Flavour '81: 3rd Weurman Symposium Proceedings of the International Conference, Munich April 28–30, 1981 9783110844061, 9783110084412

Citation preview

Flavour '81

3rd Weurman Symposium

Flavour '81 3rd Weurman Symposium Proceedings of the International Conference, Munich April 28-30,1981

Editor Peter Schreier

W DE

G

Walter de Gruyter • Berlin • New York 1981

Editor Peter Schreier, Dr. rer. nat. Lehrstuhl für Lebensmittelchemie der Universität Würzburg A m Hubland D - 8 7 0 0 Würzburg

CIP-Kurztitelaufnahme

der Deutschen

Bibliothek

[Flavour eighty-one] Flavour'81: proceedings of the internat. conference, Munich, April 28-30,1981/ 3rd Weurman Symposium. Ed. Peter Schreier. - Berlin; New York: de Gruyter, 1981. ISBN 3-11-00844-1 NE: Schreier, Peter [Hrsg.]; Weurman Symposium

Library of Congress Cataloging in Publication Data Weurman Symposium (3rd: 1981: Munich, Germany) Flavour'81. Bibliography: p. Includes indexes. 1. Flavour-Congresses. 2. Food-Analysis-Congresses. I. Schreier, Peter, 1942. II.Title. TX511.W46 1981 664 81-12518 ISBN 3-11-008441-4 AACR2

Copyright © 1981 by Walter de Gruyter&Co., Berlin 30 All rights reserved, including those of translation into foreign languages. No part of this book may be reproduced in any form - by photoprint, microfilm or any other means nor transmitted nor translated into a machine language without written permission from the publisher. Printing: Karl Gerike, Berlin. - Binding: Dieter Mikolai, Berlin - Printed in Germany.

PREFACE It is six years ago since - on Dr. W e u r m a n i n i t i a t i v e - a group of active scientists met for the first time in the Netherlands to describe and discuss the growing points and needs in flavour research. After the unexpected death of Dr. Weurman, all the participants agreed that they would, in future, connect this type of meeting with the name of this remarkable man, in recognition of his work carried out in the field of flavour research. The 3rd Weurman Symposium FLAVOUR *81 took place in Munich, April 28-30, 1981, and was organized by the Department of Food Chemistry at the University of Wiirzburg. Its purpose was to bring together especially active workers from diverse fields of flavour research, both from university and industry, to review representative aspects, hear and discuss new work, stimulate speculation and particularly, cross-fertilize between the different approaches and disciplines involved. This book contains the contributions presented as lectures or posters covering the following main topics: sensory methodology r application of sensory methods, instrumental analysis, formation of flavour, application and technology, and molecular aspects of flavour. Thus, an up-to-date compendium of almost all facets of flavour research could be published. Many people and authorities contributed to the success of the symposium. The Editor wishes to thank all who attended and participated in the Conference, and especially the authors of papers, short communications and posters. Furthermore, the Editor would like to express his sincerest thanks to the members of the Organizing Committee for their help in preparing this meeting, and for their assistance during the Conference, serving as chairpersons or in other functions, namely Prof. Dr

J. Adda, Dr. R. Emberger, Dr. D.G. Land, Dr. H. Maarse, Prof. Dr. H.E. Nursten, Prof. Dr. R.M. Pangborn and Prof. Dr. J. Solms. The Editor is particularly grateful to the members of his staff who acted as registrators, operators, technicians etc., and who helped to create a friendly working atmosphere. The Editor is also grateful to the Munich Technical University whose guests we were privileged to be. Finally, the Publisher's guidance and assistance is greatly appreciated.

P. Schreier

ACKNOWLEDGEMENTS

For economic support received to accomplish this symposium the Organizing Committee wants to express its gratitude to the following authorities and organizations DANI-Analysentechnik ERBA SCIENCE GmbH FINNIGAN GmbH HAARMANN & REIMER GmbH HAMILTON GmbH HEWLETT-PACKARD G m b H INFOCHROMA KRATOS GmbH PERKIN-ELMER & CO GmbH PHILIPS GmbH REICHELT Chemietechnik SIEMENS KG SILESIA KG VACUUM GENERATORS GmbH VARIAN GmbH Verband der Deutschen Essenzenindustrie e.V. WGA Analysentechnik. Without these contributions, the organization of the conference would not have been possible.

CONTENTS

PREFACE ACKNOWLEDGEMENTS

V VII

SECTION I SENSORY METHODOLOGY A critical review of threshold, intensity and descriptive analysis in flavor research R.M. Pangborn

3

Finding the word for it - methods and uses of descriptive sensory analysis J.R. Piggott and P.R. Canaway

33

Magnitude scales of taste and smell intensity J. Herrmann

47

Suprathreshold odour intensity assessment: individual variation in scaling H. Tuorila

53

Comparison of structured and unstructured category scales in the evaluation of sourness intensity A.C. Noble and J.O. Schmidt

63

Assessment of sensory discriminability using sensory difference tests J.E.R. Frijters

69

X

Evaluating tasters' performance in the profiling of foods and beverages A.A. Williams and C.R. Baines

83

Sensory evaluation in a "natural environment tl E.P. Köster

93

SECTION II APPLICATION OF SENSORY M E T H O D S Perception and analysis: a perspeetive view of attempts to find causal relations between sensory and objective data sets J.J. Powers

103

Some properties of odoriferous molecules R. Teranishi, R.G. Buttery and D.G. Guadagni

133

Predictive value of sensory and analytical data for distilled beverages P. Jounela-Eriksson

145

Physico-chemical studies on flavour-active compounds J.W. Gramshaw and D.R. Williams

165

The contribution of some volatile« to the sensory quality of apple and orange juice odour P. Diirr and U. Schobinger

179

The identification of the legume-flavour in raw peanuts K.H. Fischer and W. Grosch

195

Studies on quality characterization of meat-like flavourings M. Rothe, M. Specht a n d V. Ehrhardt

203

Improvements of the Seoville method for the pungency determination of black pepper L.J. van Gemert, L.M. Nijssen and H. Maarse

211

XI Flavor and texture of gels: an interlaboratory

study

B. Lundgren

217

A search for a nervous code for odor quality combining analytical and electrophysiological methods: a model of a biotest R. Selzer and N. Christoph

227

SECTION III INSTRUMENTAL ANALYSIS Recent developments in high resolution gas chromatography W. Jennings

233

The horizons of identification and analysis with mass spectrometry M.C. ten Noever de Brauw

253

Microanalysis of volatile compounds in biological materials in small quantities H. Sugisawa and T. Hirose

287

Comparison of adsorbents in head space sampling J. Schaefer

301

Quick flow-optimization of capillary columns W. Giinther, M. Meiertoberens and F. Schlegelmilch

315

Automated procedure for headspace analysis by glass capillary gas chromatography U. Ott and R. Liardon

323

Computer assisted identification in routine gas chromatographic analysis of essential oils L. Huber and H. Obbens

339

Application of HPLC for the separation of flavour compounds K.H. Kubeczka

345

A convenient method for determining volatile sulphur compounds in beer 0. Leppanen, J. Denslow, T. Koivisto and P. Ronkainen

361

XII Comparison of two isolation procedures for aroma compounds of dill R. Huopalahti, H. Kallio, P. Karppa and R. Linko

369

Analysis of natural and artificial coconut flavouring in beverages R. Eberhardt, H. Woidich and W. Pfannhauser

377

Isolation and properties of bitter-sensitive proteins via affinity chromatography I.L. Gatfield

385

SECTION IV FORMATION OF FLAVOUR Formation of flavour components from proline and hydroxyproline with glucose and maltose and their importance to food flavour R. Tressl, K.G. Griinewald and B. Helak

397

Shigematsu variation of the Maillard reaction D. de Rijke, J.M. van Dort and H. Boelens

417

Model experiments about the formation of volatile carbonyl compounds from fatty acid hydroperoxides W. Grosch, P. Schieberle and G. Laskawy

433

Influence of curing on the formation of tobacco flavour C.R. Enzell

449

Factors influencing flavour formation in fruits N. Paillard

479

Formation of "green-grassy"-notes in disrupted plant tissues: characterization of the tomato enzyme systems P. Schreier and G. Lorenz

495

Possibilities of the biotechnological production of aroma substances by plant tissue cultures F. Drawert and R. Berger

509

XIII Role of the gaseous e n v i r o n m e n t o n v o l a t i l e c o m p o u n d p r o d u c t i o n by fruit cell suspensions c u l t u r e d in vitro C. A m b i d a n d J. F a l l o t M i c r o b i a l f o r m a t i o n of E.W. Seitz

529 flavours

539

A r o m a - p r o d u c i n g fungi: Influence of strain specifity a n d culture c o n d i t i o n s o n a r o m a p r o d u c t i o n H.P. H a n s s e n and E. Sprecher

547

C h e d d a r c h e e s e flavour - its f o r m a t i o n in the l i g h t of new a n a l y t i c a l results D. L a m p a r s k y and I. K l i m e s

557

F l a v o u r c o n s t i t u e n t s of m a l t T . L . P e p p a r d , S.A. H a l s e y and D.R.J. L a w s

579

C h a r a c t e r i z a t i o n and c o m p a r i s o n of flavor v o l a t i l e s in m e a t b y - p r o d u c t s M.J. Greenberg

599

S u n l i g h t flavours in C h a m p a g n e w i n e s N. Charpentier and A. Maujean

609

SECTION V APPLICATION AND

TECHNOLOGY

A s p e c t s of the d e v e l o p m e n t of industrial flavor m a t e r i a l s R. E m b e r g e r P r o b l e m s of the industrial q u a l i t y control of

619 flavours

W. Bruhn

635

C h e m i c a l - t e c h n o l o g i c a l a s p e c t s for c o n c e n t r a t i o n of plant aromas F. D r a w e r t , P. S c h r e i e r , S. B h i w a p u r k a r a n d I. H e i n d z e

649

C r a i g d i s t r i b u t i o n in the a n a l y s i s of food flavours W . P i c k e n h a g e n , A . F u r r e r a n d B. M a u r e r

665

XIV Use of synthetic polymeric adsorbents for processing and recovering essential citrus-fruit oils F. Tateo

671

Aroma compounds which contribute to the difference in flavour between pasteurized milk and UHT milk H.T. Badings, J.J.G. van der Pol and R. Neeter

683

Inclusion complexes of potato starch with flavor compounds R. Wyler and J. Solms

693

The influence of food components on the volatility of diacetyl D.G. Land and J. Reynolds

701

Interactions of volatile flavor compounds with caffeine, chlorogenic acid and naringin B.M. King and J. Solms

707

Stability of flavouring substances in food models related to milk B.A. Gubler

717

Some remarks on the proposal for an EEC-directive on flavourings R. Grundel and H.E. Muermann

729

SECTION VI MOLECULAR ASPECTS OF FLAVOUR Structural requirements for sweet and bitter taste H.D. Belitz, W. Chen, H. Jugel, W. Stempfl, R. Treleano and H. Wieser

741

Bifunctional unit concept in flavour chemistry G. Ohloff

757

AUTHOR INDEX

771

SUBJECT INDEX

773

SECTION

I

SENSORY METHODOLOGY

A CRITICAL REVIEW OF THRESHOLD, INTENSITY AND DESCRIPTIVE ANALYSES IN FLAVOR RESEARCH

R. M. Pangborn Food Science and Technology, University of California, Davis, 95616, USA

Introduction

Psychophysics, the study of the relation between stimulus variables and sensory variables (1), has gradually extended from the psychologist's domain into the food science laboratory. Elemente der Psychophysik

From Fechner's classic,

(2), to the "New Psychophysics" (3, 4), the

field has focused on both theory and methodology (5, 6).

The theoreti-

cal behaviorist as well as the pragmatic technologist wish to understand the mechanism by which man perceives as well as how best to measure the phenomena.

Flavor has been one of the most complex, and hence most

elusive, of the sensory experiences. Technically, flavor is the attribute of a material which stimulates the sense organs that are grouped together at the entrance of the alimentary and respiratory tracts.

The sensory experience called flavor is a

mingled, but unitary experience which includes sensations of taste, smell, temperature, pain, pressure, and other cutaneous sensations.

Odor

or aroma, in contrast, is restricted to stimulation of receptors in the nasal passages and the olfactory epithelium.

By definition, flavor and

aroma, like pain and vision, are human sensations, rather than attributes of the stimulus.

Nonetheless, to adequately define and measure flavor

requires an appreciation of chemistry, physiology, and human behavior. Table 1 attempts to delineate three separate but integrated approaches to the study of flavor.

The importance of a knowledge of the physicochemical

properties of the flavorant, which is covered in detail by other authors in this proceedings, and of the transduction mechanism cannot be

© 1981 by W a l t e r d e G r u y t e r &. C o , B e r l i n • N e w Y o r k F l a v o u r '81

4

TABLE 1 Chemical, Physiological, and Behavioral Approaches to Study of Flavor Phenomenon Studied I. Physicochemical of Stimulus

Limitations

Properties

MW, MP, BP, Volatility Functional groups Stereochemistry Molecular vibration Partition coefficients, etc.

In vitro studies

II. Neural encoding, transduction Membrane solubility Membrane puncture Electrostatic transfer Spatiotemporal, Patterning etc.

Experimental animals Excised nerves or cells

III. Perception A.

Analytical

Tests

1. Sensitivity: Thresholds Detection Recognition Difference Terminal

Usually empirical Wide variability Responses dependent upon: Experiential factors Psychophysical procedures

2. Quantitative: Scaling Ordinal (Ranking) Interval (Category) Ratio (Magnitude Estimation) 3. Qualitative: Descriptive "Flavor Profile" "Dilution Profile" "Texture Profile" "Quantitative Descriptive Analysis" B.

Consumer T e s t s 3

Context dependent

1. Acceptance: Accept/reject what is available 2. Preference: Select one over another 3. Hedonic: Degree of like/dislike a

N o t validly measured in laboratory with trained judges.

5 overemphasized.

However, the information so derived must be validated

with human experimentation in order to progress beyond a narrow, in vitro status.

The analytical measures used most extensively in the study of

human perception of flavors and aromas have been threshold, scaling, and descriptive procedures, for quantification of sensitivity, perceived intensity, and quality, respectively.

Recent literature (7, 8) has called

attention to the "Consumer in the Laboratory Syndrome," the unfortunate misuse of hedonic testing of trained sensory personnel.

Whereas accept-

ance, preference, and degree of liking of products are very necessary to product development, these measures are inappropriate among analytical judges who are not representative of the consumers for whom the products are intended.

A.

Thresholds

Psychophysical^, a threshold can be defined as the minimum concentration of a stimulus that can be detected (absolute threshold), discriminated (just-noticeable-difference) or recognized (recognition threshold), usually 50% of the time.

In general, detection thresholds are lower than

recognition thresholds, but this relationship is valid only if the difficulty of the behavioral tests to measure both are equivalent. As exemplified by Allison and Katz in 1919 (9) and by Zwaardemaker in 1926 (10), chemists have long been interested in using thresholds for comparison of the relative strength of compounds.

Unfortunately, the

odor threshold values compiled by Zwaardemaker are in error by a factor of 100, as noted by Jones in 1953 (11).

The more recent computer-derived

listing assembled by Fazzalari (12), while useful, are cumbersome to decipher, due to cryptic computer notation and to the wide variety of concentration units in which the threshold values are expressed.

An illus-

tration of the information contained in the latter publication is given for hexanal in Table 2 where all concentrations have been converted to ppb.

The accuracy of the literature values reported by Fazzalari have

not been verified, as far as the present author knows.

One wonders, for

example, about the invariance of the threshold value of 4.5 ppb reported

6

TABLE 2 Examples of Odor and Taste Thresholds Reported for Hexanal Purity 3

Measurement/Author

Threshold ppb

ODOR DETECTION IN AIR Guadagni (1963)

GC

4.5

ODOR DETECTION IN WATER Guadagni (1963) Wick (1966) Flath (1967) Guadagni (1972) Ericksson (1976)

Chem.

?

GC GC GC

4.5 30.0 5.0 4.5 0.19

ODOR RECOGNITION IN WATER

1

Land (1968) Guadagni (1963)

GC

400 4.5

TASTE DETECTION IN WATER Lindsay (1969) Borovikova (1971)

GC ?

10 0.2

TASTE DETECTION IN PARAFFIN

1 1

Lea Meijboom (1964)

a

300 150

Chemically pure, gas chromatically pure, or not specified.

Source:

Fazzalari (12)

for hexanal odor detection in air, odor detection in water, and odor recognition in water attributed to Guadagni from two distinct publications, 1963 and 1972.

Note also the extreme variation between the two

entries for odor recognition in water (4.5 to 400 ppb) and for taste detection in water (0.2 to 10 ppb).

One of the lowest odor threshold

reported is that of 4 parts of thiamin dithia-8-oxaoctane) per 10^

3

(l-methylbicyclo[3.3.0]-2,4-

parts of water (13).

7

Table 2 is also included herein to underscore the specificity required in reporting and using threshold data, e.g., the type of threshold measured (detection, difference, recognition), the modality tested (taste, odor, pain), the medium of dispersion (air, water, ethanol, oil, paraffin), and the purity of the compound tested.

A serious omission from the Zwaarde-

maker and Fazzalari listings is lack of reference to the psychophysical or methodological variables which profoundly influence the validity and reliability of threshold measurement.

These include the number of test

subjects, the degree of experience of the subjects with the stimulus and with the test procedure, the nature of instruction given subjects, the specific test procedure employed, verification with replicate testing, and the manner in which the resultant data were handled statistically. A variety of single sample, paired, and triangle presentations have been used to determine detection, difference, and recognition thresholds.

An

excellent critical comparison of 14 test methods and directions for calculation of threshold values have been presented by Brown et al_. (14).

The

methods could be classified as variations of one of the following procedures: 1.

The Method of Limits (Method of Least Noticeable Difference, Method of

Minimal Changes, Method of Serial Exploration)

The detection threshold

is determined by approaching and receding from the standard stimulus by short concentration steps.

The threshold is that step where the response

shifts from one category to another.

In tests where subjects receive a

dilution series in ascending order of concentration, the experimenter must reckon with the "error of anticipation."

Conversely, when presented

with dilutions in descending order, subjects may exhibit the "error of habituation," the tendency to continue to report false positives.

These

"constant errors" are averaged out by randomizing the order of presentation of the concentration.

Calculation of difference thresholds by the

method of limits could be achieved with signal detection procedures. When presented with a blank or wvth a stimulus, the responses are categorized into one of four groups; hit, miss, false alarm, or correct rejection (15).

The major disadvantage of signal detection methods in food

research is the requirement for a large (ca. 200) number of measurements per judge.

A short-cut method, R-index, requiring approximately

8 20 judgments per subject, utilizes the respondant's degree of certainty of the detection (16, 17). 2.

The Method of Average Error (Method of Adjustment, Ad-Libiturn

Mixing.)

This difference threshold technique allows the subject to adjust

the concentration of the comparison stimulus to apparent equality with a standard (18). technique.

The mix is analyzed by an appropriate physical or chemical

This procedure is simple, requires little preparation and

serving, and subjects generally find the task interesting.

However, the

method is restricted to homogeneous, readily miscible materials for which there exists a rapid physical or chemical assay with a sensitivity equal to or greater than the sensory sensitivity for that ingredient.

The

procedure is more applicable for liquids and semiliquids assessed by mouth than for testing of aroma of materials dispersed in air or in 1iquid. 3.

The Frequency Method.

This matching procedure is used to establish

the frequency distribution of perception of a specific additive.

Each

comparison stimulus is tested against the standard stimulus an equal number of times, and the relative frequency is plotted against concentration. The 50% threshold (or 50% chance threshold) is located by interpolation or by statistical treatment, e.g., regression analysis.

This matching

procedure is reliable only when the stimulus increases in a single sensory dimension with increasing concentration, and when there is no enhancing, masking, adaptation, or other cross-interference between comparison stimuli.

An example of the inappropriate use of this technique is

illustrated by Larson-Powers and Pangborn (19) for comparison of the relative sweetness of sucrose with cyclamate where there was mutual synergism, and for saccharin where increasing concentrations became bitter at a faster rate than they became sweeter. With all test procedures, experimenters must anticipate extensive individual variability of response (14, 20), necessitating large numbers of subjects (a minimum of 25) and repeated testing.

Since subject performance

improves with practice, repeated measures are essential.

Brown et al.

(14) concluded that four replicates were necessary to establish an

9

approximate threshold using difference testing.

Table 3 illustrates the

reduction in taste thresholds for eight subjects from the first through the sixth measurement conducted within a two-week period (21). An appropriate range of stimulus concentrations must be included to bracket the individual sensitivities.

For example, out of 44 subjects

who were tested with the flavor of purified dimethyl sulfide added to beer, 37 had thresholds between 12 and 87 ug/L, while seven required retesting at a higher range (14).

Meilgaard and Reid (22) bracketed

their 16 subjects' thresholds for ethanol, dimethyl sulphide, diacetyl, and isoamyl acetate added to beer.

For each compound the most sensitive

10% of the subjects showed a threshold approximately 20 times lower than that of the least sensitive 10%.

Teranishi (23) noted that odor thresh-

olds for pyrazine compounds varied as much as 10® fold.

Using benzalde-

hyde dispersed in air, Drake et al_. (24) reported that a concentration approximately 100 times greater was required by the least sensitive of their 21 subjects than by the most sensitive to give a perceived odor intensity of "moderately strong." An innovative and potentially very useful equation, based on the ratios of the logarithms of reciprocals of flavor or odor thresholds in aqueous and oil media to octonal/water partition coefficients, has been postulated by Garner (25).

Based, unfortunately, upon literature-derived

thresholds (measured in diverse manners in different media by subjects differing widely in sensitivity), this 1ipophilicity approach invites verification under controlled experimentation. Bartoshuk (26) itemized factors which could modify taste threshold values, such as confusion of taste terms, taste distortions from compounds in saliva or from bad teeth, the amount and frequency of oral rinsing, and the fact that the taste thresholds vary inversely with the size of the tongue area stimulated.

This would invalidate the so-called "drop"

methods because the tongue area cannot be well-controlled, there are variations in the way subjects hold their tongues and in the condition on the surface of the tongue which affect the spread of the stimulus resulting in artifacts of the method rather than true taste differences (27).

10 TABLE 3 Influence of Practice on Individual Taste Recognition Thresholds SUCROSE Tst

SUBJECTS

CITRIC ACID Tst

6th

6th

NaCl

CAFFEINE

Tst

6th

1st

6th

34

30

1 6

.20

Molarity X 10" 3 1

36

22

1 6

.10

2

24

8

0 6

.04

13

3

0 6

.08

3

36

8

2 1

.01

34

5

1 6

.40

4

18

10

1 1

.10

20

3

1 1

.60

5

12

6

0 6

.01

20

7

0 6

.40

6

18

2

0 1

.03

20

5

1 1

.30

7

18

6

2 1

.04

7

3

3 1

.60

8

18

4

1 1

.03

20

9

1 6

.60

MEAN

22

8

1 2

.05

21

8

1 4

.40

Source:

Pangborn

(21)

With odor testing, there are persistent problems of background odors, adaptation, drying out of the nasal mucosa due to continuous and the potential ceptors.

interference by stimulation of trigeminal

sniffing, or pain re-

When odor samples are delivered via an air stream, the tactile

sensations produced by the pressure can modify thresholds, a condition which has discredited the Elsberg blast injection technique

(28).

