of floral, fruity, mushroom and coconut) in model wine samples as analyzed by a trained panel (n=10).


Panelist 9 354265928.17 39362880.91 32.24 < 0.0001

Flavor 11 56307312.69 5118846.61 4.19 < 0.0001

Replicate 1 27767.26 27767.26 0.02 0.880

Panelist*Flavor 99 116624965.35 1178029.95 0.97 0.571

60

Table 13. Mean separation of contrasts of the four flavors (floral, fruity, mushroom and coconut) analyzed in two compound model

wine solutions evaluated by a trained panel (n=10), p≤0.05.

Flavor Coconut Fruity Mushroom Floral

Imax LS means 47.48b 47.60b 58.05a 57.02a

Tend LS means 106.08a 89.87b 110.45a 108.85a

AUC LS Means 2657.61a 2057.97b 3152.93a 2890.19a

61

Figure 8. Pearson‘s principle component analysis factor loadings and factor scores from normalized time intensity data of two


intensity parameters, Imax (maximum intensity), AUC (area under the curve), Tend (duration) and Tmax (time to maximum intensity).

The compound listed first/capitalized was the compound panelists were told to evaluate.

COCONUT and floral

COCONUT and fruity

COCONUT and

mushroom

FLORAL and coconut

FLORAL and fruity

FLORAL and mushroom

FRUITY and coconut

FRUITY and floral

FRUITY and mushroom

MUSHROOM and

coconut

MUSHROOM and floral

MUSHROOM and fruity

Tmax

Imax

Tend

AUC

-3

-1

1

3

-5 -3 -1 1 3 5

F2 (

28.1

%)

F1 (64.7 %)

62

Table 14. Degrees of freedom and F values from analysis of variance (ANOVA) of time to maximum intensity (Tmax) for four

compounds (floral, fruity, mushroom and coconut) in model wine samples as analyzed by a trained panel (n=10)


Panelist 9 10152.01 1128.00 8.53 < 0.0001

Flavor 3 633.94 211.31 1.60 0.205

Replicate 1 465.61 465.61 3.52 0.068

Panelist*Flavor 27 3845.44 142.42 1.08 0.409

63

Table 15. Degrees of freedom and F values from analysis of variance (ANOVA) of maximum intensity (Imax) for four compounds

(floral, fruity, mushroom and coconut) in model wine samples as analyzed by a trained panel (n=10)


Panelist 9 20547.11 2283.01 17.30 < 0.0001

Flavor 3 848.84 282.95 2.14 0.110

Replicate 1 409.51 409.51 3.10 0.086

Panelist*Flavor 27 5690.04 210.74 1.60 0.089

64

Table 16. Degrees of freedom and F values from analysis of variance (ANOVA) of duration (Tend) for four compounds (floral, fruity,

mushroom and coconut) in model wine samples as analyzed by a trained panel (n=10)


Panelist 9 85751.56 9527.95 11.88 < 0.0001

Flavor 3 8659.64 2886.55 3.60 0.022

Replicate 1 667.01 667.01 0.83 0.367

Panelist*Flavor 27 15645.49 579.46 0.72 0.811

65

Table 17. Pairwise comparison of duration (Tend) evaluations for four compounds in model wine samples analyzed by a trained panel

(n=10) using Tukey‘s HSD, p≤0.05.

Contrast Difference

Standardized


Floral vs Fruity 27.90 3.12 2.68 0.017

Floral vs Mushroom 21.95 2.45 2.68 0.084



Coconut vs Mushroom 6.55 0.73 2.68 0.884



3.79

66

Table 18. Degrees of freedom and F values from analysis of variance (ANOVA) of area under the curve (AUC) for four compounds

(floral, fruity, mushroom and coconut) in model wine samples as analyzed by a trained panel (n=10)


Panelist 9 142705105.90 15856122.88 16.55 < 0.0001

Flavor 3 9369087.93 3123029.31 3.26 0.032

Replicate 1 48782.50 48782.50 0.05 0.823

Panelist*Flavor 27 22241635.66 823764.28 0.86 0.655

67

Table 19. Pairwise comparison of area under the curve (AUC) evaluations for four compounds in model wine samples analyzed by a


Contrast Difference

Standardized


Floral vs Fruity 939.70 3.04 2.68 0.021







3.79

68

Table 20. Mean separation of duration (Tend) and area under the curve (AUC) for four compounds in model wine samples analyzed by

the trained panel (n=10) using Tukey‘s HSD. Different letters indicate significant differences between the flavors (p≤0.05).