As described previously (29), thresholds have been misused in at least three ways: 1.

To express relative sensory intensity.

potency are not synonymous.

Threshold concentration and

If subjects detect one compound at a concen-

tration 150 times lower than they detect another compound, it is incorrect to conclude that the first is 150 times stronger than the second, because the ratio is valid only at threshold concentration.

If the two

compounds increased in intensity as a function of concentration at different rates, the intensity ratio of 150/1 would not hold.

Therefore,

it must be emphasized that thresholds are of restricted value, because

11

they are but one p o i n t on a dynamic c o n c e n t r a t i o n continuum.

As pointed

out by Land ( 3 0 ) , i t i s f a l l a c i o u s to assume t h a t the perceived of t a s t e or odor substances w i l l tration.

The h y p o t h e t i c a l

compounds with i d e n t i c a l

intensity

i n c r e a s e at the same r a t e with concen-

curves depicted i n F i g u r e 1 i l l u s t r a t e

thresholds

(A, B) can i n c r e a s e i n

intensity

d i f f e r e n t i a l l y , being q u i t e d i f f e r e n t at h i g h e r c o n c e n t r a t i o n s the c o n c e n t r a t i o n at which they occur i n the product under C o n v e r s e l y , compounds with d i v e r s e t h r e s h o l d s t e n s i t i e s at h i g h e r 2.

that

(possibly

investigation).

(B, D) could have equal

in-

concentrations.

To determine procedural or e x p e r i e n t i a l

F i g u r e 1 a l s o can i l l u s t r a t e s i t u a t i o n s may be a l t e r e d by s u b j e c t v a r i a b l e s D f o r nonsmokers) or t r a i n i n g experienced s u b j e c t s ) .

e f f e c t s on s e n s i t i v i t y .

in which i n d i v i d u a l

thresholds

( e . g . , curve B f o r smokers v s .

curve

( e . g . , curve B f o r naive v s . curve D f o r

T h r e s h o l d s d i f f e r , but perceived i n t e n s i t i e s

h i g h e r c o n c e n t r a t i o n s do not (or c o n v e r s e l y f o r curves A and B).

at

An

e x c e l l e n t example of c o n f l i c t i n g t a s t e c o n c l u s i o n s were presented by Bartoshuk (26, 31) wherein a p a t i e n t ' s t h r e s h o l d s to t a s t e compounds returned to normal a f t e r r a d i a t i o n therapy f o r neck c a n c e r , but the p a t i e n t ' s s c a l i n g of t a s t e i n t e n s i t y of h i g h e r c o n c e n t r a t i o n s d i d not. 3.

To s e l e c t s e n s i t i v e people as s u b j e c t s f o r f l a v o r p a n e l s .

to popular b e l i e f , there are no data to v e r i f y t h a t s u b j e c t ' s

Contrary sensitivity

to d i l u t i o n s of odors or t a s t e s i s r e l a t e d to t h e i r performance in j u d g ing s e n s o r y a t t r i b u t e s of complex food systems ( 3 2 - 3 6 ) .

There i s con-

s i d e r a b l e m e r i t , however, to o r i e n t i n g new s u b j e c t s to a v a r i e t y of t a s t e and odor compounds to v e r i f y t h e i r use of c o r r e c t d e s c r i p t o r s .

This

is

p a r t i c u l a r l y a p p l i c a b l e f o r sour and b i t t e r compounds, as the two s e n s a t i o n s are f r e q u e n t l y confused, a p p a r e n t l y a semantic t r a n s p o s i t i o n

(37).

From the f o r e g o i n g , i t i s e v i d e n t , as s t r e s s e d by O'Mahony ( 3 8 ) , t h a t a t h r e s h o l d i s not a d i r e c t measure of s e n s i t i v i t y , but o n l y an i n d i c a t o r of s e n s i t i v i t y with i n b u i l t v a r i a b i l i t y due to c r i t e r i a brought to the t e s t by the s u b j e c t and introduced by the experimenter.

12

THRESHOLD vs INTENSITY

itn z UJ

z

CONCENTRATION

FIGURE 1.

Theoretical plot of relation between threshold (on abscissa)

and rate of increase of perceived intensity with stimulus concentration. A, B, C, and D signify different compounds or different subjects.

B.

Perceived Intensity

The relationship between perceived sensation and the physical concentration of a stimulus has occupied some of the best minds and energetic spirits in psychophysics during the past century.

Although ranking is

the simplest, most expedient way of rating an intensity series, it suffers from the serious disadvantage of affording no measure of degree of difference between samples.

Therefore, the main use of ranking in the

analytical laboratory is in the initial screening of test products, where the objective is to group them into high, medium, and low intensities. Scoring of the perceived intensity on numerical category scales, the oldest of rating systems, is used extensively due to its simplicity, diversity, and ease of statistical analysis.

The optimum number of

13

rating categories, i.e., the number beyond which there will be no further improvement in discrimination, is a function of the amount of discriminability inherent in the stimuli being rated and in the subjects doing the rating.

In general, a 20-step scale leaves ample room interior to the

end, and a 10-step scale probably is the absolute minimal that should be employed (39).

The psychological error of central tendency, avoidance of

the end-points, is frequently observed in scaling, suggesting that the optimum scale should have at least two points greater than what can be discriminated.

Alternatively, non-numerical or graphic scales could be

used - a vertical or horizontal line anchored with terms such as "none" and "extremely strong" on which subjects place a mark which is converted to cm for subsequent numerical analysis.

Anderson (39) notes that graphic

ratings offer several advantages over numerical ratings, as they are less susceptible to number preferences, and to possible memory effects when stimuli are repeated. The most.common misuse of category scaling reported in the literature is the combination of intensity with qualitative or hedonic terms which render the scale non-linear and non-additive.

For example, flavor de-

fects are confused with flavor strength in scales such as those used by the American Dairy Science Association (40).

Flavor strength (weak,

strong), flavor quality (green, flowerlike, old) and fruit maturity (unripe, ripe, overripe) are confounded in the scale used by Gorin et al. (41) for apples.

Taste defects, strength, and good-bad responses are

intertwined in evaluation of drinking water (42).

Dagerskog and

Sorensfors (43) had untrained judges analyze six attributes of meat patties with several intermixtures of intensity, quality and hedonic scales.

Adam (44) and Paulus et a K

(45) presented extensive explana-

tion of a 9-point scale designed for a wide variety of foods, containing intercombinations of intensity, quality, and sensory defects.

The de-

ficiencies of the foregoing combination scales could be corrected easily by separating each characteristic, and recording intensities for each individually.

Of utmost necessity is the use of trained judges and the

elimination of hedonic terms from analytical

procedures.

14

Category scales produce data which are often linear with the log of the concentration (Fechner's law where response = K log concentration).

The

main disadvantage of category scales is that the intervals are not equal, e.g., a score of 4 is not twice 2 nor half of 8.

Since the end points

are fixed, subjects may run out of scalar intervals.

The foregoing con-

straints have led many psychophysicists to utilize ratio scaling to estimate perceived sensation.

The subject assigns the first sample in

a series any number which seems appropriate (free modulus) or the experimenter presents a reference with a specified value (fixed modulus). Values proportional to the first sample are assigned to successive samples, using numbers to reflect how many times lesser or how many times greater the intensity is compared to the first.

When a large

enough number of judgments have been made, the group geometric means will approximate a logarithmic normal distribution, fitting a function of the form: Y = k e where f = perceived magnitude and = stimulus concentration.

In log-

log coordinates, this becomes: log y = 6 log $ + log k where 6 = the slope of the line and k = the intercept.

This is the well-

known psychophysical power law, or Stevens' Law (46-48), that indicates equal ratio changes in concentration produce equal ratio changes in sensation magnitude.

An exponent of 1 specifies that the two measures

increase at the same rate.

Exponents greater than 1 indicate that the

sensation increases at a faster rate, while exponents less than 1 indicate the sensation increases at a slower rate than the concentration. To eliminate differences among subjects, it is conventional to normalize the raw data by a variety of arithmetic procedures (46-48). The three most important advantages of ratio scaling over category scaling are that (a) Subjects cannot run out of numbers as the scale is infinite: (b) One can establish a physical concentration that results in proportional changes in sensation; and (c) Judgments are expressed in proportions, allowing conversion to percentages so that one ratio scale

15

can be compared directly with another.

Disadvantages relate to the com-

plexity of the task, causing many subjects to revert to category or to a combination category-ratio scaling, and the "round number" bias overselection of 2, 5, 10, 20, 50, 100, etc. (48). It is claimed by Stevens and his followers that one need not train subjects, as magnitude estimation procedures are easily understood by naive subjects, even by children (46).

Nonetheless, one of the foregoing

authors subsequently reported that only 25% of the respondents he had tested could be typified as "Clear Understanders," who had little difficulty despite no prior experience (49).

This observation is in accord

with data from our laboratory (29, 50). Only a few investigations have been concerned with the direct comparison of ratio and category scaling among the same group of experimental subjects (51-57).

The two methods were compared in the author's laboratory,

using lemonade varying in added sucrose (4, 6, 8, 10, 14, 20, and 30%) and milk varying in amount of fat (0, 1, 2, 4, 8, and 16%).

Ratio scal-

ing with a fixed modulus, designated 10, was contrasted with scaling on a 100-wn graphic scale which was then converted to cm intervals.

Hedonic

(from like extremely to dislike extremely) and intensity (from none to extremely intense) scales were used by 51 subjects for the lemonade and by 53 different subjects for the milk.

For the lemonade, two different

reference samples were tested, 4% and 10% sucrose. As summarized in Table 4, multiple criteria were applied to the data in order to contrast the two scaling procedures.

The criticism of re-

stricted number usage and the error of central tendency leveled at category scaling (C.S.) was evident in magnitude estimation (M.E.) as well.

The full range of the 100-mm rule was utilized for C.S. while

ratios from 0.1 to 100 were used for M.E.

Approximately two-thirds of

the C.S. values fell in the middle of the range (20 - 80), while up to 94% of the M.E. responses were restricted between 1 and 20.

Using F

ratios for replications from the analysis of variance as an index of reliability, only values from hedonic responses to lemonade showed significance.

For the group, C.S. was less reproducible.

However, when

the data were subdivided according to which test the subjects completed

16

TABLE 4 Comparison of Graphic Category Scaling with Magnitude Estimation of Sweetness in Lemonade and Fat in Milk Criterion

Graphic C.S.

Magnitude Estimation

Scale Range 0 - 100 0 - 100

Hedonic Intensity Number Usage

Range:

20 -

Hedonic Intensity

0.1 - 80 0.5 - 100

80

20

1 -

72% 57

94% 71

Reliability (F ratio for reps) Lemonade Hedonic (4% sue) Group (n=51) CS first (n=26) MS first (n=25) Precision

(CV =

X

8.3** 1.8

1.8 g

1

0^9

j***

X 100)

Milk Hedonic " Intensity

35 34

Lemonade Hedonic

31

Lemonade Intensity

17

24 15 = = = =

R R R R

4% 10% 4% 10%

18 20 7 14

Accuracy (r for

x . If the converse is the case, i.e. x > x., then stimulus S will be s s j s selected as the strongest. Because both variables x^ and x a r e randomly and normally distributed, the difference between each pair of x^ and x. (= x. ) is normally distributed. If the variables x and x. are assumed 3 js s : to be mutually independent, and further it is assumed that their variances are equal, i.e. 0 2 = a 2 = 1. the distribution of the difference x. is s i js normal with variance 0. = 2 . Since the mean difference between the values js x and x. (= mean x. ) is identical to the difference between their means s 3 us y - y.(y . ), it follows that: s j js z. = y. //2 (y. > 0) (2) K js "js js The parameter y ^ can be conceived as a measure for the difference between the mean sensory effects resulting from the stimuli S^ and S_. . The value of

can be estimated from an experimentally obtained response .-distri-

bution. It is essential to note that herewith the response continuum is linked to the sensory continuum. If a pair of stimuli S^ and S^ has been presented on a number of trials, the final exxDerimental data consist of two binary proportions.These are the proportion P . that x. from S was judged stronger than x. from S., S] s s j 3 and the complementary proportion P j s - Using the cumulative distribution function of the standard normal distribution N(0,1), the values of z corresponding to these proportions can be readily obtained (5,7). The values

75 z. and z . are numerically identical, but different in sign, js sj Because the Thurstonian approach is response oriented this type of procedure can be applied for all stimuli for which a common psychological attribute can be specified. A corresponding physicochemical attribute is not necessary. However, if there is a series of stimuli S s , S. (j=l,2,..,n, D ^ s) differing with respect to the degree these posses a particular physicochemical attribute, a psychophysical function of the same status as a Fechner function can be established. If the stimulus S^ has been repeatedly presented together with each of the other stimuli S_. , a series of estimates z.

( j = l , 2 , . . , n s ) can be determined. By linear regression a

Js psychophysical function can be fit through the origin, having the following form: z . = k log (S ,/S ) js j s

(S. > S ) 3 s

(3)

This function is fitted through the origin because if S_. log (S./S ) ->- 0 and the corresponding value of z.

S^, then

0.

Measurement of Sensory Differences With The Duo, Duo-Trio And Triangular Method Despite the availability of a large variety of psychological measurement procedures the so called "Sensory Difference Tests" (8) still find large scale application in flavour research. In the present section the essentials of their internal psychological structure are briefly described. Although there is a large family of sensory difference tests (9) the present discussion will only encompass the duo, duo-trio and triangular method, because these are the most frequently used in sensory analysis. Sensory difference tests oan be simply applied. In the duo method the subject is presented with a pair of stimuli S s and S^ under the instructions to select the strongest stimulus of the pair.(It is thus identical to the method of paired comparisons described in the previous section.)In the duotrio method the subject receives a standard stimulus (Ss or S.) and a pair D of stimuli S s and S_. under the instructions to identify the stimulus of the pair which is identical to the standard. In the triangular method three

76

stimuli are presented, either two stimuli S (= S ,S') plus one stimulus s s s S., or two stimuli S. (= S . ,S') plus one stimulus S . Normally, triangles 3

3

3

3

s

of both combinations are presented with equal frequencies in random order. The instructions are to select the "odd" stimulus. All sensory difference tests are forced choice procedures; the subject must guess if in doubt. The final results of the various methods are proportions: In the duo method these proportions are P(S. > Ss ) and P(S 3

chemical intensity of

s

> S.). If the physicoj

is larger than that of S^ these proportions are

identical to the proportions of correct and incorrect responses, P^ and P^ n c respectively. In the duo-trio and the triangular method proportions P and P. are also obtained. In the duo-trio method P is equal to c mc c P(S'=S ) if stimulus S served as the standard: P is identical to the s s s c proportion P(S^=S^) if stimulus S_. was designated as standard for the subject. In the triangular method a proportion of correct responses is obtained from the frequencies stimulus S^ was selected from a triangle containing the stimuli S , S 1 and S. , and stimulus S was selected from a tris s j s angle composed of S., S! and S s . 3

3

In sensory analysis a common way to handle proportions of correct responses is to test them statistically against a theoretical chance proportion. In the duo and duo-trio method these are 0.50, whereas in the triangular method it is 0.3333. Given the number of trials and a specified significance level, a binomial test on an experimentally obtained proportion is performed. (Appropriate statistical tables have been produced, e.g.(10)). If the experimentally obtained proportion is shown to be statistically different from the theoretical chance proportion, the null hypothesis predicting no difference is rejected. It is then concluded that the stimuli Sj and S s are sensorily different. Although the testing procedure performed as described is theoretically valid, it will be clear that the conn' elusion drawn is completely arbitrary. This results from the fact that any proportion unequal to a chance proportion can be made statistically significant by increase of the number of observations. In conclusion, this means that any degree of discriminability greater than zero can serve to make stimuli "sensory different". This is the reason why a measurement approach should be preferred over a testing procedure.

77 The sensory and judgemental processes occuring in the main sensory difference tests can be described in the concepts Thurstone used for the method of paired comparisons.In the remainder of this section the assumptions made are identical to those mentioned in the discussion of Case V of the Law of Comparative Judgement. These were: There are two types of stimuli S^ and S.. These stimuli can give rise to discriminal processes x

3

pectively. The sensory difference y. - y

3

denoted

s

= d1

and x. resS J ( d" > 0) (Note that d' was

in the previous section) is identical to the difference between

the mean discriminal processes Since y. > ys ,in the duo method a response can be classified as correct or

3

incorrect. If in a particular trial x^ > x^ a correct response will be produced because the subject will select the stimulus S_. as the strongest of the pair. In the duo-trio method a correct response will be obtained (11) if: a) |x^ - x'^ | < | x s - Xj | , when S s was presented as the standard stimulus, and, b)

x. - x! < x. - x , when the stimulus S, served as standard. s D J 3 3 In the triangular method the subject will respond correctly when: a) |x x ' | < l x — x.I and |x x ' l < |x' x.l , when the triangle 1 1 1 s s1 ' s j1 s s' s D

presented contained the stimuli S^, S^ and S , or if, b) Ix. - x!| < Ix. - x I1 and |x. - x l l < |x! - x

' D

s

D

was composed of the stimuli S., S'. and S .

3

3

3

I , when the triangle

s1

s

Given these internal representations and the concomitant'decision rules, for each of the three procedures a mathematical expression relating the proportion of correct responses, P , to d' can be specified. Assuming that the variables x . and x s are mutually independently and normally distribut-

3

ed with equal variances,these expressions are (11,12,13); for the duo method: P

= $(d'//2) c for the duo-trio method:

(4)

= 1 - $(d'//2) - 4>(d'//6) + 2(d'//2)'I>(d,/>/6) c and for the triangular method:

(5)

P

p °

= 2/{(-U/3 + d'/(2/3) + $(-u/3 - d'/(2/3) }exp(-!5u2) //2ir) du o (6)

78 In the equations (4) to (6) $ is the cumulative distribution function of the standard normal distribution N(0,1). Through the above expressions the sensory difference d 1 can be estimated from an experimentally obtained proportion P . In order to enable a rapid conversion, Tables of d 1 were produced by Ura (11) and David & Trivedi (12). More recently a more elaborate Table of d' for the triangular method was published (14). The parameter d' can be estimated by each of the three sensory difference tests. Theoretically, the value of d 1 should be method independent: it should not make any difference by which method it is obtained. At the moment no data are available showing the equality of the three sensory difference tests in this respect.(however, see also next section). It should be stressed that the present models are based on the independence of discriminal processes. When applying the sensory difference tests in flavour research, this independence assumption can easily be violated due to sensory adaptation. From this point of view the duo method is to be preferred over the duo-trio and triangular method.

Multiple Alternative Forced Choice Procedures And The Theory Of Signal Detection The principles of Thurstonian measurement are rather similar to the principles of statistical decision theory on which the Theory of Signal Detection has been based (15). In the present section attention is paid to a small part of this theory. The discussion will be restricted to the measurement of sensory discriminability using one of the multiple-alternative forced choice (m-AFC) procedures. In the method of paired comparisons the subject is instructed to select the strongest stimulus from each pair presented. One could also present the subject with three stimuli, or four, or m. Discriminations paradigms in which the subject has to identify a particular stimulus out of m stimuli are thus logical extensions of the duo method. In an m-AFC procedure the subject has to select the strongest stimulus S. ( o r the stimulus contain-

79 ing a "signal") out of m stimuli, of which m - 1 are identical to S^. Expressing the sensory and judgemental processes in the notation used earlier, the stimulus S_. generates a normally distributed internal sensory response x^. Each of the remaining stimuli S^ elicits a sehsation of the value x g , so that m — 1 sensory responses x s occur. It is assumed that a correct response will be obtained if in a particular trial the momentary value of x_. is larger than each of the m - 1 values x^. Stated differently, the subject will respond correctly if x_. is larger than the largest value (x. > x ). Given a particular value of d' the probability that r s ] s.max x. > x decreases if m increases and thus the probability of a correct j s.max response decreases.

x

Again, let the means of the normally and independently distributed x . and

J

x^ be

and P s respectively. If their variances are assumed to be equal,

the sensory discriminability between stimulus S^ and

is d'

=

Vj ~

(u. > u ). For all m-AFC procedures the general expression relating the proportion of a correct response P

to d' can be specified as follows: (7)

In this expression $ is the cumulative distribution function of the standard normal distribution and if not, there is a disagreement. If judges are differentiating samples equally well, by rotating/reflecting the configurations so that they overlap as nearly as possible (still maintaining the configuration shape), an indication of the degree of confusion between adjectives is obtained. Scaling (see Example, Fig. 3). In the final operation of the Procrustes analysis configurations are matched using an isotropic scale change, the size of the scaling factor relating to the range of the scales used by the judges in differentiating samples.

The distance of each assessor from the sample centroid can be inspected at each stage in the operation to discover whether the grouping has been improved and hence how much variation is accounted for by that particular transformation. Lack of matching after applying generalised Procrustes analysis implies that judges are discriminating between the sample in ways other than those allowed for, either by using different ranges of the scales when differentiating different attributes, or because they are perceiving or merely scoring something that the other judges are not (Judge D in Example, Fig. 3). Non-shape preserving transformations such as INDSCAL (13) will help elucidate the first of these. In the latter case multiple assessments of the same samples will be required to differentiate between a judge who is grossly inconsistent and those who are consistently perceiving different stimuli from the majority of the panel. The existence of sub-groups can be further investigated using pairwise Procrustes analysis (14).

86

Example The following example discusses a simulated assessment of four stimuli (I - Astringent, II - Bitter, III - Acid and IV - Sweet) present in five beverages (1 - 5) by four judges (A, B, C and D).

In comparison with judges A and C, judge B

confuses attributes I and II, judge C scores lower than others and D behaves completely differently from the other three. is assumed that each attribute is scored using a line scale (length 12 cm). The initial configuration for attributes I and II, which illustrates confusion, is given in Fig. 1. The initial distances from each judge to the average configuration are given in Table 1. The initial translation terms for all attributes for each judge are given in Table 2.

TABLE 1.

2 Initial distances from each judge to the average configuration Judge

Distance

TABLE 2.

2

A

B

C

D

43.24

61.09

14.49

92. 47

Translation terms : means for each judge Judge A

Attribute

I II III IV

6.11 6.23 5.58 6.99

B 6.12 6. 74 5.67 6.65

C 5.90 6. 44 5.79 6.25

D 5.88 6.44 5.97 6.56

It

87

12

10-

O o

Astringency

e

4-

© ©

2-

a

0

I

6 Bitterness

Fig. I.

T

!

10

Initial configurations

Translation TABLE 3. Distances from each judge to the average configuration after translation Judges

Distances

2

A

B

C

D

42.90

60.97

14.15

91.23

Squared distances after translation (i.e. distances after bringing to common origin) are given in Table 3.

There is

only slight improvement in grouping as judges show fair agreement with respect to the region of the scale used.

88 Rotation (Fig. 2) Squared distances from average after rotation and reflection are given in Table 4. Elimination of confusion between words has improved the conformity amongst judges. 2 TABLE 4. Distances from each judge to average configuration after translation rotation/reflection Judges

Distance2

A

B

C

D

29.40

24.03

18.81

57.13

5 m

U

\ +2 %

© ®

I

1

I

i

o

I

!