Flavor Floral Coconut Mushroom Fruity

Tend LS means 116.50a 101.10ab 94.55ab 88.60b

AUC LS means 3022.00a 2708.23ab 2484.75ab 2082.30b

69

Figure 9. Pearson‘s principle component analysis factor loadings and factor scores from normalized time intensity data of four



Coconut

Floral

Fruity

Mushroom

Tmax

Imax

Tend

AUC

-1.5

0.1

-5 -3.4 -1.8 -0.2 1.4 3 4.6

F2 (

14.3

0 %

)

F1 (83.93 %)

70

Table 21. Degrees of freedom and F values from analysis of variance (ANOVA) of time to maximum intensity (Tmax) for single

compounds (floral, fruity, mushroom and coconut) in base wine samples as analyzed by a trained panel (n=10).


Panelist 9 7464.45 829.38 2.25 0.039

Flavor 3 807.30 269.10 0.73 0.541

Replicate 1 781.25 781.25 2.12 0.154

Panelist*Flavor 27 8262.45 306.02 0.83 0.691

71

Table 22. Degrees of freedom and F values from analysis of variance (ANOVA) of maximum intensity (Imax) for single compounds

(floral, fruity, mushroom and coconut) in base wine samples as analyzed by a trained panel (n=10).


Panelist 9 11461.51 1273.50 18.94 < 0.0001

Flavor 3 403.64 134.55 2.00 0.130

Replicate 1 15.31 15.31 0.23 0.636

Panelist*Flavor 27 2591.24 95.97 1.43 0.152

72

Table 23. Degrees of freedom and F values from analysis of variance (ANOVA) of duration (Tend) for single compounds (floral,

fruity, mushroom and coconut) in base wine samples as analyzed by a trained panel (n=10).


Panelist 9 60661.95 6740.22 6.39 < 0.0001

Flavor 3 1994.85 664.95 0.63 0.600

Replicate 1 793.80 793.80 0.75 0.391

Panelist*Flavor 27 24088.15 892.15 0.85 0.673

73

Table 24. Degrees of freedom and F values from analysis of variance (ANOVA) of area under the curve (AUC) for single compounds



Panelist 9 42497959.14 4721995.46 7.27 < 0.0001

Flavor 3 2093624.36 697874.79 1.07 0.371

Replicate 1 219137.11 219137.11 0.34 0.565

Panelist*Flavor 27 19027502.89 704722.33 1.09 0.401

74

Figure 10. Pearson‘s principle component analysis factor loadings and factor scores from normalized time intensity data of single

compounds in base wine evaluation by a trained panel (n=10). The plot illustrates the PC space for each compound and four time-


Coconut

Floral Fruity

Mushroom

Tmax

Imax

Tend

AUC

-2.5

-1.5

-0.5

0.5

1.5

2.5

-4 -3 -2 -1 0 1 2 3 4

F2 (

33.0

4 %

)

F1 (56.11 %)

75

Table 25. Degrees of freedom and F values from analysis of variance (ANOVA) of time to maximum intensity (Tmax) for two

compounds (all possible combinations of floral, fruity, mushroom and coconut) in base wine samples as analyzed by a trained panel

(n=10).


Panelist 9 14297.35 1588.59 33.63 < 0.0001

Flavor 11 860.63 78.24 1.66 0.092

Replicate 1 16.02 16.02 0.34 0.561

Panelist*Flavor 99 5677.95 57.35 1.21 0.155

76

Table 26. Degrees of freedom and F values from analysis of variance (ANOVA) of maximum intensity (Imax) for two compounds (all

possible combinations of floral, fruity, mushroom and coconut) in base wine samples as analyzed by a trained panel (n=10).


Panelist 9 76558.17 8506.46 67.14 < 0.0001

Flavor 11 5499.41 499.95 3.95 < 0.0001

Replicate 1 45.94 45.94 0.36 0.548

Panelist*Flavor 99 20363.88 205.70 1.62 0.006

77

Table 27. Degrees of freedom and F values from analysis of variance (ANOVA) of duration (Tend) for two compounds (all possible

combinations of floral, fruity, mushroom and coconut) in base wine samples as analyzed by a trained panel (n=10).


Panelist 9 209299.68 23255.52 30.31 < 0.0001

Flavor 11 11041.98 1003.82 1.31 0.228

Replicate 1 117.60 117.60 0.15 0.696

Panelist*Flavor 99 106011.52 1070.82 1.40 0.041

78

Table 28. Degrees of freedom and F values from analysis of variance (ANOVA) of area under the curve (AUC) for two compounds

(all possible combinations of floral, fruity, mushroom and coconut) in base wine samples as analyzed by a trained panel (n=10).


Panelist 9 268953467.42 29883718.60 40.62 < 0.0001

Flavor 11 25934898.86 2357718.08 3.21 0.001

Replicate 1 58734.46 58734.46 0.08 0.778

Panelist*Flavor 99 162128675.92 1637663.39 2.23 < 0.0001

79

Table 29. Mean separation of contrasts of the four flavors (floral, fruity, mushroom and coconut) analyzed in two compound base

wine solutions evaluated by a trained panel (n=10), p≤0.05.

Flavor Coconut Fruity Mushroom Floral

Imax LS means 46.10c 48.05bc 54.77a 53.53ab

AUC LS Means 2128.62b 2058.77b 2662.02a 2555.29ab

80

Figure 11. Pearson‘s principle component analysis factor loadings and factor scores from normalized time intensity data of two



COCONUT and floral COCONUT and fruity

COCONUT and

mushroom

FLORAL and coconut FLORAL and fruity

FLORAL and mushroom

FRUITY and coconut

FRUITY and floral

FRUITY and mushroom

MUSHROOM and

coconut

MUSHROOM and floral

MUSHROOM and fruity Tmax

Imax

Tend AUC

-2

0

2

-6.5 -4.5 -2.5 -0.5 1.5 3.5 5.5

F2 (

18.8

0 %

)

F1 (74.71 %)

81

Table 30. Degrees of freedom and F values from analysis of variance (ANOVA) of time to maximum intensity (Tmax) for four

compounds (floral, fruity, mushroom and coconut) in base wine samples as analyzed by a trained panel (n=10).


Panelist 9 8546.70 949.63 16.30 < 0.0001

Flavor 3 622.65 207.55 3.56 0.023

Replicate 1 115.20 115.20 1.98 0.168

Panelist*Flavor 27 2297.60 85.10 1.46 0.137

82

Table 31. Pairwise comparison of time to maximum intensity (Tmax) evaluations for four compounds in base wine samples analyzed

by a trained panel (n=10) using Tukey‘s HSD, p≤0.05.

Contrast Difference

Standardized



Coconut vs Floral 6.75 2.80 2.68 0.038




Floral vs Fruity 0.10 0.04 2.68 1.000


3.79

83

Table 32. Degrees of freedom and F values from analysis of variance (ANOVA) of maximum intensity (Imax) for four compounds



Panelist 9 30118.11 3346.46 31.63 < 0.0001

Flavor 3 892.44 297.48 2.81 0.052

Replicate 1 201.61 201.61 1.91 0.175

Panelist*Flavor 27 5432.44 201.20 1.90 0.033

84

Table 33. Pairwise comparison of maximum intensity (Imax) evaluations for four compounds in base wine samples analyzed by a


Contrast Difference

Standardized



Floral vs Fruity 5.60 1.72 2.68 0.326




Fruity vs Coconut 3.70 1.14 2.68 0.669


3.79

85

Table 34. Degrees of freedom and F values from analysis of variance (ANOVA) of duration (Tend) for four compounds (floral, fruity,

mushroom and coconut) in base wine samples as analyzed by a trained panel (n=10).