-1o

®

-2

-3

V@

-4

©

--5 -6

Fig. 2.

Configuration after translation and rotation/reflection

89

Scaling (Fig. 3) 5J&3) - (gÈf 3-

.

S

•

©

©

®

&

2®

1- o

-U

-3

-2

0

-1

1

2

5

3

-2

e 1 A4'

© 9

®®

(ci)

-5 -6

Fig. 3. Configuration after translation, rotation/reflection and scaling 2 TABLE 5. Distances from each judge to the average configuration after translation, rotation/reflection and scaling Judge A Distances

2

9.72

B 8.25

C 8.14

D 61.14

Squared distances from mean after rotation, reflection and scaling are given in Table 5;

further improvement in fit is

obtained and judge D indicated as being different from the other three judges.

Re-examination of judges A, B and C alone

90

produces the configuration shown in Fig. 4 and the distances given in Table 6, indicating just how close these three really are as a group once the various systematic causes of variation are removed.

44 3-

2-

iWi wL

®O

m

1-

2

-1-5

J

- 5

0

•5

1

15

2

-1-

-2-

-3(A 4} v—••-'

ffi

-5-

Fig. 4. Configuration of judges A, B and C after translation, rotation/reflection and scaling 2 TABLE 6. Distances from judges A, B and C to their average configuration after translation, rotation/reflection and scaling Judges A Distance2

0.86

B

C 2.39

2.64

91

Evaluation of how judges originally differed Inspection of individual rotation matrices for the three judges (A, B and C) who finally emerged as scoring the same stimuli (Table 7(i), (ii) and Ciii)) indicates that judges A and C used adjectives similarly when describing the product whereas judge B interchanges the meaning of adjectives I and II (astringency and bitterness).

From the scaling factors

(Table 8) it is also apparent that judge C continually underscored compared to judges A and B. TABLE 7.

Rotation matrices for Judges A, B and C Rotation Matrix - Judge A

CD

(ii)

0.00 0.00 0 .00 1.00 0.00 1.00 0.00 0 .00 O.OO 0.00 1.00 0 .00 0.00 0.00 1 .00 0.00 Rotation Matrix - Judge B 1.00 0.04 -0.01 0.06 0.05 0 .00 1.00 -0.07 1.00 -0.00 -0.04 0 .02 0.00 -0.01 -0.02 1 .00

( iii) Rotation Matrix - Judge C 0.98 0. 50 0.18 0 .01 0.04 0.99 -0.08 0 .10 0.07 -0.19 0.98 0 .07 -0.00 -0.10 -0.06 0 .97

TABLE 8.

Scaling factors for Judges A, B and C Judges A

Scaling factor

0.85

B

C 0.86

1.89

92 Conclusion Discovering the underlying causes of variation in sensory profile data, as outlined, enables those of a systematic nature to be removed prior to comparing assessments of different samples, hence improving the precision of any investigation.

Examinations of the differences between sub-

sets of panelists once systematic variations have been eliminated will also possibly help to explain why people differ in their acceptance of a given food or beverage.

References 1. Caul, J.F.: Adv. Fd Res. 7, 1-40 (1957). 2. Williams, A.A. : J. Sci. Fd Agric. 567-582 (.1975). 3. Meilgaard, M.C., Dalgliesh, C.E., Clapperton, J.F.: Amer. Soc. Brew. Chem. 37, 47-52 (1979).

J.

4. Williams, A.A., Baines, C.R., Arnold, G.M.: Proc. International Centennial Symposium of Viticulture and Enology, University of California, Davis (1980) (in press). 5. Clapperton, J.F., Piggott, J.F.: 275-277 (1979). 6. Frijters,J.E.R.:

J. Inst. Brew. 85,

Poultry Sci. 55, 229-234 (1976).

7. Vuataz, L.: Nestle Research News, 57-71 (1976/77). 8. MacFie, H.J.H., Gutteridge, C.S., Norris, J.R.: J. Gen. Microbiol. 104, 67-74 C1978). 9. Harries, J.M.: Personality and Sensory Acuity. Research Institute, Memorandum No. 23 (1975). 10. Henderson, D., Vaisey, M.: (1970).

Meat

J. Food Sci. 35, 407-411

11. Williams, A.A.: Proceedings of the 7th Long Ashton Symposium, April 1979 (in press). Applied Science Pub. 12. Gower, J.C.: Psychometrica 40, 33-52 (1973). 13. Carrol, J.D., Chang, J.J.: (1970).

Psychometrica 3j>, 283-319

14. Banfield, C.F., Harries, J.M.: 1-10 (1975) .

J. Food Technol. 10,

SENSORY EVALUATION IN A "NATURAL ENVIRONMENT"

E.P. Köster Psychologisch Laboratorium, Rijksuniversiteit Utrecht, The Netherlands

Over the past twenty five years sensory analysis of the properties of foods has made considerable progress. Better research methods have been developed, and better statistical and mathematical techniques have been employed. Although still in many places these new methods and techniques are not fully used or are used often improperly, they are now available to those who want to use sensory analysis for the characterization and.control of food properties. Much attention has also been given to the standardization of test circumstances. Samples are presented in coded containers in blind procedures, precautions are taken to get independent measurements by placing subjects in separate testing booths and all kinds of response biases arising from constant errors such as code preferences and side preferences are effectively overcome by systematic variation of conditions over subjects. Finally, good care is usually taken to overcome the effects of irrelevant cues such as colour differences in an odour test or texture differences in tests principally concerned with taste differences. All in all, the sensory labororatory has grown into a powerful tool for the analytical assessment of properties of foods and beverages. Difference testing, multidimensional scaling and good descriptive methods make it possible to answer a wide variety of questions. Further refinements of the techniques, based on the progressing psychological knowledge about the functioning of the human subject, which is used as a measuring device in sensory analysis, are constantly made. Thus, Frijters (1980) very elegantly showed the implications of modern decision the-

© 1981 by Walter d e Gruyter &. Co, Berlin • New York Flavour '81

94 ory for the use of such well-established methods as the triangle test in sensory evaluation. This direct translation of fundamental psychophysics and psychometrics into practical suggestions for better use of methods is a hopeful sign of the different way in which the sensory evaluation tool is now developing. What people can do and how they can perform in the best possible way are basic questions to this development. In the fields of selection and training of sensory panel members there is still a lot to be learned from psychology and we will have to devote more attention to this in the near future, but on the whole the progress has been remarkable and sensory analysis has become firmly established as a laboratory tool. This has also served to overcome much of the distrust that instrumental analysts in food science have for such intrinsically subjective measurements as sensory judgements. Although we can therefore be quite optimistic about the future we will nevertheless have to remain cautious. There are two main and interconnected reasons for this caution. In the first place we must admit that the positive development described, has mainly taken place in the field of quality control, concerning analytical questions such as "Is there a difference between todays' production and our internal standard" or "How can the off-taste detected be characterized?" If on the other hand questions like "Which of these samples is to be preferred?" or more complicated questions like "Will this coffee cause a bad stomach feeling after ingestion?" arise, the possibilities of coming up with solid answers are much smaller. Unfortunately, in many instances panels who were selected and trained to answer the first type of analytical questions are also used to give the second hedonic type of answers. It is often hard to convince people that experts and semi-experts by the nature of their training have lost the possibility to give valid hedonic judgements and that in such a case laymen should

95 be preferred. But even when laymen are used they are often brought into highly unnatural situations and this has a direct and deleterious influence on the validity of their answers. And this is the second reason for caution. The very nature of the laboratory experiment with its rigid control of all sorts of extraneous variables may be a limiting factor in the establishment of the right responses. In some instances we were able to show for example that preferences for non-alcoholic beverages obtained in paired comparison tests or with the best scaling methods did not predict at all the actual consumption pattern observed in natural situations like a party or at home. The so-called sweetness comparison test provides another example of such spurious results. If one gives the subject the choice between two beers, one of which is sweeter than the other, he will usually prefer the sweeter one over the other. If one repeats this test using the highest sweetness of the previous test as the lowest one in the trial, the subject will again choose the even sweeter one. After several repetitions of this test one can lead the subject to a preference for an extremely sweet beer. If after the test one proposes to the subject to have a glass of beer at the bar before leaving and one presents him under natural conditions with a beer of the sweetness he has just preferred, he will very often reject it as being appallingly sweet. Of course this trick will only work if one progresses only gradually in giving sweeter beers in the first stages of the experiment, but the result shows nevertheless that the unnatural paired comparison technique has induced the possibility of false answers. What is so unnatural about the paired comparison technique in this case ? On the whole, people are good in comparing things and comparison methods are usually highly recommended over methods involving absolute judgement like rating scales or ma-

96

gnitude estimation. So why should the use of the paired comparison method be considered to be unnatural here ? The answer is twofold. In the first place one can say in general that the method is intrinsically unnatural because each separate judgement made in paired comparison is made against a very limited background. The two pair members are only compared against each other. All other possible stimuli, or indeed the normal every day internal standards of the subjects seem irrelevant at that moment of decision in the test booth. The second point is that in this case this fundamental objection is not counteracted by making the subject compare all the possible stimulus combinations in the range. Remember that the subject gets only successive pairs in a series of slowly ascending sweetness. If we would give him also the pair containing the sweetest and the least sweet beer, the difference might be so large that he is forced to go to an internal standard as a reference point to relate both of the pair members to, or that at least he is forced to realize that the stimuli he is dealing with in the other pairs are somewhere in the middle of a wide range of stimuli. This may help him to avoid the limited background trap set by the paired comparison method more effectively. Thus, we see that even a good method like paired comparison when applied to simple preference questions has its dangers. In another context (Koster, 1981) we have already criticized the low validity of the simple preference measurements usually made as regards the prediction of long term food acceptance. There we pointed out that in itself asking questions about preferences may be more unnatural than we are inclined to think at first sight. Very many of our daily decisions about preference are made unconsciously and we very seldomly are asked to make our preferences explicit. This also explains to a large extent the discrepancies between explicit preference testing and directly measured consumption behavior. If we turn to more complex questions such as "Does this coffee

97

produce 111 feeling in the stomach after ingestion?" the inadequacy of laboratory procedures becomes only more evident.Drinking coffee in isolation in a test booth of one sguaremeter under artificial light at three o' clock in the afternoon, with a rating scale form in front of you, might in itself be a reason for the stomach to be upset, or for the person to express negative feelings on his subsequently handed out questionnaire, which by the way will undoubtedly seem to be too long to him and will contain often unnecessary and infuriatingly ambiguous questions about his feelings. Thus, in situations in which hedonic responses are required, the unnatural circumstances may themselves influence the judgement more than the actual stimuli. In the earlier mentioned paper, we have suggested the use of direct observation methods and of behavioural measurements in natural situations, like time and frequency analysis of behaviour, as a better solution to these problems and as at least a way to validate and complement the traditional measures of preference. Since then we have employed these methods with considerable success in some applied research concerning complex questions, but unfortunately I do not have as yet the liberty to disclose the results of these recent findings. Before we will make some suggestions of possible ways to carry out observational experiments, we will in a general way discuss the dangers of the laboratory situation, and point out the differences with the natural situation. In doing so we will make a difference between analytical and hedonic, simple or complex questions. Essentially in the laboratory we may impose five limitations on the subject. They are limitation of human contact (i.e. working in isolation), limitation of naive behavior (i.e. having to make explicit statements), limitation of normal use (i.e. having to eat at unusual times or under unnatural circumstances), limitation of stimulus context (i.e. presentation of a limited range of stimuli) and limi-

98

tation of response possibility (i.e. forcing certain types of responses on the subject). They are represented in Table 1 where also some indication of their effect is given. TABLE 1.

LIMITING FACTORS IN SENSORY EVALUATION HEDONIC

LIMITATION OF HUMAN CONTACT NAIVE BEHAVIOUR NORMAL USE STIMULUS CONTEXT RESPONSE POSSIBILITY

ANALYTICAL

SIMPLE

+ + + +

+

+

COMPLEX

+

Limitation of human contact has positive effects in analytical procedures. Measurements should be as independent as possible in this case and subjects are especially trained to judge independently of others. In the case of simple preference testing lack of human contact does not have to be bad, although in some instances it may lead to more negative responses than would be obtained in normal discussions between people. If complex questions are involved, this latter disadvantage becomes predominant. Here observation in the normal every day situation gives better results. Limitation of naive behavior is inevitable in good analytical work. In judgements of preference it always has negative effects. Any form of direct observation has the advantage of leaving the subject in his natural behavior. The amounts he consumes, the frequency with which he eats or drinks and the intervals between successive ingestions give usually better indications of what he really likes or dislikes than his explicit statements. Limitation of normal use may prove to be bad even when analytical procedures are involved. Eating cheese under blue light will almost certainly influence even the analytical judgements. But even less deviations from normal procedures as for instance the spitting out of alcoholic beverages contain the risk of false judgements because certain types of receptors in the larynx,

99

which play an important role in normal consumption, are no longer involved. In the case of hedonic questions, any deviation of normal use should be avoided. Limitation of stimulus context has already been discussed to some extent in the sweet beer example. Even in analytical procedures it is important that the subjects remain aware of the often large variations in the range of relevant stimuli. This is especially true when absolute judgements (rating scales etc.) are involved. It can easily be demonstrated that the range of stimuli in which a particular stimulus is embedded influences the judgement. Kelsons' adaptation level theory, which was developed in psychophysics applies here. In preference judgements a good overview of the normal stimulus context is imperative. Limitation of response possibility by forcing the subject into a limited number of response categories is often inevitable and may in some cases even be advantageous, depending on the specific demands by the method of data treatment involved. On the other hand, it is well-known that in analytic tasks open line judgements are preferable over rigid category scales. In preference testing limitations of response possibility are more often ill-advised, especially when complex questions should be answered. Nevertheless, even in these cases some specifications of permitted responses can usually not be avoided. All in all, one can conclude that the laboratory experiment is relatively well suited for analytical purposes but that preference testing makes much heavier demands on the natural conditions of the experiment. Observational methods are often more time consuming, because they demand a lot of preparation, but they certainly are more rewarding in the case of complex hedonic questions. In most cases they take the form of organising a party or a series of parties at which the food and drinks to be tested are available at an ad libitum basis.Each person receives a book of coupons which he can use to get any

100 item he or she wants. The only restrictions are that everybody has to get each item for himself and that no switching of items occurs. Also subjects may not take any of the items home but at the end of the session they will receive a package containing a number of them. Such an experimental session may involve as many as 100 subjects at a time, depending on the nature of the question and on the location where the experiment takes place. During the experiment the coupons are stamped for the time at which they are handed in or are collected at regular intervals in such a way that the experimenter can later see exactly in which interval which item was obtained by which subject. These data then provide the basis for the time and frequency analysis of the consumption behavior. During the period the people are entertained by short games (for instance a lotto in which small prices are available) or they entertain themselves. In almost all cases people are highly appreciative of this type of experiment. They seem to enjoy themselves and thus the experiment also provides good public relations for the sponsoring firm. In the earlier mentioned paper we have given some indication of the value of the results of these types of experiment. They may show preference patterns in a very clear way and they may also provide the basis for the selection of groups of subjects with special consumption habits for further questioning. In another type of experiment we will observe the eating and drinking habits of single persons in a small group with the help of video tapes. This experiment will be concerned with the influence on appetite of variety of foods available on the plate of a person. We also hope to be able to work on a project in an industry where beverages are available to the personnel from dispensing machines. Coupling these machines with devices for time clocking and asking the people to use their time clocking cards to operate the machine will be a very efficient way of collecting data over long periods of time. It is hoped that in the near future we will be able to give a more detailed account of the data obtained in this way.

SECTION

II

APPLICATION OF SENSORY METHODS

PERCEPTION AND ANALYSIS:

A PERSPECTIVE VIEW OF ATTEMPTS TO

FIND CAUSAL RELATIONS BETWEEN SENSORY AND OBJECTIVE DATA SETS

John J. Powers Department of Food Science, University of Georgia, Athens, Georgia 30602 USA

Introduction A challenge scientists and technologists often face is to distinguish between a specimen of food of acceptable grade and one still somewhat higher in grade.

Sometimes a single test

suffices to classify samples, but usually so only between those which are clearly substandard and those of acceptable quality. The difference between a good food and one of the same kind but slightly better almost always resides in the cumulative effect of minor differences among several chemicals, not a major difference in one or two chemicals.

This applies both to sensory

differences and chemical composition and to foods and other substances of sensory importance.

Not only is decision far

more complex within the range of commercial acceptability, but far more decisions have to be made in this region than as between acceptable substances and those which are substandard. Until the mid-1960s, most efforts to distinguish among products of nearly like quality depended upon applying first one test, than another, and attempting to resolve differences based on a whole series of decisions.

Figure 1 illustrates this process.

Assume that we have six food products which differ in composition.

Using test Tj, some of the sample differences can be re-

solved because the measurement values are sufficiently far apart, but others cannot be because the overlap between the

© 1981 by Walter de Gruyter & C o , Berlin • New York Flavour '81

104

MEAN VALUES

Figure 1.

Resolving of product differences using individual, scalar tests.

measurement values is too great. Test T2 can-be used to distinguish among some of the samples not resoluable by test T^. Test T3 can likewise be used to separate some samples from other samples. The difficulty with this procedure is that there is a probability risk associated with use of the first test, the second and the third; after all the decisions required have been made, one does not really know what the probability is that the final decisions are sound. The procedure is fraught with great uncertainty, inefficient as to effort and often as to results. Matrices vs. individual tests.

Figure 2 shows the data being

treated schematically as a set or a matrix.

Especially where

sensory evaluation is involved, there are good reasons for examing sets of tests rather than single, scalar tests as in Figure 1.

Well known to food scientists, perfumers, flavorists

and others who work with sensory materials are the facts that a component by itself unpleasant may be essential to good

105

Products

A B C D E F Figure 2.

Tests Ti

T2

T3

1.2 1.8 2.9 3.6 4.2 4.9

2.7 0.8 1.6 4.3 3.4 4.8

1.4 0.7 2.2 3.6 4.2

X

2.8

Schematic formation of a matrix from a set of measurements.

sensory quality, that the concentration of the solvent often affects pleasantness or unpleasantness, that certain compounds mask other compounds or exhibit additive, less-than-additive, enhancing or synergistic effects when in combination. Various attributes thus impinge upon each other in different ways depending upon the presence of other compounds and their concentrations. Moreover, they impinge upon each other at different points (or not at all) as a result of differences in perception by the individuals subject to the sensory signals. When sensory and analytical measurements are being correlated, there is the added difference of scales involved. Chemical concentration, for example, is expressed in a different way than spectral absorbancy. Figure 3 illustrates the interplay among different variables throught the third dimension. Each attribute evaluated and each objective test made constitues one more dimension. Multivariate analysis is needed to bring some kind of order out of that which initially appears to be chaos. The vector relations of Figures 2 and 3 can be expressed as correlation of coefficients since the cosine of an angle is numerically equal to the correlation coefficient. Correlation vs. causal relations.

To detect strong correlations

between multivariate sensory and objective data sets is often difficult.

Notwithstanding these difficulties, there has been

considerable success as judged by the volume of literature

106

Figure 3.

Intersecting of various attributes and objective tests in space.

appearing and by the substantial applications being made by industry. To establish causal relations is even more difficult because of crucial gaps in fundamental knowledge of sensory perception. Notwithstanding the greater difficulties, some causal relations have been discovered. The time has come to put greater effort into the task, for practical applications would undoubtedly be more fruitful than those based on correlation alone and in carrying on such studies, some of the very gaps which presently cause difficulty would begin to be closed. This discussion of attempts to find causal relations between sensory and objective data sets will therefore revolve around

107

five main themes. A.

They are:

Problems involved in seeking out correlations and causal relations,

B.

Practices of value in attaining the goals above,

C.

Pitfalls to be avoided,

D.

Principles and pathways likely to be productive, and

E.

Potential for success, for improved practical applications and for expansion of basic knowledge.

Beneficial applications. Before discussing problems, the success attained in detecting correlations between multivariate sensory and objective data should be mentioned lest the impression be given that the task is almost hopeless. Actually, the benefits which have flowed from deriving correlations between sensory and objective data sets is impressive. Today there is a steady flow of publications and some of these go beyond the correlation stage to the establishment of causal relations (1, 2, 3). The publication of Hoff et al. (4) should perhaps also be mentioned. Through the use of sensory, chemical and discriminant analysis (DA), the Miller Brewing Company (USA) successfully modified the flavor of beer being produced in one of its branch plants so that chemically and sensorially it was indistinguishable from the beer being produced in another one of its plants. The flavor of the beer produced in the second plant was apparently the flavor Miller wanted in its beer. Originally sensory analysis and analysis of the volatiles in the headspace of the beers showed that the beers were different. From DA of the headspace volatiles, the significant chemical differences were ascertained, and from knowledge of these, clues were obtained as to process variables which needed to be changed to bring the sensory pattern of the first beer into line with that of the second. After making the process changes, neither by sensory nor by chemical analysis could the two products be distinguished any longer.

108

The success attained by Galetto and Bednarczyk (1) is a true cause-and-effeet study.

Using stepwise regression analysis

(SKA), investigators of McCormick & Company were able to identify three components of natural oil of onion which accounted for 87% of the assessors' perception of flavor.

Three other compo-

nents were separated by gas-liquid chromatographs (GLC) which accounted for some of the remaining sensory character.

The

investigators were then able to formulate a synthetic mixture which subsequent sensory analysis verified as having the character of natural oil of onion. The studies just cited are probably but the tip of the iceberg as to industrial applications. Because of the propensity of industry not to publish, there probably are many more examples of effective applications of sensory/objective analysis as judged from those cited above and others (5, 6). Multivariate correlations are useful in quality assurance programs, in establishing purchasing specifications and in detecting adulteration, in guiding process changes as Hoff et al. (4) report and to advance basic knowledge of relations between sensory attributes and chemical composition. Problems Having pointed out that effective applications can be derived from sensory/objective analysis, let us now turn to three of the crucial problems which face those attempting to establish causal relations.

These problems are:

A.

Antipodal kinds of data acquired,

B.

Integration of signals upon perception, and

C.

Having to describe perception as a sensation instead of describing the properties of the compounds acting as stimuli.

109

Antipodal kinds of data. One of the difficulties in correlating sensory and objective results is that chemists and sensory assessors produce different kinds of data sets. Chemists do their utmost to separate, identify and quantify compounds. Sensory assessors would like to separate sensory nuances as cleanly too, but they have to work against the background of all the other sensory attributes present. Figure 4 illustrates the problem. If one examines the pieces of a Chinese puzzle ball separately, partly assembled and fully assembled, the three stages are much like the consequences of chemical analysis, panel assessment and consumer response. Like the designer or craftsman who fashioned the pieces, the chemist deals with the configuration of individual units. Like a sensory assessor attempting to tease from the assembled product each individual attribute, the one dissembling the puzzle ball has to study the interlocking of the pieces lest they be damaged so that the ball cannot be assembled again. The consumer is interested in the product assembled, in its form or symmetry, in the grain and polish of the wood, and in other qualities which give it esthetic appeal.

Figure 4.

Chinese puzzle ball, disassembled, partly assembled and fully assembled.