Panelist 9 107742.76 11971.42 12.77 < 0.0001

Flavor 3 3483.14 1161.05 1.24 0.309

Replicate 1 525.31 525.31 0.56 0.459

Panelist*Flavor 27 18594.49 688.69 0.74 0.798

86

Table 35. Degrees of freedom and F values from analysis of variance (ANOVA) of area under the curve (AUC) for four compounds



Panelist 9 102600318.84 11400035.43 18.74 < 0.0001

Flavor 3 8258197.41 2752732.47 4.52 0.008

Replicate 1 23375.70 23375.70 0.04 0.846

Panelist*Flavor 27 26959861.87 998513.40 1.64 0.077

87

Table 36. Pairwise comparison of area under the curve (AUC) evaluations for four compounds in base wine samples analyzed by a


Contrast Difference

Standardized



Floral vs Fruity 471.60 1.91 2.68 0.240




Fruity vs Coconut 435.65 1.77 2.68 0.305


3.79

88

Table 37. Mean separation of time to maximum intensity (Tmax), maximum intensity (Imax), and area under the curve (AUC) for four

compounds in base wine samples analyzed by the trained panel (n=10) using Tukey‘s HSD. Different letters indicate significant

differences between the flavors (p≤0.05).

Flavor Floral Coconut Mushroom Fruity

Tmax LS means 14.15b 20.90a 17.00ab 14.05b

Imax LS means 51.35a 42.05b 47.40ab 45.75ab

AUC LS means 2587.05a 1679.80b 2166.48ab 2115.45ab

89

Figure 12. Pearson‘s principle component analysis factor loadings and factor scores from normalized time intensity data of four



Coconut Floral

Fruity

Mushroom Tmax

Imax Tend

AUC

-2

0

2

-5 -3 -1 1 3 5

F2 (

18.0

3 %

)

F1 (77.43 %)

90

Table 38. Degrees of freedom and F values from analysis of variance (ANOVA) table of time to maximum intensity (Tmax) for three

commercially produced Washington State Chardonnay samples analyzed by a trained panel (n=10), p≤0.05


Panelist 8 2428.67 303.58 22.21 < 0.0001

Wine 2 46.33 23.17 1.70 0.203

Replicate 1 48.17 48.17 3.52 0.072

Panelist*Wine 16 166.33 10.40 0.76 0.712

91

Table 39. Degrees of freedom and F values from analysis of variance (ANOVA) table of maximum intensity (Imax) for three



Panelist 8 16322.67 2040.33 18.48 < 0.0001

Wine 2 427.00 213.50 1.93 0.165

Replicate 1 60.17 60.17 0.55 0.467

Panelist*Wine 16 1342.67 83.92 0.76 0.712

92

Table 40. Degrees of freedom and F values from analysis of variance (ANOVA) table of duration (Tend) for three commercially

produced Washington State Chardonnay samples analyzed by a trained panel (n=10), p≤0.05


Panelist 8 75210.48 9401.31 14.36 < 0.0001

Wine 2 3466.82 1733.41 2.65 0.090

Replicate 1 937.50 937.50 1.43 0.242

Panelist*Wine 16 5096.52 318.53 0.49 0.932

93

Table 41. Degrees of freedom and F values from analysis of variance (ANOVA) table of area under the curve (AUC) for three



Panelist 8 110842634.58 13855329.32 15.47 < 0.0001

Wine 2 1317948.11 658974.06 0.74 0.489

Replicate 1 3110400.00 3110400.00 3.47 0.074

Panelist*Wine 16 4114535.89 257158.49 0.29 0.994

94

Table 42. Degrees of freedom and F values from analysis of variance (ANOVA) of consumer panel (n=60) evaluation of Washington

State Chardonnay finish with timed finish perception, p≤0.05.


Panelist 59 120104.06 2035.66 6.96 < 0.0001

Wine 2 1788.81 894.41 3.06 0.051

95

Table 43. Pairwise comparison of finish time of Washington State Chardonnays analyzed by a consumer panel (n=60) using Tukey‘s

HSD, p≤0.05.

Contrast Difference

Standardized


High oak vs unoaked Chardonnay 7.20 2.31 2.37 0.059

High oak vs medium 1.18 0.38 2.37 0.924

Medium oak vs unoaked

Chardonnay 6.02 1.93 2.37 0.136


3.36

96

Table 44. Mean separation of finish time (seconds) of Washington State Chardonnays analyzed by a consumer panel (n=60) using

Tukey‘s HSD. Different letters represent a significant difference at *(p<0.1), ** (p<0.05) or *** (p<0.001)

Wine High Oak Chardonnay Medium Oak Unoaked Chardonnay

LS means 53.0a* 51.8ab 45.8b*

97

Table 45. Degrees of freedom and F values from analysis of variance (ANOVA) of consumer panel (n=60) evaluation of Washington

State Chardonnay finish with a 7-point hedonic scale, p≤0.05


Panelist 59 225.53 3.82 2.31 < 0.0001

Wine 2 9.64 4.82 2.91 0.059

98

Table 46. Mean separation of 7-point hedonic scale rating of Washington State Chardonnays analyzed by a consumer panel (n=60)

using Tukey‘s HSD, *(p<0.1), ** (p<0.05) or *** (p<0.001).

Wine Unoaked Chardonnay Horse Heaven Hill Reserve Chardonnay

LS means 4.8a* 4.483ab 4.2b*

99

Table 47. Coding of branched willingness to purchase responses determined by the consumers (n=60) for the evaluation of the

Washington State Chardonnay.

Code Dollar Range that the consumer would pay/bottled of a

particular wine

1 $0-$5.99

2 $6.00-$7.99

3 $8.00-$9.99

4 $10.00-$11.99

5 $12.00-$13.99

6 $14.00+

100

Table 48. Pairwise comparison of wines for branch willingness to purchase response to Washington State Chardonnays by consumer

panel (n=60). Data were analyzed using an ordered logit series as described in Table 53. (p≤0.05).

Comparison p-value

Unoaked Chardonnay vs medium oak Chardonnay 0.316

Unoaked Chardonnay vs high oak Chardonnay 0.036

Medium Oak Chardonnay vs high oak Chardonnay 0.111

101

Figure 13. Standard curve for cis-oak lactone in model wine (9% EtOH, 0.6% fructose, pH 3.0) with three replicates.

y = 1.5832x - 0.069

R² = 0.8547

0

0.5

1

1.5

2

2.5

3

0 0.2 0.4 0.6 0.8 1 1.2 1.4

lact

on

e a

rea

/in

tern

al

sta

nd

ard

are

a

mg/L cis oak lactone

102

Figure 14. Standard curve for trans-oak lactone in model wine (9% EtOH, 0.6% fructose, pH 3.0) with three replicates.

y = 1.4852x - 0.0381

R² = 0.9174

0

0.5

1

1.5

2

2.5

3

3.5

4

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2

lact

on

e a

rea

/in

tern

al

sta

nd

ard

are

a

mg/L trans oak lactone

103

CHAPTER 6

CONCLUSIONS AND FUTURE RESEARCH

Single compound analysis in model wines supported the hypothesis that fruity flavor

finish earlier than coconut and mushroom flavors. Numerically, floral flavor finished earlier

than coconut and mushroom flavors, but this observation was not significant. Therefore, single

compound analysis in model wine did not support the hypothesis that floral flavor would finish

earlier than coconut and mushroom flavors.

Multiple compound analysis in model wines resulted in more variation in results. This

was potentially due to interaction between flavor compounds presented together in solution. For

two compound model wine solutions, mushroom had a suppressive effect on Imax determination.

Analysis of Tend data supported the hypothesis that fruity flavor would finish earlier than coconut

and mushroom flavors. Fruity flavor perception in the presence of floral flavor showed the

significantly earlier finish than other solutions. Contrasting the four flavors overall in two

compound model wine solutions also supported the hypothesized early finish of fruity flavor.

Principle component analysis (PCA) of normalized data further supported differences between

fruity and mushroom flavors. Four compound model wine model wine results showed that fruity

flavor had the shortest perceived finish, supporting hypothesized results for fruity flavor.

Hypothesized early finish of floral flavor was not supported by two compound or four compound

model wine data.

Analysis of base wine showed different trends than those seen in model wine analysis.

Significant results were not observed for finish length in single or multiple compound analysis.