110

While the chemist's data and the sensory analyst's data are disjointed, this has not precluded the discerning of correlations between the two kinds of data sets, for correlation says nothing about a relationship. The gap between the very separate kinds of data the chemist generates and the somewhat less discrete data the sensory assessors provide merely lessens the chance of finding correlations. Integrative effects: The gap which exists as to the detecting of correlations become a chasm when causal relations must be established. We humans are marvelous signal receiving-integrating devices. We sense hundreds of compounds, but we then integrate their signals to provide ourselves with relatively simple perceptual patterns. We make separate assessments as to appearance, odor, taste, feel and sound and as to various sensory nuances within each one of the sense modalities, but perception is almost invariably expressed as a simple decision. We decide the product is highly desirable, acceptable, nondescript or perhaps even repulsive. The response is simple though it may have been based upon perception of hundreds of sensory signals. Let us use an example. Meilgaard reports (7) that at least 840 compounds have been identified in typical market beers and that approximately one-eighth of them are "flavour-active", that is, they are close to or above threshold levels. We receive signals from the 110 compounds, integrate our response to them and make a decision. In integrating signals from the 110 compounds, we have combined them in some new fashion. In most instances our ultimate decision is not based upon the separate stimuli alone but upon some unknown number of combinations drawn from the millions of combinations into which the 110 signals can be formed. Table 1 lists the number of possible combinations if only 16 chemicals are present but combined in different ways. Table 2 lists the almost incomprehensible number of combinations resulting from various numbers of

111

Table 1.

Number of combinations when 16 chemicals are combined in all possible ways.

Unit

Number of Combinations

Single Pairs Triplets Quadruplets Sextuplets Septuplets Octuplets

16 120 560 1, 820 4,368 8,008 12,870

Unit

Number of Combinations

Nonatuplets Decatuplets Undecatuplets Dodecatuplets Tridecatuplets Quadecatuplets Quindecatuplets Hexadecatuplets

11,440 8, 008 4, 368 1, 820 560 120 16 1 65,535

compounds up to 100; yet rarely does a food, save some formulated foods, contain as few as 100 chemicals of sensory importance (8, 91, 10) . We are familiar with our responding to combination of signals. When we listen to a symphony orchestra, our response at any given moment is to some combination of instruments and notes being played. Unless one is a musician, one does not normally recognize each instrument and note played. We respond to some pattern which we decide is melodic, dissonant, played sotto or forte. In fact, a physicist might sometimes be unable to Table 2.

Number of sensory chemicals and possible combinations.

Number of Chemicals 5 10 15 25 50 75 100

Number of Possible Combinations 31 1,023 32,767 3.35543 1.12588 3,77789 1.26768

x x x x

107 10i5 10 10 3 0

112 decompose the sound of the music into each of the originating frequencies and intensities. In that regard, chemists do better. Today, they can separate, identify and quantify hundreds of compounds, all in the same sensory material and often present initially only in a few parts per million or billion (11) .

Imaginary chemicals: The consequence of our perceiving separate signals but of integrating them is that we no longer are relating 110 stimuli to our sensory response. There is a veiled domain where an unknown number of integrated signals of unknown combination are at least partial determinants of sensory response. This is just as if out of the vast number of possible combinations new chemicals had been formed. The compounds have not reacted chemically, but they are perceived as if they had. Quite obviously, the chemist cannot measure these mythical chemicals which exist only in the recesses of the brain. Because of sensory integration, the task now is to try to match up dozens of chemicals with a response when perhaps no one of the separate chemicals is even reminescent of the sensation generated. The dichotomy in the form of answers chemists and sensory analysts arrive at is not too serious if correlations are all that are sought, but if one wants to establish causal relations, then the fact we usually do not know the immediate cause of the sensory response is a serious handicap indeed. Figures 5 and 6 illustrate the dichotomy. Figure 5 shows the geometric relations of eight snap bean products analyzed chemically. At the fourth step of a stepwise discriminant analysis (SDA), four chemical tests permitted proper catagorization of all 32 samples according to the four brands of canned beans, the three of frozen beans and a fresh product.

113

-18

-12

-6 CANONICAL

Figure 5.

Fik-a i— 6 12

0

VARIABLE

24

30

I

Geometric location of eight green bean products, cooked for serving, based on the fourth SDA step using chemical tests. C = canned brands, F = frozen brands and Fh = fresh, cooked green beans. Numbers following letters stand for different brands.

1.3 _

-2.0

-1.5

-1.0

-OS

0

0.5

CANONICAL VARIABLE

Figure 6.

1.0

1.5

20

I

Geometric location of same products as in Figure 5 except based on SDA using the seven best sensory discriminators in combination.

114

Both sensory and chemical tests were performed on the beans cooked as for serving. Figure 6 shows the sensory classification. It makes much more sense. The patterns of Figures 5 and 6 can be correlated with each other, but quite obviously the pattern for Figure 6 was determined by some re-arrangement of the chemical signals upon perception. Descriptions, not properties. A third critical gap in fundamental knowledge is that we rarely can define the properties of a compound which makes it a sensory substance. There are exceptions. Saltiness and sourness are fairly easy to explain in terms of other properties of salts and acids. There are theories as to the perception of sweetness (12), to bitterness (13, 14, 15), to olfaction (16, 17) and the role of hydrophobic and ir-electron interactions in the complexing of purines and flavor compounds (18). By and large, however, instead of being able to specify compounds in terms of properties which make them sensory substances, we have to fall back on descriptions of the sensation they generate. Word descriptions and reference compounds have been suggested (19, 20, 21, 22); still, we often have to denote the presence of one substance in terms reminescent of some other substance, i.e., like cabbage (7), like a wet dog (2 3) or by some rather general effect such as causing one's nose to sting or be cough provoking (24). Procedures and Practices In most instances correlation methods need to precede efforts to detect causal relations. The various correlative procedures will first be described as to their general function and some of their features; then practices of aid in refining data will be commented upon. Further description can be found elsewhere (25, 26, 27, 28, 29). DA will be discussed first because it is widely used. DA is one of the procedures for classifying individual cases into groups so that things in the same group have properties or attributes in common. DA was devised in

115

1936 by Fisher (30). He was interested in classifying irises according to species. He had four measurement values: petal width and length and sepal width and length. All four of these measurements overlapped greatly as to each of the three species. This is the typical situation most of us encounter in assessing differences between foods or other sensory materials. In the case of foods, the classes may consist of brands, treatments or classes, the last simply because a product has been evaluated as to flavor, astringency or other particular attribute. DA may be used to form classes (31, 32), to rank variables according to their efficiency as discriminatory variables (4, 9, 33, 34, 35, 36, 37) or the discriminant function may be calculated and then later used as a predicting equation by substituting in new measurement values (37, 38, 39, 40, 41, 42, 43). Levitt's (44) publication furnishes an illustration of the stepwise procedure because he reported the classification results at each step. Levitt evaluated 17 6 gel specimens using 22 different sensory tests and 13 objective tests. Table 3 shows that when the best sensory test, K, was applied 53% of the gels were misclassified. When variable S was coupled with K as a pair, then 60% of the 176 gels were correctly classified. By the eighth step, classification was as successful as it was to be, 88%. Stepwise programs are set up to add variables as predicting agents until all the samples are correctly classified or else the program terminates because there are no remaining variables which can add to discrimination (42, 43, 45). When Levitt applied SDA to the 13 objective tests, he was able to classify all the gel specimens he had into the proper catagory by the fourth step.

Aishima et al. (45) classified 160

specimens of soy sauce, based on 29 GLC peaks, into the 20 brands the samples represented. sample.

They misclassified only one

In one of our early studies (38) we classified correct-

ly 70 samples of Coca-Cola and Pepsi-Cola blended in mixture

116 Table 3.

Order of inclusion of first eight out of 22 sensory attributes to classify eight gels into separate categories, adapted from Levitt (1974).

Step

Variable Included

% Misclassified

1 2

K K+S

53 40

3 4

K+S+M

26

K+S+M+A

20 18

5 6

K+S+M+A+B K+S+M+A+B+O

7 8

K+S+M+A+B+O+U K+S+M+A+B+O+U+C

U

Statistic F 108 . 3 64.4

7/164 14/326

49.9

21/466

36.4 29.4

28/582

0.017 0.014

25.0

0.013

19.1

42/749 49/806 56/851

0.178 0.071 0.042 0.026 0.020

12 12

df

21.7

35/675

ratios of 80/20, 67/33, 60/40, 50/50, 33/67 and 30/70, but we had to use 39 steps to do so. In contrast, the sensory panel could not always resolve differences between adjoining mixtures. Often, neither the sensory nor the objective tests alone are sufficient.

By combining sensory and objective tests, a

further degree of resolution sometimes can be attained (47). Once SDA has been used to classify samples and to rank variables according to their efficiency as discriminators, one probably will want to turn to multiple DA to calculate a discriminant function: Z = A^Xi + X2X2

*kxk

where the As are the weighting values and the Xs are the measurement values (38, 40).

Upon substituting in measurement

values for the new samples and predicting the class to which they belong, one in effect is predicting their sensory identity. Ancillary correlative methods.

There are several other corre-

lation methods of value in analyzing sensory or objective data.

117

Canonical analysis allows the geometric location of samples to be calculated and depicted (Figures 5 and 6). Principal component analysis (PCA), cluster analysis (CA) and multidimensional scaling (MDS) can be used for the same purpose (48, 49, 50). Mention needs to be made of differences among correlative techniques. DA and canonical analysis are external methods of analysis. The correlation is between data sets. PCA, CA, MDS and factor analysis (FAO are internal methods of analysis. Here, correlations are among data within the same data set. Some statisticians, on that account, frown upon placing too much reliance on the results obtained on the ground that separation into factors or components is likely to be parochial to the particular set of assessors (44) or samples evaluated. There are some further distinctions between the correlation methods. PCA and CA merely show that certain things or events happen to be together. There may not be any real relation among members of the component. FA, on the other hand, is an attempt to discern underlying order or structure amongst variables (51), and members of the same factor are assumed to be at least partially related to the other members. The chief function of PCA and Ca, and one function of FA, is to reduce the number of variables which have to be dealt with to a lesser number of components. Aishima et al. (46), for example, applied PCA to the 29 GLC peaks they used as discriminating variables. They derived eight components; then in turn they showed that these eight components were correlated with sensory quality. Lyon (52) compressed description of 12 attributes to four factors. The compressing of variables into factors, clusters or components is really nothing new either to chemists or sensory analysts. Chemists, for example, are not at all surprised that fatty acids from C4 to Cs have certain chemical properties in common, nor are sensory experts surprised that the fatty acids differ in their total sensory effect but that they also have certain notes in common. The example given is an illustration of FA, but PCA, CA and FA all

118 have the common objective of reducing the number of variables to a lesser number of components or factors. Practices The chief use we make of these correlative methods is to group assessors into homogeneous subsets (50, 53, 54, 55) . So as to segregate experimental error within the panel, the correlations between assessors should invariably be examined before proceeding further with analysis. Table 4 shows the extent to which results differed when two homogeneous subsets of 13 and 8 assessors were formed from amongst 27 assessors evaluating cooked, snap beans. Figure 7 shows partitioning of a panel into homogeneous subsets (56). (For as few as 12 assessors, we normally would visually classify assessors from the correlation coefficients themselves, but CA was applied here so as to be able to point out later other advantages to examining the panel for differences in scoring). Table 5 shows the success one of the panel subgroups had in classifying three cheese products. Product No. 1 was a natural cheese; products 2 and 3 were formulated (imitation) cheese. The assessors could not unequivocally distinguish between the natural cheese and imitation cheese No. 3. When the panel of 12 assessors was not segregated into homogeneous subsets, their performance was poorer. Success of classification averaged 71.7%. Pitfalls While multivariate correlation methods have proved to be very useful for the resolution of sample differences, there are some pitfalls which need to be avoided.

One is placing too great a

reliance on seemingly significant correlation coefficients. Where there are a large number of correlations being examined, a certain percentage will appear to be significant but in fact

119

Table 4.

Mean scores assigned by two subsets of assessors. Scores Assiqned to Canned Products — —•— Accept. Appear. Color Flavor Mouthfeel

Subset No. 1, 13 assessors

5 .6

5.5

5.0

5.6

5.5

No. 2, 8 assessors

4 .9

4.7

4.4

4.8

5.0

Scores Assigned •to Frozen Products NO. 1

4 .5

5.4

5.4

4.4

4.7

No. 2

6 .2

6.5

6.7

6.0

6.0

Clustering Stage

1

7

2

4

Assessors 5 10 12 8

3

6

9

11

1

X

X

X

X

X

X

X

X

X

X

X

X

2

X

X

X

X

X

X

X

X

X

X

X

X

3

X

X

X

X

X

X

X

X

Figure 7.

Table 5.

Clustering of 12 assessors into three subsets according to their responses in evaluating the intensity of 15 attributes of cheese. Classification of three cheese products based on sensory and discriminant analysis.

Actual Product

Predicted 1

1

2

Success % 3

12

1

2

80

2

0

3

2

15

0

100

0

13

86.7 88.8

120 be simply chance co-fluctuation between variables. Persson et al. (57) have cautioned that correlation coefficients must be exceptionally high and occur regularly before much stock should be put in them. They stated that experimenters should conduct experiments with new samples, preferably some time apart and possibly with different assessors to learn whether the correlation is a general one or is singular to a particular data set. This is the procedure we followed in conducting one of our studies (50). Degrees of freedom. Another problem in carrying on multivariate analysis is that one has to have sufficient degrees of freedom (Df) to go through several SDA steps or to have matrices reasonably symmetrical so that they can be inverted (45). Table 6 illustrates the first of these problems. In this instance, the experimenters wanted to have only three replications to resolve differences amongst six brands of blue cheese. Out of 18 Df, only 12 were left for the step process. Very fortunately for the investigators, they could resolve sample differences with 12 sensory terms. Had they had to go to the 13th step, Table 6.

Step 1 2 3 4 5 6 7 8 9 10 11 12

Sensory terms permitting resolution of differences among six brands of blue cheese, at 12th step Df = 5/1. Term

F-value

Veiny, distinctness and intensity Acceptability Whiteness Curd appearance Pastiness Yellowness Crumbliness Sharpness Soapiness Moldiness Appearance Flavor

100.9 84.6 49.8 15.1 11.3 6.5 7.3 4.4 4.3 58 .6 14.6 5.7

121

the program would have terminated for lack of Df.

Ample degrees

of freedom have to be provided since it is not known in advance how many steps will be needed to resolve sample differences. Co-variates. Somewhat disconcerting, though not a problem, is the fact that two or more variables may be highly correlated with sample differences and also with each other (43, 54). The function of SDA and DA is to search for descriptors which add to discrimination, not merely provide redundancy. Note in Table 6 that "moldiness" had a high F value but it was not incorporated into the step process until the 10th step. Distinctness of the mold veins and moldiness in general both reflect the amount of mold present. They were correlated with each other. Veininess had the highest F value; it was used as a discriminator first. Once it was selected, other variables having lower F values than moldiness were selected because they provided new information whereas moldiness merely re-affirmed part of the information already contained in veininess. Unlike the situation in Table 6 where the best, single discriminator was retained, sometimes a variable already selected may be deleted later on in the step process. Table 7 illustrates this type of occurrence. A variable may be displaced by some other variable because the latter—in combination with the other variables selected—provides more powerful discrimination. Related to the combination of circumstances just described is the experience of Galetto and Bednarczyk (1) who found that the best equation for relating chemical composition to sensory quality did not include GLC peaks (compounds) they thought would be likely prospects for inclusion in a synthetic oil of 19 onion. They would have had to examine 5.1 x 10 formulae if they were to examine all possible combinations based on statistical considerations alone. Having knowledge of the literature and their own experience, they were able to select 47 combinations to include peaks they thought should be considered; from

122 Table 7.

Order of inclusion and deletion of variables by two subsets of cheese assessors. Subset 1

Steps

Variable

1

Firmness Cheesiness

2

F Value 37.0 6.5

Subset 2 Approx. F Value

Variable

F Value

Approx. F Value

37.0

Dryness

32 .1

32 .1

Saltiness

10 . 8

17.4

Chewiness

2.6

16.3

4.6

13.6

3 4

Springiness Caramellike

6.2

19.1 15 . 8

8.6

15.3

Sweetness

5

Cheddarlike

3.5

13.6

Firmness

15 . 3

13.9

6

Cheesiness deleted Aftertaste

16.5

16.5

Nuttylike Acidulous

8.7

12.7

3.1

12 .5 13.6

7 8

4 .1

14. 7

Saltiness deleted

the resulting regression functions they were able to deduce which compounds were essential to the flavor of oil of onion. From a correlation point of view there is only one equation which is the most discriminating one. Often, however, there are several others nearly as efficient. The investigator may choose to force in a variable (test) he or she wants in the discriminant function by forcing out a better test which is correlated with the one wanted. An example of this might be the substitution of a test, simplier to run, cheaper or less dangerous than the more efficient test. The benefits can be greater than the slight loss of resolving power. Principles and Pathways A considerable amount of discussion has been devoted to the discerning of correlations.

This was done because all the problems

existing as to the finding of correlations also apply to the

123

determining of causal relations; moreover, in most cases, correlation methods need to be employed before one is ready to seek out causal relations. The determination of correlations normally goes as follows: 1. Develop a list of descriptors cte novo or use glossaries; evaluate the terms by discussion, sniffing of GLC peaks, use of reference compounds or other substances; carry on preliminary trials; refine the list by applying one-way analysis of variance (ANOVA) to discard descriptors which are not also discriminators . 2. Carry on the main trial, preferably using reference compounds or substances and some form of descriptive analysis (28, 58). 3. Re-examine the assessors after the trial for efficiency, eliminate those who are not acceptable (58), then apply one-way ANOVA to the descriptors to ascertain which descriptors are only that, not discriminators. 4. Carry on concurrently chemical, physical, spectral and mechanical tests as appropriate. 5. Apply SDA to sensory and to objective data to select useful discriminators. If 100% classification is obtained, discard the variables not needed, even if they are significant; it is wasteful of effort to include variables not needed for classification, but if the establishing causal relations is the ultimate objective, all significant discriminators should be retained. 6. Often, neither sensory nor objective data will be 100% efficient in classifying samples. If not, combine sensory and objective data sets to ascertain if joint use of the two types enables greater classification to be attained (45, 47). 7. Carry on trials apart in time and with different samples to seek out variables universally correlated with quality differences, not those correlated only with a particular trial (50, 57).

124

The procedures above lead to the selection of variables useful for quality assurance and other technological functions not depending upon identity of the compounds or the discerning of causal relations. Establishing causal relations 1. Nowadays, chemical identification probably will have been done during the correlation stage because gas chromatographs and mass spectrometers (MS) are so frequently coupled. If not, identification needs to be carried out using the DA as a guide to which compounds should be identified first. Sniffing of the GLC effluents and collection of the peaks in bands (60) also should give clues as to compounds which are of sensory importance. 2. Factorial, response surface and other appropriate statistical designs should be considered for adoption so that compounds can be tested singly and in combination with other compounds at different concentrations to permit statistical analysis for main treatment effects and interaction effects (3, 61, 62, 63) . At the correlation stage the investigator was seeking to form classes from individual cases. Now the investigator has a class (products, brands, or classes established by the sensory analysis); the objective is to impose treatments upon the class to learn whether the treatments bring about changes in sensory quality or attributes. Appropriate statistical methods are ANOVA, co-variance (64) and regression analysis (RA). Stepwise regression analysis (SRA) had been frequently used by investigators to seek out regression effects and thus correlation. Provided the dependent variable (sensory quality or intensity) is determined by the concentrations of the independent variables, the chemical compounds, RA is an appropriate method of analysis. Qvist et al. (2) pointed out that DA is generally preferred to RA to detect correlations because in RA

125

the dependent variable has to be continuous and normally distributed or, in other words, to have additivity properties. They considered the last criterion as being unlikely for the kinds of complex odor qualities generally investigated. None of the criteria above apply to DA. RA has been used with apparent success by many investigators (3, 65, 66, 67) 3.

Concurrent sensory tests must be made to detect and

quantify sensory change.

Co-variance or multiple regression

methods are appropriate to learn whether the dependent variable (sensory response) changes in direction and magnitude as the independent variables (chemicals and concentrations) are changed. For both the chemical and sensory stages, there is no problem if compounds have been identified and they are commercially available, and they are being applied at levels above the normal level in the product under test. When the experimenter wants to test levels below the normal level in the food, this may present a difficulty. Volatiles can be stripped (68) and then added back at different levels, but procedures for removing non-volatiles are not as easy to find. Sometimes the problem can be overcome by blending of products high and low in the particular non-volatiles being investigated. 4. As at the correlation stage, replicate trials will need to be made to be sure that effects observed are general ones, not happenstance occurrences. The investigator seeking to establish causal relations should peruse publications of others (1, 2, 3) who have had the same objectives. Unfortunately there are very few publications to serve as models. One of the merits of the publication of Spencer et al. (3) is that they considered the effect both of volatiles and non-volatiles. Spencer et al. (3) in many ways took on a larger task than did Galetto and Bednarczyk (1) though the latter investigators were successful in their endeavor. The latter had to deal only with volatiles. Not only

126 are the non-volatiles important in their own right in many products, but there often is interaction (chemical and perceptual) between and within the non-volatiles and the volatiles (69, 70, 71). This makes the relating of composition to sensory character several times more difficult. 5.

Multiple regression analysis needs to be applied to

the sensory determinations to ascertain the importance of the main sense modalities in determining acceptance (42, 53, 55) and to ascertain the importance of each attribute within each sense modality and to relate sensory and chemical data sets. Potential Although the problems are great, several developments have taken place which make the prospect of being able to establish causal relations much better than they were only a few years ago. The development of more effective methods has been much like the formation of a giant river. When we first commenced our studies in the early 196 0s, we could utilize four fundamental tools developed by others. They were: DA as devised by Fisher (30), the coupling of sensory and statistical methods as a part of the same procedure, credit for which probably should go to Bengtsson (72) , gas chromatography and high-speed computing;GLC enabled many volatiles to be separated concurrently unlike prior methods of analysis, and high-speed computing provided the capacity to deal with large data sets. Since that time, other tributaries—in fact rivers in their own right— have joined in to add greater power to methods available. Highperformance liquid chromatography (HPLG), enabling non-volatiles to be separated en masse, is increasingly filling a crucial gap. Since perception of volatiles and non-volatiles are so interrelated, one can hardly establish causal relations without analyzing for both types of compounds. Exceedingly important has been the interfacing of GLC and HPLC with mass spectrometry. Mussinan (73) has commented upon the role interfacing has had

127

in the advancement of analytical chemistry and flavor creation. Equally important have been advances in methods of sensory analysis and in the flexibility and ease of use of statistical programs written for computer calculations. While not perfect, the old and the newer tools now available make the establishing of causal relations practical today whereas not too many years ago the task would have been overwhelming. Since there is a good prospect that considerable success can be achieved, the effort needs to be made for various reasons. Quality assurance programs would unquestionably be more efficient if tests were for determinants of sensory difference or quality rather than correlates which may or may not be directly related to sensory attributes. Tighter specifications could be set for the procurement of raw ingredients. Process changes and innovations could be specified with greater particularity so as to bring about desired sensory change. Of greatest importance in the long run is the fact that once progress begins to be made in establishing determinate relations between chemical composition and sensory response, then we will begin to have a better understanding of the processes which go on between perception and response. This is not a task for food scientists and technologists alone. The efforts of biochemists, physiologists and behavioral scientists—to name a few—will be needed. However, just as applications of multivariate correlation methods to sensory/objective problems were initiated largely by technologists and other practitioners who had problems to solve, the liklihood is that the same kinds of individuals will have to be in the forefront in establishing causal relations. Acknowledgement The preparation of this paper was supported by State tax funds allocated to the Georgia Agricultural Experiment Stations.