However, base wine analysis of Imax and AUC parameters indicated that certain flavors were

influential as suppressors of intensity and perception. Two compound base wine analysis

showed that mushroom and floral flavors were perceived as significantly more intense than fruity

104

and coconut flavors. Mushroom had significantly greater AUC than other flavors. Four

compound base wine analysis supported this observation, as floral flavor was perceived to be

significantly more intense than coconut flavor and have significantly greater AUC.

Consumer panel analysis of commercially available Washington State Chardonnays

indicated that panelists were able to determine finish differences between unoaked and high oak

Chardonnay finish. High oak Chardonnay was perceived to have significantly longer finish than

unoaked Chardonnay. Additionally, consumers expressed significant differences in acceptance

for finish. Consumers preferred the finish of unoaked wine over high oak wine. This acceptance

was reflected in willingness to purchase analysis of the Chardonnay samples. Consumers were

willing to pay significantly larger dollar values for unoaked Chardonnay than high oak

Chardonnay. This indicated that barrel aging, although associated with quality, does not dictate

consumer preference or willingness to purchase.

Chemical analysis with SPME-GC-MS quantified the differences in cis- and trans- oak

lactone among unoaked, medium, and high oak Chardonnay samples. Quantification of oak

lactone showed that unoaked wine did not contain cis- or trans- oak lactone. Concentrations of

cis- and trans- oak lactone were highest in high oak Chardonnay samples.

Future research on wine finish could include a number of different types of time intensity

studies. Cataloging of additional flavor compound in white model wine could be performed,

creating a profile of flavor finish for white wine. Similar work could be performed in red model

wine systems. This work could then be expanded to catalog flavor finish utilizing different

varieties of white and red base wines.

Evaluation of finish of barrel aged compounds could also be performed. Consumer

panelists would be presented model solutions containing individual compounds associated with

105

barrel aging and rate the flavor compound finish on a hedonic scale. This would indicate which

compounds resulting from barrel aging contribute to a favorable finish. These results could be

paired with analytical analysis of different oak/barrel sources. Winemakers could then determine

what barrels will yield the most favorable compounds and therefore improve the quality of wine

finish.

106

REFERENCES

Aldave L, Almy J, Cabezudo M, Caceres I, Gonzalez-Raurich M, Salvador M. 1993. The shelf-

life of young white wines. Dev Food Sci 33:923-44.

Amerine MA, Roessler EB. 1983. Wines: their sensory evaluation. New York: W.H. Freeman

and Company. 230 p.

Arnold RA, Noble AC. 1978. Bitterness and astringency of grape seed phenolics in a model

wine solution. Am J Enol Vitic 29(3): 150-2.

Arnold RA, Noble AC, Singleton VL. 1980. Bitterness and astringency of phenolic fractions in

wine. J Agri Food Chem 28(3):675-8.

Bardi L, Crivelli C, Marzona M. 1998. Esterase activity and release of ethyl esters of medium-

chain fatty acids by saccharomyces cerevisiae during anaerobic growth. Can J Micro

44(12):1171-6.

Bertrand A. 1986. Utilisation de la chromatographie en phase gazeuse pour le dosage des

constiuants volatils du vin. Univ Bordeaux.

Bonnans S, Noble A. 1993. Effect of sweetener type and of sweetener and acid levels on

temporal perception of sweetness, sourness and fruitiness. Chem Senses 18(3):273-83.

Bramley B, Curtin C, Cowey G, Holdstock M, Coulter A, Kennedy E, Travis B, Mueller S,

Lockshin L, Godden P, Francis L. 2008. Wine style alters the sensory impact of ‗Brett‘ flavour

compounds in red wines. Aust Wine Res Inst. Available from (www.winepreferences.com).

Brown RC, Sefton MA, Taylor DK, Elsey GM. 2006. An odour detection threshold

determination of all four possible stereoisomers of oak lactone in a white and red wine. Aus J

Grape Wine Res 12(2):115-8.

Burdock GA. 2009. Fenaroli's Handbook of Flavor Ingredients. 6th ed. Boca Raton, FL: CRC

Press. 2159 p.

Carrillo JD, Garrido-Lopez A, Tena MT. 2006. Determination of volatile oak compounds in

wine by headspace solid-phase microextraction and gas chromatography-mass spectrometry. J

Chrom A 1102:25-36.

Clarke RJ, Bakker J. 2004. Wine flavour chemistry. Oxford, UK: Blackwell Pub. 324p.

Coombe B. 1992. Research on development and ripening of the grape berry. Amer J Enol Vitic

43(1):101-10.

DeMan JM. 1999. Principles of food chemistry. 3rd ed. Gaithersburg, MD: Aspen Publishers.

520p.

Ebeler SE. 2001. Analytical chemistry: Unlocking the secrets of wine flavor. Food Rev Int

17(1):45-64.

107

Etaio I, Elortondo FJP, Albisu M, Gaston E, Ojeda M, Schlich P. 2008. Effect of winemaking

process and addition of white grapes on sensory and physiochemical characteristics of young red

wines. Aus J Grape Wine Res 14(3):211-22.

Ezquerro O, Tena MT. 2005. Determination of odour-causing volatile organic compounds in

cork stoppers by multiple headspace solid-phase microextraction. J Chromat A 1068(2):201-8.

Francis IL, Sefton MA, Williams PJ. 1992. A study by sensory descriptive analysis of the effects

of oak origin, seasoning, and heating on the aromas of oak model wine extracts. Am J Enol Vitic

43(1): 23-30.

Gallart M, Francioli S, Viu-Marco A, López-Tamames E, Buxaderas S. 1997. Determination of

free fatty acids and their ethyl esters in musts and wines. J Chromat A 776(2):283-91.

Genovese A, Gambuti A, Piombino P, Moio L. 2007. Sensory properties and aroma compounds

of sweet fiano wine. Food Chem 103(4):1228-36.

Guth H. 1997. Quantitation and sensory studies of character impact odorants of different white

wine varieties. J Agric Food Chem 45(8):3027-32.

Ishikawa T, Noble A. 1995. Temporal perception of astringency and sweetness in red wine. Food

Qual Pref 6(1):27-33.

Jackson RS. 2000. Wine science: principles, practice, perception. 2nd

ed. San Diego, CA:

Academic Press. 648 p.

Jackson RS. 2002. Wine tasting: a professional handbook. 1st ed. San Diego, CA: Academic

Press. 295 p.

Kadim D, Mannheim C. 1999. Kinetics of phenolic extraction during aging of model wine

solution and white wine in oak barrels. Amer J Enol Vit 50(1):33-9.

La Guerche S, Dauphin B, Pons M, Blancard D, Darriet P. 2006. Characterization of some

mushroom and earthy off-odors microbially induced by the development of rot on grapes. J

Agric Food Chem 54(24):9193-200.

Lawless HT ,Heymann H. 1998. Sensory evaluation of food: principles and practices. 1st ed.

New York: Chapman and Hall. 848 p.

Lee SJ, Noble AC. 2003. Characterization of odor-active compounds in Californian chardonnay

wines using GC-olfactometry and GC-mass spectrometry. J Agric Food Chem 51(27):8036-44.

Lund ST, Bohlmann J. 2006. The molecular basis for wine grape quality-A volatile subject.

Science 311:804-5.

MacFie HJH, Liu YH. 1992. Developments in the analysis of time-intensity curves. Food Tech

92-7.

Marais J. 1983. Terpenes in the aroma of grapes and wines: a review. S Afr J Enol Vitic 4(2):49-

60.

108

Marais J, van Wyk C. 1986. Effect of grape maturity and juice treatments on terpene

concentrations and wine quality of vitis vinifera l. cv. weisser riesling and bukettraube1. S Afr J

Enol Vitic 7(1):26-35.

Meilgaard M, Civille GV, Carr BT. 2007. 4th ed. Sensory evaluation techniques. Boca Raton,

FL: CRC Press. 448 p.

Mialon V, Ebeler S. 1997. Time-intensity measurement of matrix effects on retronasal aroma

perception. J Sens Stud 12(4):303-16.

Noble AC. 1995. Application of time-intensity procedures for the evaluation of taste and

mouthfeel. Amer J Enol Vit 46(1):128-33.