128 References 1.

Galetto, W. G., Bednarczyk, A. A.: J. Food Sei. 40^ 11651167 (1975).

2.

Qvist, I. H., von Sydow, E. C. F., Akesson, C. A.: Lebensm.Wiss. u. Technol. 9, 299-303 (1976).

3.

Spencer, M. D., Pangborn, R. M., Jennings, W. G.: J. Agric. Fd. Chem. 26_, 725-732 (1978) . Hoff, J. T., Helbert, J. R., Chicoye, E.: Tech. Quart., Master Brewers Association of the Americas 12, 209-213 (1975). Biggers, R. E., Hilton, J. J., Gianturco, M. A.: J. Chromat. Sei. 7, 453-472 (1969).

4. 5. 6. 7. 8. 9. 10.

Gianturco, M. A., Biggers, R. E., Ridling, B. H.: J. Agric. Fd. Chem. Z2, 758-764 (1974). Meilgaard, M. C.: Beer Flavour, in: "Brewing Science", Vol. 3, Pollock, J. R. A. (ed.), Academic Press, London (in press). Schreier, P.: CRC Critical Rev. Food Sei. Nutr. 12, 59-111 (1979) . Schreier, P., Drawert, F., Schmid, M.: J. Sei. Fd. Agric. 29, 728-736 (1978). ter Heide, R., de Valois, P. J., Visser, J., Jaegers, P. P., Timmer, R.: Concentration and Identification of Trace Constituents in Alcoholic Beverages, in: "Analysis of Foods and Beverages--Headspace Techniques", Charalambous, G. (ed.), Academic Press, New York, pp. 249-281 (1978).

11.

Takahaski, K., Ohba, T., Takagi, M., Sato, S., Namba, Y.: Hakkokogaku 57, 148-157 (1979) .

12.

Shallenberger, R. S., Acree, T. E.: Nature 216, 480-482 (1967) .

13.

Guigoz, Y., Solms, J.: Chemical Senses and Flavour 2, 71-84 (1976) .

14. 15.

Gardner, R. J.: Chemical Senses and Flavour 4, 275-286 (1979). Tancredi, T., Lelj, F., Temussi, P. A.: Chemical Senses and Flavour 4, 259-265 (1979).

16.

Boelens, H.: Molecular Structure and Olfactive Properties, in: "Structure-Activity Relationships in Chemoreception", Benz, G. (ed.), Information Retrieval Limited, London, p. 197-206 (1976).

17.

Ohloff, G.: Stereochemistry-Activity Relationships in Human Odor Sensation: The Trixial Rule, in: "Olfaction and Taste VII", Starre, H. van der (ed.), IRL Press Ltd., London, p. 3-12 (1980) .

129

18.

19.

King, B. M., Solms, J.: Air-Water Partitioning of Flavor Compounds in Systems Containing Purines, in: "Olfaction and Taste VII", Starre, H. van der (ed.), IRL Press Ltd., London, pp. 23-26 (1980). Harper, R., Land, D. G., Griffiths, N. M., Bate-Smith, E. C.: Br. J. Psychol. 59_, 231-252 (1968).

20.

Dravnieks, A., Bock, F. C., Powers, J. J., Tibbetts, M., Ford, M.: Chemical Senses and Flavour 3 1 9 1 - 2 2 5 (1978).

21.

Clapperton, J. F., Dalgleish, C. E, Meilgaard, M. C.: J. Inst. Brew. 82, 7-13 (1976) . Meilgaard, M. C., Dalgliesh, C. E., Clapperton, J. F.: J. Inst. Brew. £5, 38-42 (1979).

22. 23. 24. 25.

El1 Ode, K. E., Dornseifer, T. P., Keith, E. S., Powers, J. J.: J. Food Sei. 31, 351-358 (1966). Williams, A. A.: J. Sei. Fd. Agric. 2j5 , 567-585 (1975). Harries, J. M.: J. Sei. Fd. Agric. 2£, 1571-1581 (1973).

26.

Palmer, D. H.: J. Sei. Agric. 25, 153-164 (1974).

27.

Vuataz, L., Sotek, Rahim, H. M.: Proceedings IV Intern'1. Congress Food Sei. and Technol. Ia:25 (1974).

28.

Vuataz, L.: Nestle Research News, 76/77, Tech. Documentation Center, Lausanne, Switzerland, pp. 57-71.

29.

Williams, A. A.: International Flavours and Food Additives, Vols. 9 and 10, 131-133 (1978-1979). Fisher, R. A.: Annals of Eugenics (London) VII, 179-188 (1936) .

30. 31.

Powers, J. J., Keith, E. S.: Abstracts of papers, 2nd International Congress of Food Sei. and Technol., Warsaw, p. 439 (1966).

32.

Powers, J. J., Keith, E. S.: J. Food Sei. 25_, 224-228 (1968) .

33.

Schreier, P., Reiner, L.: J. Sei. Fd. Agric. 30^ 319-327 (1979). Brown, D. G. W., Clappeton, J. F.: J. Inst. Brew. 84, 318323 (1978) .

34. 35. 36.

Noble, A. C., Flath, R. A., Forrey, R. R.: J. Agric. Fd. Chem. 28, 346-353 (1980). Flora, L. F., Wiley, R. C., Moon, L. C.: J. Food Qual. 1, 341-348 (1977).

37.

Dravnieks, A., Reilich, H. G., Whitfield, J., Watson, G. A.: J. Food Sei. 3£, 34-39 (1973).

38.

Young, L. L., Bargmann, R. E., Powers, J. J.: J. Food Sei. 35:219-223 (1970).

130 39. 40. 41. 42.

43.

44.

Milutinovic, L. , Bargmann, R. E., Chang, K., Chastain, M., Powers, J. J.: J. Food Sei. 35, 224-228 (1970). Powers, J. J.; Quinlan, M. C.: Lebensm.-Wiss. u. Technol. 6:209-214 (1973). Powers, J. J. : Food Technol. 22^, 39-44 (1968). Powers, J. J.: Experiences with Subjective/Objective Correlation, in: "Correlating Sensory Objective Measurements— New Methods for Answering Old Problems", Powers, J. J. and Moskowitz, H. R. (eds.), STP 594, Am. Soc. for Testing and Materials, Philadelphia, PA, pp. 111-122 (1976). Powers, J. J.: Correlation of Sensory Evaluation Data with Objective Test Data, in: "Sensory Evaluation Methods for the Practicing Food Technologist", Johnston, M. R. (ed.), Institute of Food Technologists, Chicago, IL, USA, pp. 91 to 9-35 (1979). Levitt, D. J.: J. Texture Studies 5, 183-200 (1974).

45.

Powers, J. J.: Measurement of Relative Contribution of Constituents to Flavor Composition, in: Proceedings, 3rd International Congress of Food Sei. and Technol., Washington., D.C., Institute of Food Technologists, Chicago, IL, USA, pp. 395-402 (1970).

46.

Aishima, T., Nagasawa, J., Fukushima, D.: J. Food Sei. 44, 1723-1727, 1731 (1979) . Powers, J. J., Smit, C.J.B., Godwin, D. R.: Lebensm.-Wiss. u. Technol. 11, 275-278 (1978).

47. 48.

Lisle, D. B., Richards, C. P., Wardleworth, D. F.: J. Inst. Brew. 84, 93-96 (1978).

49.

Clapperton, J. F., Piggott, J. R.: J. Inst. Brew. £5, 271274 (1979).

50.

Godwin, D. R., Bargmann, R. E., Powers, J. J.: J. Food Sei. 43, 1229-1234 (1978) . Harper, R.: Applied Statistics 5, 32-48 (1956). Lyon, B. G.: J. Food Sei. £5, 1341-1346 (1980).

51. 52. 53. 54.

Powers, J. J.: Validation of Sensory and Instrumental Analyses for Flavor, in: Proceedings, IV International Congress of Food Sei. and Technol. 11, 173-182 (1974). Powers, J. J.: Tech. Quarterly, Master Brewers Assocn. of the Americas 18(1) , 11-21 (1981) .

55.

Powers, J. J.; Quinlan, M. C.: J. Agric. Fd. Chem. 22, 744-749 (1974).

56.

Choi, I. K.: Ph.D. Dissertation, University of Georgia, Athens, GA, USA (1981).

57.

Persson, T. and von Sydow, E.: J. Food Sei. 37_, 234-239 (1972) .

131

58. 59. 60. 61. 62. 63. 64.

Stone, H., Sidel, J., Oliver, S., Woolsey, A., Singleton, R. C.: Food Technol. 29, 24-34 (1974). Wu, L. S., Bargmann, R. E., Powers, J. J.: J. Food Sci. 42, 944-952 (1977). Parliment, T. B., Scarpellino, R.: J. Agric. Food Chem. 25, 97-112 (1977). Meilgaard, M. C., Elizondo, A., Mackinney, A.: Wallerstein Lab. Communications XXXIV (No. 114), 95-110 (1971). Persson, T. , von Sydow, E., Akesson, C.: J. Food Sci. 38, 682-689 (1973). Shaw, P. E., Ahmed, E. M., Dennison, R. A.: Proc. Int. Soc. Citriculture 3, 807-840 (1977). Kosaric, N., Duong, T. B., Svrcek, W. Y.: J. Food Sci. 38, 369-373 (1973).

65.

Bouton, P. E., Ford, A. L., Harris, P. V. , Ratcliff, D.: J. Texture Studies 7, 315-328 (1975).

66.

Jackson, M. G., Timberlake, C. F., Bridle, P., Vallis, L.: J. Sci. Fd. Agric. 29^, 715-727 (1978).

67.

Simon, P. W., Peterson, C. E., Lindsay, R. C.: J. Agric. Fd. Chem. 28, 559-561 (1980).

68.

von Sydow, E., Anderson, J., Anjou, K., Karlsson, G., Land, D. , Griffiths, N.: Lebensm.-Wiss. u. Technol. 2' 11-17 (1970) .

69.

King, B. M., Solms, J.: J. Agric. Food Chem. 21_, 1331-1334, (1979) . Beyeler, M. , Solms, J.: Lebensm.-Wiss. u. Technol. 1_, 217219 (1974).

70. 71.

Hardwick, W. A.: Interaction Flavor Influence of Some Materials in Different Foods and Beverages, in: "Flavor: Its Chemical, Behavioral, and Commercial Aspects", Apt, C. M. (ed.), Westview Press, Boulder, CO, USA, pp. 67-78, (1978) .

72.

Bengtsson, K.: Svenska Bryggareforeningens Manadsblad 58, 59-71, 102-111, 149-157 (1943). Mussinan, C. J.: The Role of Analytical Chemistry in Flavor Creation, in: "Flavor: Its Chemical, Behavioral, and Commercial Aspects", Ibid, pp. 13-30 (1978).

73.

SOME PROPERTIES OF ODORIFEROUS MOLECULES

R. Teranishi, R. G. Buttery, and D. G. Guadagni U.S. Department of Agriculture, Western Regional Research Center, 800 Buchanan Street, Albany, California 94710, USA

Introduction In 1958, Dr. C. Weurman visited me at the Western Regional Research Center, Albany, California, to discuss flavor chemistry research.

In

subsequent years, we had many interesting discussions, and my respect and admiration for his scholarly approach increased.

I am very pleased

and honored to be here at this Weurman Symposium and to have this opportunity to dedicate this paper to Dr. Weurman. Property is defined as an effect that a material object or substance has on another object or on one or more of the senses of an observer. In chemistry, sane of the properties of compounds that must be considered are:

1. Physical- melting point, boiling point, density,

solubilities, etc.

2.

Spectral- ultraviolet, infrared, Raman, nuclear

magnetic resonance, mass spectrometry, etc.

3. Chemical - reactivity,

what kind of bonds are made and/or broken, how much and hew fast. These properties are utilized to characterize compounds.

It is assumed

that a compound is pure when there is a constancy of properties even with continued efforts for further purification. In the biological sciences, we are concerned with complex living systems

© 1981 by Walter de Gruyter &. Co, Berlin • New York Flavour '81

134

in which properties are not as easily determined as in the physical sciences.

A branch of mathematics has been developed to handle

variabilities found in biological systems so that with systematic compilation of instances, there can be an inference of general truths. Odor threshold The most important biological property of an odorant is that it has an odor.

The effectiveness, or potency, of this compound must be firmly

established, just as with the effectiveness of a drug.

It seems reason-

able to determine this property of odor effectiveness as the odor threshold, the amount of the chemical stimuli needed to produce olfactory responses, the detection odor threshold. If a molecule has an odor, it has a characteristic or distinctive trait, or quality.

A description of the characteristic quality is very diffi-

cult, but the amount necessary for recognition can be established, the recognition threshold. In order to classify data, it is necessary to establish definitions and terminology.

Various terms, such as detection, difference, recognition,

supra-, absolute, and terminal, used with odor thresholds are defined by Stahl (1).

We will be concerned predominantly with detection odor thresh-

holds and will be referring to detection threshold when we use the term "threshold". The concept of odor threshold is not new.

One of the early workers,

referred to in "Compilation of Odour Threshold Values in Air and Water", edited by van Gemert and Nettenbreijer (2), is Passy, whose work was done

135 in the 1890's (3). Odor unit In 1957, Patton and Josephson (4) proposed a method for determining significance of volatile flavor compounds in foods.

In this method,

concentration of a compound in a food is determined chemically, and the flavor threshold value is determined psychcmetrically, as illustrated in Table 1. Table 1 FLAVOR THRESHOLD AND QUALITY (4) Compound

Medium

Threshold (ppti)

Quality

Butyric acid

Hcmog. milk

Methional

Skim milk

0.05

broth-like

Methyl sulfide

Water

0.012

malty, cowy

Methyl mercaptan

Water

0.002

putrid, cooked cabbage

25

rancid milk

Compounds in excess of their threshold are considered to make contributions to flavor, whereas those below their thresholds are thought to have little or no effect.

This intuitively acceptable relationship has been

expressed mathematically and referred to as '• aroma value1' by Rothe and Thanas (5) in 1963 and as ''odor unit" by Guadagni et al. (6) in 1966. In the article by Guadagni et al. (6), it is clearly stated that the definition and calculation of odor units are strictly arbitrary, that the validity of the procedure depends on the fact that these and other

136 studies (7-9) indicate that at least certain mixtures of odors are additive at threshold levels.

In this article (6), it is also clearly stated

that the ratio of concentration and threshold gives the number of odor units that a specific component or fraction contributes to the total odor of the mixture, but the ratio says nothing about the odor quality of the final mixture nor does it imply anything about the relationship between stimulus concentration and intensity of sensation above threshold.

Obviously, extrapolation beyond regions which any relationship is

intended may lead to erroneous conclusions.

Nevertheless, when used as

intended, within limits specified, odor units can provide a logical approach in odor research problems.

A.

Hop oil

In the study of odor intensitites of hop oil (6), hydrocarbon components were shown to contribute considerably to the total odor, Table 2.

Table 2 RELATIONSHIPS AMONG HOP OIL, FRACTIONS, THRESHOLDS, AND ODOR UNITS (6) Material

Whole Oil

%

Threshold (Tc, X 10 9 )

Od§r Units (UQ, 10~ 6 )

% of Total Odor Units

100

12

83.3

100

Hydrocarbon

86

15

57.3

69

Oxygenated

14

5

28.0

34

*

U

_ o ~

F c , fraction of whole oil in ppb T c , threshold in ppb

This study indicates that the contribution that hydrocarbons make to

137 odors cannot be ignored as previously assuned.

Detailed study of the

oxygenated canpounds showed that two compounds, methyl thiohexanoate and methyl dec-4-enoate, accounted for more than half of the contribution made by 20 components of oxygenated compounds.

With the addition of five

more canpounds, about 90% of the odor contributed by 20 oxygenated canpounds was accounted for.

Thus, even though a large number of canpounds

make up the oxygenated traction, a relatively small number of these components appear to account for the major portion of the odor of this fraction.

With the use of odor units to gather information and to guide

formulation, an oil was constituted fran 27 synthetic chemicals that was very closely similar to an oil isolated from hops as judged in sensory panel studies. B.

Bay oils

The volatile oil from the leaves of the California bay tree (Umbellularia californica) was studied and compared to the oils fran Mediterranean bay, laurel (Laurus nobilis) and fran West Indian bay (Pimenta racemosa formerly Myrcia acris) (8). Table 3 shows part of the data obtained on the California bay oil Table 3 CALIFORNIA BAY OIL (8) Compound

Rei. % in Oil

Odor Units (C/T X 10" 6 )

Sabinene

6

0.8

% Odor Units of Whole Oil 0.5

1,8-Cineole

19

Umbellulone

39

0.5

0.3

5

0.8

0.5

Methyleugenol

146

95

138 Although umbellulone is by far the major constituent, it contributes little to the total odor.

However, 1,8-cineole, though half the concen-

tration of umbellulone, contributes to about 95% of the total odor. Table 4 shows sane data obtained on the components of Mediterranean bay oil. Table 4 MEDITERRANEAN BAY OIL (8) Compound

Rel. % in Oil

Odor Units (C/T X 10~ 6 )

% Odor Units of Whole Oil

a-pinene

12

20

5

1,8-Cineole

30

230

58

Linalool

11

18

2

3

Eugenol

4.5 0.75

In this case, 1,8-cineole is the major constituent and does contribute to most of the odor of this oil, about 58%. Table 5 shows sane of the data obtained on the West Indian bay oil. Table 5 WEST INDIAN BAY OIL (8) Compound

Rel. % in Oil

Myrcene

21

1,8-Cineole Eugenol Oct-l-en-3-ol

Odor Units (C/T X 10~ 6 ) 16

% Odor Units of Whole Oil 5.7

9.7

75

26

33.8

56

20

1.6

12

4.2

139 In this oil, eugenol is the major constituent and does contribute considerably, 20%, but 1,8-cineole still contributes substantially, 26%, to the total odor. Probably one of the reasons why California bay leaves are used as a spice in a similar manner as Mediterranean bay leaves is that in both 1,8cineole plays a large role.

Table 6 shows seme compounds cannon to the

three bay oils (8). Table 6 COMPOUNDS COMMON IN BAY OILS (8) Compounds

Calif.

Medit.

19%

30%

5

4

1,8-Cineole Methyl eugenol Eugenol

West Indian 10% 0.2

2

34

0.3

11

3

Terpinyl acetate

< 0.1

10

—

Umbellulone

39

—

—

—

Linalool

The arena impact of 1,8-cineole in Mediterranean bay oil is softened with the contribution of linalool.

West Indian bay oil is quite different

from both California and Mediterranean bay oils, but 1,8-cineole still plays a substantial role. Most of the California bay oil odor is accounted for, but approximately 30-40% of the odor of the Mediterranean and West bay oils is still unaccounted for in our studies.

Relatively minor components, not yet

characterized and with low thresholds, could account for this discrepancy.

140 C.

Rose oil

Ohloff has used "odor values" to explain the importance of minor components in flavors and fragrances (9).

In the example of Bulgarian rose

oil, 9 constituents account for 86% of the total oil, but a mixture of these do not reproduce the odor pattern of rose oil at all.

Seven

constituents amounting to about 1% of the essential oil are decisive for the development of rose fragance.

Although B-ionone is 200 times lower

in concentration than that of nerol, g-ionone has an odor value approximately 200 times higher than nerol because of the extremely low threshold of B-ionone. Thresholds and Characteristic Odors Use of odor units, or odor values, can be used to indicate whether most of the total odor is accounted for or not, as shown by the above examples. Some relationships between thresholds and characteristics were discussed previously (10).

Although nootkatone has the very characteristic odor of

grapefruit, it has a relatively high threshold.

However, compounds that

contribute to characteristic odors generally have low thresholds, such as 2-methoxy-3-isobutylpyrazine. The threshold of B-ionone was the lowest threshold that we had measured by 1968 (11).

Since that time, 2-methoxy-3-isobutylpyrazine (12) and the

compound responsible for the thiamine off-odor (13) have been added to our list of thresholds, shown in Table 7.

141

Table 7 Odorant

yg/liter of water

Ethanol

100,000

Nootkatone

170

Myrcene

15

n-Amyl acetate

5

a-and g-Sinensal

0.05

ß-Ionone

0.007

2-Methoxy-3-isobutylpyrazine

0.002

Thiamine odor compound

0.0004

Surely, compounds with lower thresholds will be added to the list as very effective odorants are identified. We are not unaware that olfaction is a very complicated phenomenon. real life, rarely do we encounter a single odorant.

In

Usually, we inte-

grate a multitude of sensations from a multitude of stimuli to form a composite sensation.

In most foods we are anelling compounds near their

thresholds, whereas with perfumes, we are dealing with much more concentrated materials and with constituents which may be well over the threshold concentrations.

As concentrations exceed the threshold values,

the sensations evoked may not be similar to those near thresholds.

Vapor

pressures vary considerably, several orders of magnitude, with different media. minutes.

Sane molecules are very unstable and decompose in a matter of Nevertheless, systems must be isolated tor simplification so

1 42 that we may begin to understand seme of the fundamental aspects of olfaction. In our hands, the concept of odor units, or odor values, has provided a logical and methodical approach to study characteristic odors in topics and problems such as hops (6), bay oils (8), apples (11), bell peppers (12), carrots (14), potato chips (15), volatile components of jute sacks (16), cabbage, broccoli, and cauliflower (17), artichokes (18), offensive odor compounds from potato processing wastes (19, 20), etc.

References 1.

Stahl, W. H.: Editor. Compilation of Odor and Taste Threshold Values Data, ASTM Data Series DS 48, American Society for Testing and Materials. Philadelphia 1973.

2.

van Gemert, L. J., and Nettenbreijer, A. H.: Editors. Compilation of Odour Threshold Values in Air and Water, National Institute for Water Supply, Voorburg, Netherlands, Central Institite for Nutrition and Food Research TNO, Zeist, Netherlands, 1977.

3.

Passy, J.:

4.

Patton, S., and Josephson, D. V.:

5.

Rothe, M., and Thcmas, B. : Z. (1963).

6.

Guadagni, D. G., Buttery, R. G., and Harris, J.: 17, 142-144 (1966).

7.

Guadagni, D. G., Buttery, R. G., Okano, S., and Burr, H. K.: Nature, Lond. 200, 1288-1289 (1963).

8.

Buttery, R. G., Black, D. R., Guadagni, D. G., Ling, L. C., Connolly, G., and Teranishi, R.: J. Agric. Food Chem. 22(5), 773-777 (1974).

9.

Ohloff, G.: The Importance of Minor Components in Flavors and Fragrances, in Proceedings of VII International Congress of Essential Oils, Oct 7-11, 1977, Kyoto, Japan, Japan Flavor and Fragrance Manufactures' Association, Kyoto 1977, pp. 69-74.

10.

Teranishi, R.: Odor and Molecular Structure, in Gustation and Olfaction, Ohloff, G., and Thcmas, A. F., Editors, Academic Press, London 1971, pp. 165-177.

C. R. Hebd.

Seances Acad. Sci. 114, 1140-1143 (1892). Food Res. 22, 316-318 (1957).

Lebensm.-Unters. Forsch 119, 302-310 J. Sci. Fd. Agric.

143 11. Guadagni, D. G.: Requirements for Coordination of Instrumental and Sensory Techniques, in Correlation of Subjective-Objective Methods in the Study of Odors and Taste, Special Technical Publication No. 440, Am. Soc. for Testing and Materials, Philadelphia, 1968, pp. 36-48. 12. Buttery, R. G., Seifert, R. M., Guadagni, D. G., and Ling, L. C.: J. Agric. Food Chem. r7(6), 1322-1327 (1969). 13. Buttery, R. G., Seifert, R. M., Turnbaugh, J. G., Guadagni, D. G., and Ling. L: J. Agric. Food Chem. 29, 183-185 (1981). 14. Buttery, R. G., Seifert, R. M., Guadagni, D. G., Black, D. R., and Ling, L: J. Agric. Food Chem. 16(6), 1009-1015 (1968). 15. Guadagni, D. G., Buttery, R. G., and TUrnbaugh, J. G.: J. Sci. Fd Agric. 23, 1435-1444 (1972). 16. Seifert, R. M., Buttery, R. G., and Guadagni, D. G.: J. Sci. Fd Agric. 26, 1839-1845 (1975). 17. Buttery, R. G., Guadagni, D. G., Ling, L., Seifert, R. M., and Lipton, W.: J. Agric. Food Chem. 24(4), 829-832 (1976). 18. Buttery, R. G., Guadagni, D. G., and Ling, L: J. Agric. Food Chem. 26(4), 791-793 (1978). 19. Buttery, R. G., Guadagni, D. G., and Garibaldi, J. A.: J. Agric. Food Chem. 27(3), 646-647 (1979). 20. Buttery, R. G., and Garibaldi, J. A.: Agric. Food Chem. 28(1), 158-159 (1980). 21. Meilgaard, M. C.: MBAA Tech. Quart. 12(3), 151-168 (1975). 22. Meilgaard, M. C., and Reid, D. S.: Determination of Personal and Group Thresholds and the Use of Magnitude Estimation in Beer Flavour Chemistry, in Progress in Flavour Research, Land, D. G., and Nursten, H. E., Editors, Applied Science Publishers, Ltd., Essex 1979, pp. 67-77. 23. Rothe, M.: Arcma Values - A Useful Concept?, in Proc. Int. Symp. Aroma Research, Zeist, 1975, Maarse, H., and Groenen, P. J., Editors, Centre for Agricultural Publishing and Doc orientation, Wageningen 1975, pp. 111-119.

P R E D I C T I V E V A L U E O F S E N S O R Y A N D A N A L Y T I C A L DATA DISTILLED

Paula

FOR

BEVERAGES

Jounela-Eriksson

R e s e a r c h L a b o r a t o r i e s of the S t a t e A l c o h o l M o n o p o l y Box 350, S F - 0 0 1 0 1 H e l s i n k i 10, F i n l a n d

(Alko),

INTRODUCTION T h e a r o m a of a l c o h o l i c b e v e r a g e s

is always a c o m b i n a t i o n of

various different components. Aroma compounds play a key

role

in d i s t i l l e d b e v e r a g e s , such as v o d k a s , w h i s k i e s ,

cognacs,

b r a n d i e s and rums, p a r t i c u l a r l y

sensations

t h r o u g h the a r o m a

p e r c e i v e d by the o l f a c t o r y sense and the more or less c o m p o u n d s p r o d u c i n g an e f f e c t u p o n

volatile

them.

A l t h o u g h the d i f f e r e n c e s b e t w e e n for instance w h i s k y ,

cognac

and rum can be e a s i l y d i s t i n g u i s h e d by t a s t i n g , the m a i n c o m p o u n d s are the same in these b e v e r a g e s d i f f e r e n t raw m a t e r i a l s used

in spite of

(1). D i s t i l l a t i o n

aroma

the

techniques

r e g u l a t e the q u a n t i t i e s of a r o m a c o m p o n e n t s , but the y e a s t

and

f e r m e n t a t i o n c o n d i t i o n s c h i e f l y d e t e r m i n e the m a i n

compounds.

In fact, t r e a t m e n t of raw m a t e r i a l s can a f f e c t the

occurrence

of a r o m a c o m p o u n d s

in d i s t i l l a t e s and t h r o u g h

it the a r o m a of

a b e v e r a g e , e.g. the use of p e a t e d malt as a raw m a t e r i a l Scotch malt whiskies Distillates

for

(2).

f r e q u e n t l y d e v e l o p their c h a r a c t e r i s t i c a r o m a

a g e i n g . A r o m a c o m p o u n d s are f o r m e d and d i s a p p e a r e d d u r i n g m a t u r i n g w h e n the c o r r e c t i n t e r r e l a t i o n s h i p s B a r r e l s a l s o release c o m p o u n d s a f f e c t i n g

© 1981 by Walter de Gruyter & Co, Berlin • New York Flavour '81

are

in the

achieved.

the b e v e r a g e

taste

146 and aroma. E v e n the m o s t d e v e l o p e d p r o c e s s i n g t e c h n i q u e s not d i s p l a y the i m p o r t a n c e of the s k i l f u l b l e n d i n g of lates of d i f f e r e n t ages and types, for instance and g u a r a n t e e i n g w h i s k y w i t h a c h a r a c t e r i s t i c T h e a r o m a c o n t e n t of d i s t i l l e d b e v e r a g e s

do

distil-

in p r o d u c i n g

aroma.

is today

well-known

(3-5). H o w e v e r , as the n u m b e r of i d e n t i f i e d c o m p o u n d s

grows,

determining

aroma

the m o s t e s s e n t i a l c o m p o u n d s

b e c o m e s more d i f f i c u l t . F u r t h e r m o r e , components beverages

is n a t u r a l

to p e r c e i v e d

a v a r i a t i o n of

to a c e r t a i n e x t e n t

(6), w h i l e p e r c e i v e d b e v e r a g e

in m a n y

aroma

distilled

aroma r e m a i n s

accept-

able.

W i t h a l c o h o l i c b e v e r a g e s , a h u m a n being p e r c e i v e s

an o v e r a l l

s e n s a t i o n , w h e r e a s c h e m i c a l a n a l y s e s m e a s u r e single c o m p o u n d s and t h e i r c o n c e n t r a t i o n s

aroma

(7). T h i s p e r c e i v e d

sensa-

tion m u s t , h o w e v e r , be b r o k e n d o w n into clearly d e f i n e d m e n t s and their v a r i a t i o n m e a s u r e d u n d e r control o b t a i n v a l i d d a t a for d e t a i l e d a n a l y t i c a l

results. T h i s

p r e s e n t s some e x a m p l e s of the a p p l i c a t i o n of r e c e n t methods

ele-

in o r d e r

to

paper

sensory

to the d e t e r m i n a t i o n of w h i s k y a r o m a and how

these

results have c o m p l e m e n t e d c h e m i c a l a n a l y s i s data on w h i s k y aroma. CHEMICAL COMPOSITION OF DISTILLED

BEVERAGES

T h e aroma c o m p o s i t i o n of d i s t i l l e d a l c o h o l i c b e v e r a g e s sists of several h u n d r e d d i s t i n c t c h e m i c a l c o m p o u n d s .

conFusel

a l c o h o l s , fatty acids and their esters usually d o m i n a t e

over

c a r b o n y l , p h e n o l i c , s u l p h u r and n i t r o g e n c o m p o u n d s but also lactones, acetals, hydrocarbons, components

s u g a r s , etc. rank a m o n g

aroma

(3-5).

T h e f o r m a t i o n of fusel a l c o h o l s , q u a n t i t a t i v e l y

the

g r o u p of a r o m a c o m p o n e n t s , does not d e p e n d on raw

largest

materials

147 b u t instead on the y e a s t e m p l o y e d and f e r m e n t a t i o n

conditions

(1). The fusel a l c o h o l c o n c e n t r a t i o n w i d e l y varies

in d i s t i l -

lates and e v e n in d i f f e r e n t types of b l e n d e d w h i s k y .

Canadian

w h i s k y c o n t a i n s a p p r o x i m a t e l y one tenth of the total level of fusel a l c o h o l s

in A m e r i c a n s t r a i g h t w h i s k e y ,

i.e.

Bourbon

w h i s k e y , and S c o t c h w h i s k i e s a b o u t half of the q u a n t i t y

in

B o u r b o n , or 0.8 - 1.0 g/1

(6). T h e m a i n c o m p o u n d

40-70 % of fusel a l c o h o l s

is 3 - m e t h y l - l - b u t a n o l , but also 2 -

methyl-l-butanol

comprising

and 2 - m e t h y l - l - p r o p a n o l are p r e s e n t

tilled a l c o h o l i c b e v e r a g e s , b e i n g 10-35 % of fusel (8).

in d i s -

alcohols

Of the a r o m a c o m p o u n d s s y n t h e s i z e d by y e a s t , fatty

c o n c e n t r a t i o n s may also vary o v e r a wide range c o n t r i b u t i o n s of a c i d s a d d i t i o n a l l y vary beverages

(9).

acid

Relative

in d i f f e r e n t

(3), to m e n t i o n 2 - e t h y l - 3 - m e t h y l b u t y r i c acid, w h i c h

has b e e n d e t e c t e d as a c h a r a c t e r i s t i c

i n d i c a t i v e of rum o n l y

(10).

F a t t y acid e s t e r s f o r m not only the n u m e r i c a l l y l a r g e s t

group,

b u t are a l s o an e s s e n t i a l g r o u p of aroma c o m p o u n d s c a u s i n g a r o m a intensity of d i s t i l l e d b e v e r a g e s to vary a c c o r d i n g t h e i r c o n c e n t r a t i o n . A l c o h o l i c b e v e r a g e s are very o f t e n

to clas-

s i f i e d as light a r o m a and heavy a r o m a b e v e r a g e s p r i m a r i l y the basis of the e s t e r c o n c e n t r a t i o n s . The r e l a t i v e tions of i n d i v i d u a l e s t e r s do not, h o w e v e r , vary, for e x a m p l e Ethyl acetate

whiskies.

in w h i s k i e s .

Scotch

w h i s k y c o n t a i n s 1 2 0 - 1 7 0 mg/1 of ethyl a c e t a t e , C a n a d i a n half of that a m o u n t , and in B o u r b o n w h i s k e y

whisky

the q u a n t i t y

e t h y l a c e t a t e may be t h r e e f o l d c o m p a r e d to that of whisky

on

contribu-

significantly

in C a n a d i a n , S c o t c h or B o u r b o n

is the m o s t a b u n d a n t ester

the

of

Scotch

(6).

C o n c e r n i n g c a r b o n y l c o m p o u n d s , a l d e h y d e s are i n t e r m e d i e n t s

in

the f o r m a t i o n of fusel a l c o h o l s . R e s e a r c h on c a r b o n y l

content

in d i s t i l l e d b e v e r a g e s has r e v e a l e d that d i s t i l l a t i o n

and

pyrolytic reactions occurring during distillation

influence

b o t h the a l d e h y d e and d i k e t o n e c o n c e n t r a t i o n s . D u r i n g

ageing,

148 the a l d e h y d e c o n t e n t declines both

increases, w h e r e a s

the level of

diacetyl

in c o g n a c s and w h i s k i e s . S c o t c h w h i s k y

c o n t a i n s more d i a c e t y l

evidently

than C a n a d i a n w h i s k y or B o u r b o n

whiskey

(12). A m o n g o t h e r v o l a t i l e c o m p o u n d s of interest

in d i s t i l l e d

ages are p h e n o l i c and s u l p h u r c o m p o u n d s . C r e s o l s and

bever-

guaiacol

s e e m to be p h e n o l s t y p i c a l of S c o t c h w h i s k y , even to the tent to a f f e c t

its a r o m a , a l t h o u g h q u a n t i t a t i v e l y

p h e n o l s n u m b e r the h i g h e s t c o m p o u n d s are p o t e n t i a l

in M a r t i n i q u e

ex-

volatile

rum (13).

Sulphur

causes of o f f - a r o m a s due to their

low

s e n s o r y t h r e s h o l d s . S u l p h u r c o m p o u n d s are of s i g n i f i c a n c e

in

the control of d i s t i l l a t i o n . T h e y have also b e e n o b s e r v e d

to

decrease distinctly during ageing

and

in oak b a r r e l s . B o u r b o n

Canadian whiskies contain noticeably fewer sulphur than Scotch whisky PERCEIVED AROMA

compounds

(14).

INTENSITY

T h r e s h o l d s and r e l a t i v e c o n t r i b u t i o n to a r o m a As techniques employed reliable quantitative

in c h e m i c a l a n a l y s i s p r o d u c e rapid r e s u l t s as w e l l ,

and

it w o u l d be t e m p t i n g

to

d e t e r m i n e the a r o m a of d i s t i l l e d b e v e r a g e s solely by m e a n s of gas c h r o m a t o g r a p h y . Y e t d i f f e r e n t b r a n d s of S c o t c h w h i s k y ,

for

instance, c a n n o t be d i s c e r n e d from each o t h e r m e r e l y on the b a s i s of c h e m i c a l a n a l y s i s

(6). A l t h o u g h aroma c o m p o u n d s

their c o n c e n t r a t i o n s are k n o w n , they do not d e t e r m i n e p e r c e i v e d aroma

and

the

intensity.

A n u n d e r s t a n d i n g of w h a t c o m p o u n d s have an e f f e c t on the is a r r i v e d at by s p e c i f y i n g c o m p o u n d s by m e a n s of

the t h r e s h o l d values for

aroma

aroma

iterative and c o n t r o l l e d m e t h o d s . In the

same w a y s o - c a l l e d o d o u r untis based on the t h r e s h o l d c o n c e n t r a t i o n of a r o m a c o m p o u n d s

in the food or

and

beverage

s t u d i e d could reveal some c r i t i c a l c o m p o u n d s c o n t r i b u t i n g

to

149

the p e r c e i v e d a r o m a

(15-17). H o w e v e r ,

the use of the

u n i t as an e s t i m a t e of r e l a t i v e a r o m a c o n t r i b u t i o n d i c t o r y to the valid p s y c h o p h y s i c a l

odour

is

contra-

f u n c t i o n as F r i j t e r s

(18)

has s h o w n in d e t a i l , and can give a w r o n g p i c t u r e of the potential, especially

at the s u p r a t h r e s h o l d

level of

aroma

compo-

nents . S t u d i e s on the c o n t r i b u t i o n of a r o m a c o m p o u n d s to the of aroma in a S c o t c h w h i s k y 3-methyl-l-butanol,

imitation

(15) has

strength

indicated

that

ethyl a c e t a t e and 3 - m e t h y l b u t y l a c e t a t e

the e t h y l e s t e r s of C g - C 1 2

acids can g r e a t l y a f f e c t the

a r o m a intensity of a w h i s k y

and

total

i m i t a t i o n . The c o n t r i b u t i o n of

c a r b o n y l c o m p o u n d s w a s a s s e s s e d to be the g r e a t e s t , a l t h o u g h f a l s e l y . C o n v e r s l y , fatty acids m a d e no g r e a t e r c o n t r i b u t i o n to the a r o m a of a w h i s k y

Variation

imitation

(16).

in a r o m a c o m p o n e n t s and p e r c e i v e d

C o n c e n t r a t i o n s of m a n y a r o m a c o m p o u n d s

intensity

in d i f f e r e n t

production

b a t c h e s of the same w h i s k y brand may vary e v e n m o r e w i d e l y than b e t w e e n d i f f e r e n t w h i s k y b r a n d s . T h i s p o s e s the q u e s t i o n of how g r e a t an increase detected

in c o n c e n t r a t i o n can be

generally

in p e r c e i v e d a r o m a . F i g u r e 1 i l l u s t r a t e s the

tion t h r e s h o l d s of a d d e d a r o m a c o m p o u n d s or m i x t u r e s

detecwhen

c o m p a r e d to the a m o u n t s of these c o m p o u n d s found in C a n a d i a n , S c o t c h and B o u r b o n w h i s k i e s . The t h r e s h o l d s are e s t a b l i s h e d

by

an a s c e n d i n g c o n c e n t r a t i o n s e r i e s using m e t h o d of l i m i t

(19).

O n l y the p a n e l a v e r a g e s of t h r e s h o l d s are p r e s e n t e d and

the

individual variation

is d i s r e g a r d e d

in this c o n t e x t . The

ure i n d i c a t e s that e v e n an a v e r a g e d e t e c t i o n t h r e s h o l d

is

g e n e r a l l y h i g h e r than the u p p e r c o n c e n t r a t i o n limits of m a i n fusel a l c o h o l ,

i.e. 3 - m e t h y l - l - b u t a n o l ,

the m o s t

a r o m a c o m p o n e n t s c o u l d thus increase two- or t h r e e f o l d in the aroma. The m o s t e a s i l y

tected d i f f e r e n c e s c o u l d be c a u s e d by e t h y l a c e t a t e

the

common

e s t e r s or the c a r b o n y l c o m p o u n d s . T h e c o n c e n t r a t i o n s of any d e t e c t a b l e d i f f e r e n c e

fig-

these without de-

in S c o t c h

150

mg/l 1400T

1200

mg/l

1000

C - Canadian S - Scotch B - Bourbon

500r

mg/l

400

C S B 3-Mathyl-l-butanol

C S B Ethyl acatata

C S B Carbonyl compounds

Fig. 1. Detection thresholds of added aroma compounds ([!•) compared to the amounts found in three types of whisky. Vertical lines indicate normal variation quantities except for carbonyls partly based on estimates. Composition of ester and carbonyl mixtures are given in Table 1.

whisky and the ethyl esters of the C 8 - C 1 2 acids in Bourbon whiskey. Aroma intensity measured by magnitude estimation Even an untrained assessor can readily discern different whisky types, for instance Scotch, Bourbon and Canadian whiskies. Difficulties are not encountered until the strength of single and multiple sensations have to be quantified. The use of ratio scales offer a solution by applying a magnitude estimation method. In this method panelists are instructed to assign

151 n u m b e r s to a s e q u e n c e of stimuli

so that the ratios of

n u m b e r s r e f l e c t the ratios of the p e r c e i v e d donic responses

intensity or h e -

(20).

A s the b e s t r e s u l t s have b e e n a c h i e v e d tion intensity

the

in e v a l u a t i n g a s e n s a -

in p a r t i c u l a r by the m a g n i t u d e e s t i m a t i o n

this m e t h o d w a s used to study the r e l a t i o n s h i p of the

(21),

aroma

s t r e n g t h and of a typical w h i s k y a r o m a to the c o n c e n t r a t i o n aroma components

(22). It w a s e s p e c i a l l y sought to e x a m i n e

d i f f e r e n c e s b e t w e e n v a r i o u s types of w h i s k i e s a f t e r compounds contributing

to the aroma of a w h i s k y

b e e n added. S c a r c e l y any d i f f e r e n c e s

in

of the

aroma

imitation

r e s p e c t to the

p e a r a n c e of a typical w h i s k y a r o m a b e t w e e n C a n a d i a n ,

had

disap-

Scotch

and B o u r b o n w h i s k i e s w a s o b s e r v e d , and e v e n the s t r e n g t h of a r o m a grew

in the same d i r e c t i o n ,

methyl-l-butanol,

i r r e s p e c t i v e of w h e t h e r

e s t e r s or c a r b o n y l c o m p o u n d s w e r e

3-

added.

O n l y the, a d d i t i o n of ethyl a c e t a t e seems to make a g r e a t e r c o n t r i b u t i o n to C a n a d i a n w h i s k y and B o u r b o n w h i s k e y

than to

S c o t c h w h i s k y . C a r b o n y l c o m p o u n d s w e r e s h o w n to be far significant

in all three w h i s k y types than had b e e n a s s u m e d

the basis of their low t h r e s h o l d s , as also w e r e a r o m a pounds,

less

such a m i x t u r e of 2 - m e t h y l - l - p r o p a n o l and

butanol, seemingly contributing

on

com-

2-methyl-l-

little to the w h i s k y

aroma.

A l t h o u g h the r e l a t i o n s h i p of a t y p i c a l w h i s k y a r o m a to the q u a n t i t i e s of the m a i n a r o m a c o m p o n e n t s was c l a r i f i e d by m e a n s of ratio s c a l e s , the use of these scales p r o v e d to be p r o b l e m a t i c even to a trained p a n e l d e t e r m i n i n g

such a c o m p l i c a t e d

s e n s a t i o n as a t y p i c a l w h i s k y aroma. The p a n e l i s t s

tended

e m p l o y the scales as a c a t e g o r y scale and also d e v e l o p e d vidual scales during testing. The questionable of ratio scales has e m e r g e d particularly

in o t h e r c o n t e x t s as w e l l

(23),

(24). T h e

m a g n i t u d e e s t i m a t i o n m e t h o d and the n i n e - p o i n t c a t e g o r y d e s r i p t o r s of w h i s k e y

in the e v a l u a t i o n of

indi-

serviceability

in i n v e s t i g a t i o n s of h e d o n i c r e s p o n s e s

m e t h o d w e r e also c o m p a r e d

to

scale

sensory

sour m i x e s . N o d i f f e r e n c e was

observed

152 in the m e t h o d s w h e n s a m p l e s d i f f e r e d f r o m each o t h e r b u t the use of m a g n i t u d e e s t i m a t i o n r e s u l t e d

enough,

in p a n e l i s t s -

s a m p l e i n t e r a c t i o n , w h i l e use of c a t e g o r y scales c a u s e d v a r i a b i l i t y due to p a n e l i s t s and r e p l i c a t i o n

(25).

of the intensity of w h i s k y a r o m a by the m a g n i t u d e m e t h o d showed no d e g r e e of s t a t i s t i c a l

Evaluation estimation

significance

v i d u a l v a r i a t i o n a m o n g p a n e l i s t s , but it r e q u i r e d

in

indi-

anchoring

the scale to a g i v e n s t a n d a r d sample, u s u a l l y p u r e QUANTITATIVE AROMA

more

whisky.