Noble AC, Arnold RA, Buechsenstein J, Leach EJ, Schmidt J, Stern PM. 1987. Modification of a

standardized system of wine aroma terminology. Amer J Enol Vitic 38(2):143-6.

Nordstom K. 1964. Studies on the formation of volatile esters in fermentation with brewer‘s

yeast. Svensk Kem Tidskr 76:510-43.

O'Mahony M. 1986. Sensory evaluation of food: statistical methods and procedures. CRC Press.

487 p.

Ovejero-Lopez I, Bro R, Bredie W. 2005. Univariate and multivariate modelling of flavour

release in chewing gum using time-intensity: a comparison of data analytical methods. Food

Qual Pref 16(4):327-43.

Perez-Coello M, Sanz J ,Cabezudo M. 1997. Analysis of volatile components of oak wood by

solvent extraction and direct thermal desorption-gas chromatography-mass spectrometry. J

Chrom A 778(1-2):427-34.

Piggott JR, Hunter EA, Margomenou L. 2000. Comparison of methods of analysis of time-

intensity data: application to scotch malt whisky. Food Chem 71(3):319-26.

Ramey D, Bertrand A, Ough C, Singleton V, Sanders E. 1986. Effects of skin contact

temperature on Chardonnay must and wine composition. Amer J Enol Vit 37(2):99-106.

Robichaud JL, Noble AC. 1990. Astringency and bitterness of selected phenolics in wine. J Sci

Food Agric 53(3):343-53.

Ross CF, Hinken C, Weller K. 2007. Efficacy of palate cleansers for reduction of astringency

carryover during repeated ingestions of red wine. J Sens Studies 22(3):293-312.

Servili M, De Stefano G, Piacquadio P, Sciancalepore V. 2000. A novel method for removing

phenols from grape must. Amer J Enol Vit 51(4):357-61.

Schreier P, Drawert F, Junker A. 1977. Gaschromatographische bestimmung der inhaltsstoffe

von garungagetranke part X. Chem Mikrobiol Technol Lebensm 5:45-52.

109

Stump C, McCluskey JJ, Ross CF. 2011. Consumer preferences and willingness to pay for oak

attributes in washington chardonnays. [PhD Dissertation]. Pullman, WA: Washington State

Univ.

Towey JP, Waterhouse AL. 1996. The extraction of volatile compounds from french and

american oak barrels in chardonnay during three successive vintages. Am J Enol Vitic 47(2):163-

72.

Towey JP, Waterhouse AL. 1996. Barrel-to-barrel variation of volatile oak extractives in barrel-

fermented chardonnay. Am J Enol Vitic 47(1):17-20. 2

Valentová H, Skrovánková S, Panovská Z, Pokorn J. 2002. Time-intensity studies of astringent

taste. Food Chem 78(1):29-37.

Van Buuren S. 1992. Analyzing time-intensity reponses in sensory evaluation. Food Tech

46(2):101-4.

Washington Wine Commision: state facts [Internet]. Seattle, WA: Washington State Wine

Commission; c2007 [Accessed 2011 Feb 27]. Available from:

http://washingtonwine.org/washington-wine/state-facts.php

Waterhouse AL, Towey JP. 1994. Oak lactone isomer ratio distinguishes between wine

fermented in American and French oak barrels. J Agric Food Chem 42(9):1971-4.

Wilker KL, Gallander JF. 1988. Comparison of seyval blanc wine aged in barrels and stainless

steel tanks with oak chips. Amer J Enol Vit 39(1):38-43.

Wilson B, Strauss CR, Williams P. 1986. The distribution of free and glycosidically-bound

monoterpenes among skin, juice, and pulp fractions of some white grape varieties. Am J Enol

Vitic 37(2):107-11.

Zamora MC, Goldner MC, Galmarini MV. 2006. Sourness-sweetness interactions in different

media: white wine, ethanol and water. J Sens Studies 21(6):601-11.

Documents

PERCEPTION OF FINISH IN WHITE WINE By EMILY GOODSTEIN · mushroom acting as a suppressor of other flavors. Normalization of model wine time intensity analysis showed that data variation