DESCRIPTION

A l t h o u g h d i s c e r n i n g d i f f e r e n t types of w h i s k y f r o m e a c h p o s e s few d i f f i c u l t i e s , far more

their a r o m a d e s c r i p t i o n

other

becomes

i n t r i c a t e . I n f o r m a t i v e p r o f i l e p i c t u r e s can be o b -

tained t h r o u g h q u a n t i t a t i v e d e s c r i p t i v e a n a l y s i s

(26-27)

e.g.

for the a r o m a of d i f f e r e n t types of w h i s k i e s . F i g u r e 2 d e p i c t s d i f f e r i n g a r o m a p r o f i l e s for S c o t c h w h i s k y and B o u r b o n k e y , a l t h o u g h their a r o m a i n t e n s i t i e s d i s p l a y no

d i f f e r e n c e s . The i n t e n s i t y of the C a n a d i a n w h i s k y a r o m a instead c l e a r l y lower, a l t h o u g h

its a r o m a p r o f i l e

Aroma

is

greatly

r e s e m b l e s that of B o u r b o n w h i s k e y , at least from the p o i n t of

whis-

distinct

Finnish

view.

profiles

A n a r o m a p r o f i l i n g s y s t e m of 14 d e s c r i p t i v e ence has b e e n d e v e l o p e d

terms and p r e f e r -

to d e t e r m i n e w h i s k y aroma. T h e

de-

s c r i p t i v e terms are b a s e d on the v o c a b u l a r l y used to d e s c r i b e whisky flavour

(28), and our o w n e x p e r i e n c e g a i n e d

in the

a s s e s s m e n t of w h i s k y s a m p l e s . M e t h o d s r e p o r t e d by W i l l i a m s (29, 30) have been used to s e l e c t the final terms, b u t the o p p o r t u n i t y

to p r e p a r e m o d e l s a m p l e s d e s c r i b i n g

the

also terms

has had an influence on s e l e c t i o n . A c o m m a n d of as m a n y as 15 terms on an a s s e s s m e n t o c c a s i o n is s e e m i n g l y u n w i e l d y e v e n trained assessors,

and t h e r e f o r e

limiting the n u m b e r of

for

terms

153

Grain-like

Fig. 2. P r o f i l e p i c t u r e o b t a i n e d by q u a n t i t a t i v e d e s c r i p t i o n a n a l y s i s of S c o t c h ( — ) , B o u r b o n ( ) and C a n a d i a n ( ) whisky. T h e r e l a t i v e intensity for each c h a r a c t e r i s t i c is d e c i p t e d by the length of the line from the c e n t r e .

to below that in s e v e r a l d e s c r i p t i v e a n a l y s e s has b e e n

endeav-

oured . T h e d e s c r i p t i v e a n a l y s i s w a s a n c h o r e d to a r e f e r e n c e ,

usually

the u n a d u l t u r a t e d w h i s k y s t u d i e d . The d e g r e e of d i f f e r e n c e intensity of each d e s c r i p t o r f r o m the r e f e r e n c e was u s i n g an u n s t r u c t u r e d

line labeled

ends, and the c e n t e r l a b e l e d the r e f e r e n c e p o i n t

in

rated

"Less" and "More" at the

" R e f e r e n c e " . The d i s t a n c e

from

indicated the degree of the p e r c e i v e d

intensity, same as has b e e n r e p o r t e d for e x a m p l e

in d e s c r i p -

154

T a b l e 1 C o n c e n t r a t i o n s of the a r o m a c o m p o u n d s s t u d i e d w i t h q u a n t i t a t i v e d e s c r i p t i v e a n a l y s i s m e t h o d in three types of whisky

Compound

C a n a d ian whisky mg/1

Scotch whisky mg/1

Bourbon w h iskey mg/1

3-Methyl-l-butanol

120

210

830

47

90

160

Ethyl

acetate

Ethyl

hexanoate

0.5

0.8

3.6

Ethyl

octanoate

0.6

2.8

16.2

Ethyl

decanoate

0.9

8.8

28.5

Ethyl

dodecanoate

3-Methylbutylacetate Total Acetaldehyde

0.5

8.0

12.4

0.3

5.6

5.1

2.8

26.0

65.8

13.2

18.7

33.0

iso-Butyraldehyde

5.5

7.8

13.8

iso-Valeraldehyde

3.3

4.5

8.3

Diacetyl

0.5

0.8

1.0

22.5

31.8

56.1

tive a n a l y s e s of w i n e each t e s t i n g ,

(31) and b e v e r a g e s

and g e l a t i n s

the p a n e l c o n s i s t e d of at least 12

(32).

t r a i n e d to e v a l u a t e w h i s k y s a m p l e s but w h o w e r e not e x p e r t s whisky

In

assessors in

brands.

Table 1 indicates

the c o n c e n t r a t i o n s of

3-methyl-l-butanol,

some fatty acid e s t e r s and c a r b o n y l c o m p o u n d s

in three

ent types of w h i s k y s t u d i e d . F i g u r e s 3, 4 and 5 p r e s e n t

differthe

p a n e l a v e r a g e s of a r o m a d e s c r i p t o r s for C a n a d i a n , B o u r b o n

and

S c o t c h w h i s k i e s . O n l y the m o s t i m p o r t a n t d e s c r i p t o r s have

been

155

+2.0

+2.0

3-Methyl-l-butanol

+ 1.0

.Ik - 1.0

n l r IT

tr

n~

1.0

Ik

-1.0

-2.0

•2.0

z

1

+2.0

+ 1.0 - 1.0

Ethyl acetate

+2.0

j

Eta

"0"

u

-2.0 M

Z

+1.0 T F

- 1.0

n

HF

-2.0

z + 2.0

P

-1.0 -2.0

+2.0

Carbonyl compounds

+1.0

+ 1.0

-2.0

-2.0

-+2.0

Esters

+1.0

- 1.0

-1.0

>

S +2.0 -

I

+1.0

1 1

2

3

4

5

6

7

8

9

10 11

12

13 14

1

-1.0 -2.0

15

Descriptors

Fig. 3. Aroma profile of Canadian whisky augmented (IZH) or lacking (^B) the aroma components mentioned. Descriptors are 1. Typical whisky aroma 2. Solvent-like 3. Fusel oil (3-Methyl-lbutanol ) 4. Estery (fruity) 5. Grain-1ike 6. Woody 7. Spicy (peppery)

8. 9. 10. 11. 12. 13 . 14. 15.

Tallowy Medicinal (phenolic) Tarry Malty Burnt Aroma of sweet sap Pungent Preference

156

+2.i

+2.0

3-Methyl- -butanol

+ 1.1

1.0

i

T -l.i

-1.0

-2J

-2.0

Z

Z +2.i

ll

+ 1.

Ethyl acetate

+1.0

0.

-l.i -2. >•

S -3 B „

+2.0

D-

n

I

U

i

"J

Z

^ + 1,

1

-1. -2.

n 4

a

q .

I -1.0 --2.0

73.0 ?

+2.

Carbonyl compouds

+ 10 H

-2.

-3.0

•+1.0

-3.

-1j

-2.0

|+2.0

Isters

o> + 2 .

-1.0

U D1

J 1 2

IT

+2.0

a

+1.0

-1.0

-2.0

3

4

5

6

7

8

9

10 11

12 13 14 15

Descriptors

Fig. 4. Aroma profile of Bourbon whiskey augmented (CD) or lacking ( M ) the aroma components mentioned. Descriptors are the same as in Figure 3.

157

selected, since the individual variation of panelists has been omitted. The figures give two concentration levels of aroma components. The left-hand column describes conditions where each aroma component has been added to whisky in amounts

Descriptors

Fig. 5. Aroma profile of Scotch whisky augmented (CD) or lacking (•§) the aroma components mentioned. Descriptors are the same as in Figure 3.

158

corresponding

to the u p p e r d e t e c t i o n limit of the a r o m a

nent, so that each p a n e l i s t s h o u l d d i s t i n g u i s h the

compo-

augmented

s a m p l e from the p u r e w h i s k y s e r v i n g as the r e f e r e n c e . The r i g h t - h a n d c o l u m n a g a i n r e p r e s e n t s c o n d i t i o n s w h e r e the component

is m i s s i n g f r o m the w h i s k y sample.

be c o n s i d e r e d a r t i f i c i a l ,

The s a m p l e

as the r e m a i n i n g a r o m a

h a v e b e e n s u p p l e m e n t e d by c h e m i c a l

aroma

fcompounds.

Five p e r cent of

w h i s k y w a s a d d e d to the s a m p l e s to c o m p e n s a t e for this c i e n c y , and e x a m i n a t i o n of the p r e f e r e n c e c o l u m n s

defi-

(see d e -

s c r i p t o r 15) r e v e a l s no s y s t e m a t i c i n f e r i o r i t y of these ples

in c o m p a r i s o n to the a u g m e n t e d w h i s k y

can

components

sam-

samples.

T h e v a r i a t i o n b e t w e e n d i f f e r e n t types of w h i s k i e s turned to be m i n o r

in these a r o m a p r o f i l e s as w e l l . A d d i n g or

m o v i n g 3 - m e t h y l - l - b u t a n o l , the m a i n c o m p o n e n t of fusel h o l s , m a i n l y b e c o m e s a p p a r e n t only

in g e n e r a l

alco-

descriptors,

such as "typical w h i s k y a r o m a " and " p r e f e r e n c e " . E t h y l of the C g - C 1 2

esters

a c i d s s e e m to cause the m a j o r i t y of c h a n g e s

the a r o m a of S c o t c h w h i s k y and c a r b o n y l c o m p o u n d s the changes

out

re-

in

least

in the a r o m a of C a n a d i a n w h i s k y , w h e r e a s e t h y l

acetate

p r o d u c e s an e f f e c t on m o s t d e s c r i p t o r s of B o u r b o n w h i s k e y . Of e s s e n t i a l whiskies

interest

in the a r o m a p r o f i l e s for

in q u e s t i o n p r o d u c e a c h a n g e

in the same a r o m a

f r e q u e n t l y even in the same d i r e c t i o n , w i t h emerging

different

is that both r e m o v i n g and a d d i n g the a r o m a

component

descriptor,

differences

in the intensity only. T h u s p a n e l i s t s did not

describable differences

detect

in p e r c e i v e d w h i s k y aroma, e x c e p t

for

c e r t a i n i n b a l a n c e s w h i c h m a y h a v e b e e n solely c a u s e d by the a r t i f i c i a l nature of the

samples.

Individual variation was considerable, however. The only

figures

include those d e s c r i p t o r s w i t h an intensity of at

one of the s a m p l e s e x c e e d i n g

the s t a n d a r d d e v i a t i o n of

p a n e l . R e s u l t s to be p u b l i s h e d later w i l l deal d e t a i l w i t h the role of i n d i v i d u a l

variation.

least the

in g r e a t e r

159

P r o j e c t i o n s of a r o m a

profiles

In the e n d e a v o u r to d e v e l o p the best p o s s i b l e a r o m a tors for d i f f e r e n t w h i s k y types, the m u l t i p l e

descrip-

relationships

among a r o m a c o m p o n e n t s and d e s c r i p t o r s w e r e f u r t h e r tried

to

depict more clearly. A principal

on

c o m p o n e n t a n a l y s i s based

c o v a r i a n c e s b e t w e e n d e s c r i p t o r s w a s a p p l i e d to the a r o m a p r o files using multivariate

techniques developed

by V u a t a z

F i g u r e s 6, 7 and 8 p r e s e n t the p r o j e c t i o n s o b t a i n e d

(33).

for

C a n a d i a n , B o u r b o n and S c o t c h w h i s k i e s . As the f i r s t two p r i n cipal axes e x p l a i n 74-91 % of the total v a r i a t i o n whiskies,

the f i g u r e s can be c o n s i d e r e d fairly

in d i f f e r e n t

sucessful.

2nd axis 20 °L

Fig. 6. P r o f i l e in two d i m e n s i o n s for the a r o m a p r o f i l e of C a n a d i a n w h i s k y . T h e a r o m a d e s c r i p t o r s , w h i c h are g i v e n in F i g u r e 3, are r e p r e s e n t e d by v e c t o r s and a r o m a c o m p o n e n t s by d o t s . The scale is the same for both.

160

2nd a x i s 21%

F i g . 7. P r o f i l e in two d i m e n s i o n s for the a r o m a p r o f i l e of B o u r b o n w h i s k e y . T h e a r o m a d e s c r i p t o r s , w h i c h are g i v e n in F i g u r e 3, are r e p r e s e n t e d by v e c t o r s and a r o m a c o m p o n e n t s by dots. The scale is the same for both.

T h e f i g u r e s show that for S c o t c h w h i s k y the t y p i c a l aroma descriptor

whisky

(vector no. 1) w i t h the g r e a t e s t v a r i a t i o n

is

p o s i t i o n e d on the o t h e r side of the first p r i n c i p a l

axis

for C a n a d i a n and B o u r b o n w h i s k i e s . The f i g u r e s also

indicate

that d i f f e r e n t a r o m a d e s c r i p t o r s

negatively correlate with

t y p i c a l w h i s k y a r o m a in d i f f e r e n t w h i s k i e s . T h e t y p i c a l of B o u r b o n w h i s k e y c o u l d be best d e s c r i b e d by the "aroma of s w e e t s a p " y, b u r n t and s p i c y "

the

aroma

descriptor

(no. 13), w h e r e a s S c o t c h w h i s k y by

"malt-

(no. 11, 12 and 7) and the C a n a d i a n w h i s k y

by "estery and g r a i n - l i k e " "estery"

than

(no. 4 and 5). The

descriptor

is a l s o e s s e n t i a l for B o u r b o n w h i s k e y , but it does

161

2nd axis 13%

F i g . 8. P r o f i l e in t w o d i m e n s i o n s f o r the a r o m a p r o f i l e of S c o t c h w h i s k y . T h e a r o m a d e s c r i p t o r s , w h i c h a r e g i v e n in F i g u r e 3, a r e r e p r e s e n t e d by v e c t o r s a n d a r o m a c o m p o n e n t s b y d o t s . T h e s c a l e is the s a m e f o r b o t h .

not correlate

as f u l l y w i t h t h e t y p i c a l

key. The aroma profile projections descriptor-vector

clusters,

indicating overt

of c e r t a i n a r o m a d e s c r i p t o r s . "grain-like" and each other

"woody"

in a n y w h i s k y

o n the o t h e r h a n d ,

For

descriptors from whiskey,

is p o r t r a y e d o n l y by 5 - 6 d e s c r i p t o r s , for Scotch whisky

axes that new and

experience gained from these profile

flat

better

aroma

de-

o n the b a s i s

pictures.

while

remains so

a r e to be f o u n d . C o n s e q u e n t l y ,

are today further being d e v e l o p e d

whis-

reveal

autocorrelation the

T h e a r o m a of B o u r b o n

in r e s p e c t to the s e c o n d p r i n c i p a l scriptors

instance,

also

(no. 5 a n d 6) d i d n o t d i f f e r type.

the aroma profile projection aroma descriptors

a r o m a of B o u r b o n

of w h i s k i e s

of

162 The p o s i t i o n i n g of the a r o m a c o m p o n e n t s s t u d i e d - m a i n

fusel

a l c o h o l , t y p i c a l e s t e r s and c a r b o n y l c o m p o u n d s - near the f i r s t p r i n c i p a l axis

in the p r o j e c t i o n s u p p o r t s

the view

that

their c o n t r i b u t i o n to the a r o m a of d i f f e r e n t types of w h i s k i e s is largely s i m i l a r . T h e s e c o m p o n e n t s c o n t r i b u t e

to a r o m a

in high c o n c e n t r a t i o n v a r i a t i o n s , w h i c h m a i n l y a f f e c t b a l a n c e of w h i s k y , but w i t h h a r d l y any e f f e c t on the

only

the charac-

t e r i s t i c a r o m a of d i f f e r e n t types of w h i s k i e s . The projections

are a l s o useful

v a r i a t i o n of p a n e l i s t s

in f o l l o w i n g

the

in the use of d i f f e r e n t a r o m a

tors. W h e n a p p l y i n g the same m e t h o d to the w h o l e taking

descrip-

material

into a c c o u n t v a r i a t i o n b e t w e e n both s a m p l e s and p a n e l -

ists, p r o j e c t i o n s

c o u l d no l o n g e r be p r e s e n t e d

s i o n a l form, but at least three d i m e n s i o n s of significance emerged. As occurs

individual

in r e a l i t y ,

in a two d i m e n statistical

individual personal variation

the i n t e r r e l a t i o n s h i p s

between personal

sample v a r i a t i o n call for f u r t h e r study w h e n the of a r o m a c o m p o n e n t s v a r i e s w i t h i n n a t u r a l

also and

concentration

limits.

References 1. S u o m a l a i n e n , H., L e h t o n e n , M.: J. Inst. B r e w . , L o n d o n 1 4 9 - 1 5 6 (1979). 2. S w a n , J . S . , B u r t i e s , S.M.: Chem. Soc. R e v . 7, (1978 ) .

85,

201-211

3. L e h t o n e n , M., S u o m a l a i n e n , H.: in "Economic M i c r o b i o l o g y , V o l . 1. A l c o h o l i c B e v e r a g e s " (A.H. R o s e , ed.) 5 9 5 - 6 3 3 , A c a d e m i c P r e s s , L o n d o n 1977. 4. P o s t e l , W . , D r a w e r t , F., A d a m , L.: in " G e r u c h - und G e s c h m a c k s t o f f e " , (E. D r a w e r t , ed.) 9 9 - 1 1 1 , V e r l a g H a n s C a r l , N ü r n b e r g 1975. 5. Kahn, J . H . : J. A s s . O f f i c . A n a l . C h e m . 52, (1969 ) .

1166-1178

6. L e h t o n e n , M., S u o m a l a i n e n , H.: P r o c e s s B i o c h e m . 5 - 9 , 26 (1979).

N o 2,

163

7. Jounela-Eriksson, P.: in "Olfaction and Taste VI, Paris 1977", Proc. (J. Le Magnen, P. MacLeod eds.) 409-419, Information Retrieval Ltd., London 1977. 8. Suomalainen, H., Lehtonen, M.: Kemia-Kemi 3, 69-77

(1976).

9. Nykänen, L., Puputti, E., Suomalainen, H.: J. Food Sci. 33, 88-92 (1968). 10. Lehtonen, M.J., Gref, B.K., Puputti, E.V., Suomalainen, H.: J. Agr. Food Chem. 25, 953-955 (1977). 11. Suomalainen, H., Nykänen, L.: Wallerstein Lab. Commun. 35, 185-202 (1972). 12. Leppänen, O., Ronkainen, P., Koivisto, T., Denslow, J.: J. Inst Brew. London 85, 278-281 (1979). 13. Jounela-Eriksson, P., Lehtonen, M.: to be published in "Quality of Foods and Beverages; Recent Developments in Chemistry and Technology", Academic Press, New York 1981. 14. Leppänen, 0., Denslow, J., Ronkainen, P.: J. London 85, 350-353 (1979).

Inst.Brew.,

15. Salo, P., Nykänen, L., Suomalainen, H.: J. Food Sci. 37, 394-398 (1972). 16. Salo, P.: in "Aroma Research", (H. Maarse, P.J. Groenen, eds.) 121-130, Centre for Agricultural Publishing and Documentation, Wageningen 1975. 17. Rothe, M.: Nahrung 20, 259-266

(1976).

18. Frijters, J.E.R.: Chem. Senses Flavour 3, 227-233

(1978).

19. Standard Practice for Determination of Odor and Taste Thresholds by a Forced-choice Ascending Concentration Series Method of Limits, ANSI/ASTM E 679-79 Am. Soc. Test. Mater., Book of ASTM Stand, Philadelphia, 1979. 20. Moskowitz, H.R.. J. Food Sci. 28, No 11, 16-21

(1974).

21. Meilgaard, M.C., Reid, D.S.: in "Progress in Flavour Research" (D.G. Land, H.E. Nursten, eds.), 67-73. Applied Science Publishers Ltd, London (1979). 22. Jounela-Eriksson, P.: in "Flavor of Foods and Beverages", Chemistry and Technology (G. Charalambous, G.E. Inglett, eds) 339-354, Academic Press, New York 1978. 23. O'Mahony, M.: J. Dairy Sci 62, 1954-1962

(1979).

24. Pangborn, R.M.: in "Carbohydrate Sweeteners in Foods and Nutrition" (P. Koivistoinen, L. Hyvönen, eds) 87-110, Academic Press, New York 1980. 25. McDaniel, M.R., Sawyer, F.M.: J. Food Sci 46, 178-181, 189 (1981) .

164

26. S t o n e , H., S i d e l , J., O l i v e r , S., W o o l s e y , A . , R . C . : J. F o o d Sei 28, No 11, 24-34 (1974).

Singleton,

27. S t o n e , H . , S i d e l , J . L . , B l o o m q u i s t , J.: C e r e a l F o o d W o r l d 25, 642-644 (1980). 28. P i g g o t , J . R . , J a r d i n e , S.P.: J. 82-85 (1979). 29. W i l l i a m s , A . A . :

J. Sei Fd.

Inst. B r e w . , L o n d o n

Agric.

.28, 1090-1104

30. W i l l i a m s , A . A . . in: "Progress in F l a v o u r R e s e a r c h " Land, H . E . N u r s t e n , eds) 2 8 7 - 3 0 5 , A p p l i e d S c i e n c e P u b l i s h e r s Ltd, L o n d o n 1979.

85, (1977).

(D.G.

31. N o b l e , A . C . : in "Analysis of Food and B e v e r a g e s : H e a d s p a c e T e c h n i q u e s " (G. C h a r a l a m b o u s , ed.) 2 0 3 - 2 2 8 , A c a d e m i c P r e s s , N e w Y o r k (1978). 32. L a r s o n - P o w e r s , N . , P a n g b o r n , R.M. : J. Food S e i . 43^, 47-51 (1978 ). 33. V u a t a z , L.:

in "Nestle R e s e a r c h N e w s 1 9 7 6 / 7 7 " 57-71, N e s t l e

P r o d u c t s T e c h n i c a l A s s i s t a n c e , Co. Ltd

(1977).

PHYSICO-CHEMICAL STUDIES ON FLAVOUR-ACTIVE COMPOUNDS

J.W. Gramshaw and D.R. Williams Procter Department of Food Science, The University of Leeds, Leeds LS2 9JT, England

Introduction A number of organic compounds naturally present as flavour components in foods are notable for the very low odour thresholds which they exhibit. Outstanding in this respect are the alkylmethoxypyrazines, of which 2-isobutyl-3-methoxypyrazine and 2isopropyl-3-methoxypyrazine have detection thresholds of 2 x — 6

10 ppm in water (1,2) and dialkylthiazoles, of which 4,5dimethylthiazole can be detected in aqueous solution at a concentration of 4.7 x 10" ppm (3). Such compounds clearly are potent odourants which require very few molecules within the vicinity of the olfactory epithelium in order to evoke a response. In no example of such powerful odourants has the threshold been determined directly in air, which would be an extremely difficult, perhaps impossible, task. If it is necessary to know the minimum concentration of an odourant in air which is required to stimulate response, this is readily available, at least in principle, as the product of the threshold, measured as an aqueous solution, and the air-water partition coefficient, Ka//W- It is possible, however that potent odourants may exhibit surface activity and thus become concentrated into the interface between atmosphere and aqueous solution. If this were to occur, K , would increase a/ W

as the concentration decreased and more odourant molecules would be released into the gas phase than predicted from the accepted value of K , (which would have been measured using

© 1981 by W a l t e r de Gruyter & Co, Berlin • New York Flavour '81

166 a solution concentration many orders of magnitude greater than the threshold concentration).

Results and Discussion This possibility was examined by measuring the air-water partition coefficient,

K

a/W/

for two alkylmethoxypyrazines over a

range of concentrations. The compounds chosen, 2-ethyl-3-methoxypyrazine and 2-isopropyl-3-methoxypyrazine, are believed to contribute

strongly to potato aroma (4-7). The first named

has a detection threshold 200 times greater than the second (1, 2) and it seemed not unreasonable to suppose that some difference in behaviour might exist between their dilute aqueous solution. Static method for determination of partition coefficients. A "static" method was devised which relied upon collection of the total pyrazine in the headspace above a dilute aqueous solution of known concentration. The collection method involved adsorption of the headspace vapour within a carefully designed trap containing charcoal from which the adsorbed pyrazine subsequently was transferred directly to a capillary gas chromatographic column and the quantity measured by use of an external standard. Use of a relatively large headspace volume enabled the necessary sensitivity to be achieved and painstaking development of the charcoal trap enabled quantitative recovery of alkylmethoxypyrazines to be achieved consistenly down to a -9 level of 5 x 10 g. 3 A large volume of vapour (1526 cm ) was obtained by using a vapour chamber3 in addition to a solution chamber which contained 100 cm of an aqueous solution of the alkylmethoxypyrazine under examination. A double acting bellows pump was used to circulate the vapour very slowly between the two chambers in order to reach equilibrium without disturbing the surface

167

layer in the solution chamber. All internal surfaces of the apparatus, which is shown diagramatically in Fig. 1, were of PTFE in order to reduce adsorption effects to a minimum. After 3 h (equilibrium was established within 21/2 h), the solution chamber was isolated from the rest of the system and the vapour (1387 cm^) which remained outside the solution chamber and its connecting lines was circulated repeatedly through the activated charcoal trap during a period of 3 h. The trap was then removed and the amount of alkylmethoxypyrazine which it Fig. 1 "Static" Apparatus

Closed circuit loop tor absorbing volatiles

>

had adsorbed was determined by gas chromatography using a specially designed injection device to transfer the pyrazine directly to a gas chromatographic column. Measurement of the area

168

of the resulting peak and comparison with the area of a peak generated from a precisely known and similar amount of the pyrazine directly injected onto the chromatographic column using a syringe was used to calculate the total amount of alkylmethoxypyrazine present in the vapour phase of the static apparatus and thus the concentration in the vapour phase and the value of were easily computed. It was essential that the apparatus was thoroughly cleaned before each series of experiments and that the air which it contained was purified immediately before use in each individual experiment. The apparatus was immersed in a water bath whose temperature was controlled very closely (25° - 0.005°C) since fluctuations could have led to condensation of water vapour within the vapour circuit and this would have given erroneous results. Partition coefficients of 2-ethyl-3-methoxypyrazine and 2-isopropyl-3-methoxypyrazine. The partition coefficients of the two alkylmethoxypyrazines were measured in ascending order of concentration over a range from 0.02 ppm to 20 ppm. Within the limits of experimental error, the same value of partition coefficient, K a / W = 1.9 x 10 , was obtained for 2-isopropyl-3methoxypyrazine at each concentration at which measurements were made. The value obtained for the partition coefficient of 2-ethyl-3-methoxypyrazine, however, showed a progressive de-3 crease (Fig. 2) from 1.2 x 10 , when measured using an aqueous -4 solution containing 0.02 ppm, to 5.6 x 10 when an aqueous solution containing 20 ppm was used. The possibility was considered that adsorption of pyrazine from the vapour circuit onto the surface of the PTFE might occur and that some, or all, would subsequently be released from the PTFE surface during collection of the headspace pyrazine onto the trap. Obviously this would have led to an erroneously high value of K a / W being recorded. However, such adsorption was shown not to occur under the experimental con-

169

Fig. 2 Variation in air/water partition coefficient as a function of solution concentration (2-ethyl-3-methoxypyrazine, static method)

0.02

0.2

2

20

200

Concentration (ppm log scale)

ditions used by placing a quantity of PTFE wool into the vapour circuit such that the effective surface area was approximately doubled. The partition coefficient measured with the PTFE wool in place was unchanged from the value recorded in its absence and thus no contribution due to adsorption onto PTFE surfaces had occurred. The increase in the value of K a ^ for 2-ethyl-3-methoxypyrazine as the degree of dilution is increased is considered to be due to concentration of pyrazine molecules into the surface layer of the aqueous solution - an effect which is more obvious as the solutions studied are made increasingly dilute. Since par-

170 tition must, of necessity, occur between the surface layer of the solution and the air to which it is exposed, the measured partition coefficient will alter if the surface layer becomes unrepresentative of the bulk concentration in the solution. As the imbalance between the bulk solution and the surface layer becomes more marked, which is thought to have occurred in the case of 2-ethyl-3-methoxypyrazine, the measured value of will increase. Dynamic method for determination of partition coefficients. The dynamic method of measuring the partition coefficients of the two alkylmethoxypyrazines was a development of the saturation cell described by Burnett (8). In Burnett's method, a gas was passed, as very fine bubbles, through a relatively small volume of the solution of interest and the concentration of the solute in the exhaust gas was determined by gas chromatography. Measurement was achieved by filling a small gas sampling loop with exhaust gas and transferring this to the chromatographic column. The concentration of solute in the gas phase was measured at intervals and the rate of its decrease with time was used to calculate the partition coefficient. The method, as described, is not sufficiently sensitive for use with the very dilute solutions encountered in the present investigation and it was therefore modified by collecting the alkylmethoxypyrazine molecules from a large volume of exhaust gas onto a charcoal trap in a fashion similar to that described for the static method. Also, a large volume of aqueous solution (relative to that used by Burnett) was employed and the partition coefficient was calculated directly from the solute concentration in the exhaust gas and not by measuring the rate of decrease of solute concentration in the aqueous solution. A gas volume of 60 cm

was adequate when solutions containing 200

ppm (the most concentrated) were used but alkylmethoxypyrazine from 1200 cm 3 was required when the most dilute solutions (0.02 ppm) were studied.

171

Fig. 3

The modified apparatus is shown diagrammatically in Fig. 3 and described in the Experimental section. The saturation cell was immersed in a thermostated water bath but it was unnecessary to control the temperature (25 + 0.5°C) of this as closely as in the static method. It was, however, essential to control the air flow rate very accurately since the volume of air which had passed through the cell entered into the calculations; it was also essential to prevent the occurrence of pressure fluctuations within the saturation cell. The design of the apparatus enabled replicate analyses to be performed upon each so-

172

lution placed in the cell. In the case of the most concentrated solutions (200 ppm) up to six analyses could be performed without causing a significant concentration change in the aqueous solution but no more than duplicate analyses were possible for the most dilute solutions (0.02 ppm). As the static method, 2-isopropyl-3-methoxypyrazine showed substantially the same value of partition coefficient, K , in the _3 a/w range 1.8 - 1.9 x 10 , for each aqueous solution concentration studied. Indeed, this value is the same as that obtained using the static method. 2-Ethyl-3-methoxypyrazine showed an increase of partition coefficient from a value of 6.4 x 10~4 -4 at 200 ppm, through 6.2 x 10 at 20 ppm and 2 ppm to a value -4 of 8.2 x 10 at 0.02 ppm (Fig. 4). Thus,values of for 2ethyl-3-methoxypyrazine and the increase of with dilution are smaller than those obtained using the static method. This is consistent with the hypothesis outlined above that, under static conditions, the compound is increasingly concentrated into the surface layer (relative to the bulk solution) as the solution is diluted, thus increasing the measured value of In the dynamic method, one would expect the surface layer to be continually disrupted by the passage of air bubbles which tend to move bulk solution to the surface and to mix the surface layer back into the bulk solution. Thus one would anticipate that the surface layer is much less well formed and relatively ineffective in its influence on and the results from the dynamic experiment are consistent with this interpretation. The reason why 2-ethyl-3-methoxypyrazine showed an increase in the value of with dilution whilst 2-isopropyl-3-methoxypyrazine did not is unclear at present. The value of the partition coefficient for 2-isopropyl-3-methoxypyrazine at all aqueous solution concentrations studied is similar to the largest value observed for 2-ethyl-3-methoxypyrazine - i.e. that of the most dilute solution using the static method. Thus, it could be that a surface layer unrepresentative of the bulk solution is

173

Fig. 4

Variation in air/water partition coefficient as a function of solution concentration (2-ethyl-3-methoxypyrazine, dynamic method)

present for each of the concentrations studied. However, evidence from the dynamic method is contrary to this because very similar values were recorded for the partition coefficient of 2-isopropyl-3-methoxypyrazine by the static and dynamic methods. This is in distinction to 2-ethyl-3-methoxypyrazine for which lower values of K_,yw were observed by using the dynamic method, presumably because of the disruption of the surface layer. The best interpretation appears to be that 2-isopropyl-3-methoxypyrazine does not form an unrepresentative surface layer at the concentrations which were studied; however, it may well do so in more dilute solutions and this possibility has yet to be

174

explored.

Experimental Construction of "static" apparatus. The apparatus (Fig. 1) was constructed with all internal surfaces of PTFE. Standard thick wall PTFE screw cap bottles (Gaflon) were incorporated as so3 lution chamber (150 cm ) and vapour chamber (1 1). A pneumatically operated double-acting bellows type pump (Asti dispensing pump, model PCS1), of which all internal surfaces were of PTFE, was used to circulate vapour and the circulation was controlled by magnetically actuated shuttle valves (also with PTFE surfaces) integral with the pump. The holder for the activated charcoal trap (see below) was a length of PTFE rod (100 mm x 10 mm o.d.) drilled throughout its length to accept the trap which fitted fairly tightly within it. The end of the holder into which the trap was inserted was slightly enlarged, and also tapered, so that a strip of PTFE tape wound around the trap, 1 cm from its end, could be tamped gently into the taper to retain the trap and prevent vapour from flowing past it. The charcoal trap was fitted between two PTFE three-way taps (Gaflon) so that vapour could either be circulated through the trap or trap and vapour could be isolated each from the other. Connections between pump, vapour chamber, solution chamber and three-way taps were made using PTFE tubing and compression couplings. Standard commercial couplings of slightly different type were employed as part of the taps (Gaflon) and pump (Asti) whilst fittings made in the Procter Department Workshop were used to connect the solution and vapour chambers into the system. Activated charcoal traps. Activated coconut shell charcoal

175 (0.2 mg, 120-150 mesh), which had previously been well extracted using diethyl ether, was retained in the required position within a length of Pyrex glass capillary tube (80 mm x 1.0 mm o.d. x 0.4 mm i.d.) by minute plugs of silanized glass wool at either end of the charcoal. It was essential that the minimum quantity of glass wool was used and, in the case of the dynamic apparatus, the traps were prepared to have very similar permeabilities. Use of static apparatus. Prior to introduction of the sample, the apparatus was thoroughly cleaned and assembled with a glass tube (100 mm long, 10 mm o.d. x 6 mm i.d.) filled with activated charcoal in place of the charcoal trap holder and totally immersed in a thermostated water bath (25°C + 0.005°C). Air contained in the apparatus was purified by circulation through the charcoal and it was ascertained that no leaks were present. The apparatus was briefly removed from the water bath 3 and replaced after a charge of 100 cm had been placed in the solution chamber and the charcoal trap in its holder had been substituted for the charcoal-containing glass purification tube. The solution chamber was charged by first adding water followed by the necessary volume either of the undiluted alkylmethoxypyrazine or of a relatively concentrated aqueous solution (>1000 ppm). The vapour was circulated between the solution and vapour chambers for 3 h after which time the threeway taps were turned so that the solution chamber was by-passed and the vapour could be circulated between the vapour chamber and the charcoal trap. After 3 h, adsorption of the alkylmethoxypyrazine on the trap was complete and the apparatus was lifted from the water bath and dried and the charcoal trap was removed and placed in the injection device (see below) in order that the amount of alkylmethoxypyrazine adsorbed could be measured. 3 The total volume occupied by the vapour was 1526 cm 1387 cm 3 was exposed to the charcoal trap.

of which

176

Construction and use of dynamic apparatus. The saturation cell (Fig. 3) was cylindrical in basic form and made of glass. A porosity 4 sinter (20-30 pm pore size) served as the floor of the cell and supported the aqueous solution (100 cm^) of the pyrazine under examination. Highly purified air was passed at 3 -1 20 cm • min through the sinter, the flow being accurately controlled using a mass flow controller. Exhaust gas passed from the cell via a screw cap joint (Quickfit SQ13, silicone rubber compression ring and PTFE seal) which carried a threeway glass tap fitted with a PTFE key. A small, loose, plug of glass wool ensured that, should aerosol formation occur, none would leave the cell. Each of the two upper limbs of the threeway tap carried an activated charcoal trap (as described above) attached by means of a PTFE compression fitting designed and made in the Procter Department Workshop. Thus, the path between solution and trap was entirely of glass and PTFE. The exit from each trap was connected via an individual on-off valve to a tee-junction which led to a mass-flow controller and thence to a vacuum pump; a second tee-junction just before the pump carried an adjustable air bleed. Before alkylmethoxypyrazine solution was introduced into the cell, the air flow into the cell and the exhaust gas flow from the activated carbon trap were each adjusted to 20 cm3 by means of the two mass flow controllers and the air bleed into the vacuum pump. Calibrated rotameters were used for this purpose and these were removed once the required flow rates had been established. During sampling of the exhaust gas, a water manometer was connected between the atmosphere and a side arm connected to the exit from the saturation cell which served to confirm that the pressure within the cell did not differ from atmospheric by more than + 1.5 mm Hg and thus that conditions of flow rate and pressure had remained substantially constant. Once operating conditions had been established, the rotameters were removed, the required solution (100 cm3) placed in the

177

cell and air allowed to pass through the cell and a charcoal trap for 2 min to allow equilibrium to be attained. The exhaust gas was then switched to flow through the alternative trap and pyrazine adsorbed from the gas stream during an appropriate time interval (3-60 min). During this time also, the trap which was in place for the initial 2 min period was replaced by a fresh trap which was then used to perform a duplicate analysis of the exhaust gas. Further replicate samples were collected, if appropriate, and the amount of alky lmethoxypyrazine present on each trap was assayed by gasliquid chromatography as described below. Assay of collected pyrazine by gas-liquid chromatography. The instrument used was a Pye 104 gas chromatograph fitted with a flame ionization detector modified for use with capillary columns. A glass SCOT column (65 m x 0.5 mm i.d.) containing Carbowax 20M as stationary phase and Silanox 101 as the support was coupled to an injection device which was a modification of that described by Clark and Cronin (9) and designed to accept a charcoal trap and transfer the adsorbed volatiles directly to the capillary column. The trap was placed in the cool region of the injector, the carrier gas (nitrogen) allowed to -1 3 flow and establish equilibrium at 4 cm • min and the trap then lowered into the heated zone so that the adsorbed volatiles were rapidly desorbed and flushed onto the column. A temperature programme, isothermal at 90°C for 12 min, rising by 49°C«min~1 to 135°C and remaining at this temperature to the end of the analysis, was employed and the area of the alky lmethoxypyrazine peak computed by triangulation and compared with that given by an external standard. Acknowledgements

The authors wish to thank Professor Harry Nursten, who initi-

178

a t e d the p r o j e c t , for h i s c o n t i n u e d e n c o u r a g e m e n t a n d the S c i e n c e R e s e a r c h C o u n c i l for a m a i n t e n a n c e g r a n t

(to D . R .

Williams).

References 1. B u t t e r y , R . G . , S e i f e r t , R . M . , L u n d i n , R . E . , L i n g , L . C . : C h e m . Ind. (London), 490-491 (1969). 2. S e i f e r t , R . M . , B u t t e r y , R . G . , G u a d a g n i , D . G . , B l a c k , D . R . , Harris, J . G . : J. a g r i e . F d C h e m . , J[8, 246-249 (1970). 3. B u t t e r y , R . G . , G u a d a g n i , D . G . , L u n d i n , R . E . : J. a g r i e . C h e m . , 24, 1-3 (1976) .

Fd

4. B u t t e r y , R . G . , S e i f e r t , R . M . , L i n g , L . C . : J. a g r i e . F d C h e m 1 8 , 538-539 (1970). 5. G u a d a g n i , D . G . , B u t t e r y , R . G . , S e i f e r t , R . M . , V e n s t r o m , J. F d S e i . , 36, 363-366 (1971). 6. M u r r a y , K . E . , W h i t f i e l d , F . B . : J. S e i . F d A g r i e . , 26, 986 (1975). 7. N u r s t e n , H . E . , Sheen, M . R . : J. S e i . F d A g r i e . , 25,

D.M

973-

643-663

(1974) . 8. B u r n e t t , M . G . : A n a l . C h e m . , 35^, 1 5 6 7 - 1 5 7 0

(1963).

9. C l a r k , R . G . , C r o n i n , D . A . : J. S c i . F d A g r i e . , 26, (1975) .

1615-1624

T H E CONTRIBUTION OF SOME VOLATILES TO THE SENSORY QUALITY O F APPLE AND ORANGE JUICE ODOUR

P. Dürr, U. Schobinger Eidg. Forschungsanstalt für Obst-, Vein- und Gartenbau CH-8820 Wädenswil, Switzerland

1. Introduction

About 330 volatiles found in orange fruit and its products and about 230 volatiles in apple fruit and its products have been listed (l). These figures correspond to the economic importance and the work done on the respective fruit and its products-.With these figures in mind, the contribution of a single volatile component has to be considered. Another problem: the aroma of a fruit juice undergoes rather strong changes once a fruit is crushed and during juice processing and storage. The objects of research are fresh juice aromas, essences and oils, which are recovered from the fresh juice during its concentration and the aroma of reconstituted, packed and stored juices. What are the possible contributions of a single volatile to the sensory quality of a fruit juice aroma? - it contributes strongly to the typical odour of the juice. Such a volatile is called a '"character impact compound" ( 2 ) ,

e.g. citral in lemon;

- it is part of the typical aroma, it contributes and it is desirable, e.g. sinensal in orange juice aroma; - it is responsible for an off-odour or aftertaste and therefore undesirable; this is called an off-flavour component, e.g. oL-terpineol in orange juice; - it acts as a precurser for off-flavour components, e.g. limonene in orange juice; - it contributes to the intensity of an aroma, e.g. trans-2-hexenal in apple juice aroma;

© 1981 by W a l t e r d e G r u y t e r & C o , B e r l i n • N e w Y o r k F l a v o u r '81

180 - it does not contribute directly to the aroma but is related to aroma quality, e.g. ethanol, isobutanol; - it does not contribute at all due to a concentration level far below sensory threshold.

2. Orange juice volatiles 2.1. Contribution to the typical aroma Lists of volatile compounds found in orange juice have been published by SHAW (3) and SCHBEIER etal ( 4 ) . A concise review on the aroma of orange juice and its changes during storage was given by BIELIG etal (5). The most comprehensive study on the contribution of single volatiles was carried out by DOUGHERTY and AHMED ( 6 ) , WRIGHTSON (7) and SCHINELLER (8) in Florida. They determined thresholds of 33 known orange juice components in water and evaluated some of these components singly or in combination. The components were added to evaporated juice in concentrations as found in orange juice. Compounds tested were: acetaldehyde

citral

d-limonene

octanal

citronellal

linalool

nonanal

trans-2-hexenal

decanal

ethyl-vinyl-keton

oL-pinene

myrcene

dodecanal

ethylbutyrate

/3-sinensal

Highest ratings were received by the following combinations: a) d-limonene ethylbutyrate citral

b) d-limonene ethylbutyrate citral acetaldehyde

c) d-limonene ethylbutyrate citral acetaldehyde o(.-pinene

d) d-limonene ethylbutyrate citral ot-pinene

e) d-limonene ethylbutyrate nonanal

f) d-limonene ethylbutyrate oC-pinene

181

Addition of either decanal, citronellal or trans-2-hexenal lowered ratings. Two components axe part of all 6 mixtures: ethylbutyrate and d-limonene. Pure ethylbutyrate has a neutral fruity odour with a pineapple undernote, slightly rancid. Its odour threshold is in the ppb range. But its level in fresh orange juice is near 1 ppm. Quantitatively it is one of the main components, d-limonene, which is quantitatively dominating with 10-300 ppm, contributes also to the typical aroma. But it is more important as a precurser of undesirable components. The aldehydes are the most important components to the typical aroma. The isomers of citral, neral and geranial, are character impact compounds for lemon flavour and form part of the orange juice aroma with rather high levels of 100-200 ppb in fresh juice. Other aldehydes reported to contribute are octanal, nonanal and the sesquiterpene isomers cL- and/i-sinensal.

«. -Terpineol

o

glass o

soft

90

Pig. 1:

days

Increase of ot-terpineol in orange juice during storage in cardboard package and glass bottle

182

2.2. Deterioration of aroma

One component has long been recognized as critical to orange juice aromas oC-terpineol (9). Pure oC-terpineol has a lilac like odour. When added to orange juice in the amount of 3 ppm, the juice aroma was described as terpeney, old, partly rancid (5). The threshold of OL-terpineol in orange juice is in the range of 1 ppm (5). oC-terpineol and related compounds are formed from limonene and linalool in a nonoxidative, acid catalized pathway during storage of orange juice. Due to this regular, temperature dependent increase of oC-terpineol during storage, ASKAR etal (10) recommended the level of oC-terpineol as a quality parameter of orange juice. TATUM etal (ll) isolated 10 degradation compounds from canned juice stored at 35°C for 12 weeks. Three of these compounds , oC-terpineol, 4-vinylguaiacol and 2,5-dimethyl-4-hydroxy~3-(2H)furanone exhibited malodourous properties. Their taste panel described the odour of oC-terpineol, when added to fresh juice, as stale, musty or piney. 4-vinylguaiacol imparted an "old-fruit" or rotten flavour to the juice. 2,5-dimethyl-4-hydroxy-3-(2H)-furanone is responsible for the pineapple like odour of aged orange juice. "Terpeney" off-flavours have also been attributed to carvone and carveol formed by the oxidation of d-limonene (12). The several pathways of aroma deterioration have been reviewed and studied comprehensively by BIELIG etal (5).

183

2.3» A shelflife experiment

In a recent experiment of our own juice from polyethylene lined cardboard package was compared with the same juice from glass bottles. Other variables were storage time, storage temperature and peel oil content (100 ppm, 200 ppm added to reconstituted juice). It was known that orange juice, filled in cardboard package, looses a part of its limonene within a few days by absorption into the polyethylene lining of the package. The consequences on the aroma of the juice have been evaluated. The decrease of limonene was 40

within 6 days after filling into the cardboard package,

but sensorically no difference was detectable in comparison to the glass bottled juice. The increase of o£.-terpineol was linear over 90 days at 20°C and stronger in the glass bottled juice (Fig. l).

ppb

Neral

Fig. 2: Decrease of neral in orange juice during storage

184

The increase depends much stronger on the storage temperature than on the initial limonene content of the juice. After 13 days at 32°C the juice was described as stale and musty, after 27 days as rotten, detrimental. The desirable aldehydes neral, geranial and octanal decrease depending on the storage temperature and independent from the package (Fig. 2,3).

Fig. 3: Decrease of octanal in orange juice during storage

2.4. Summary

Table 1 represents a summary of this review on orange juice aroma. The volatiles are classified according to data from quantitative and sensorial analysis.

185

Table Is

The contribution of volatiles to orange juice aroma

contribution to typical aroma

contribution to off-aroma

important

precursers

e thylbutyrate neral geranial

desirable linalool limonene oi.-pinene valencene acetaldehyde octanal nonanal oL-sinensal /3-sinensal

linalool limonene valencene

detrimental oi-terpineol carvone trans-carveol nootkatone hexanal trans-2-hexenal hexanol 4-vinylgnaiacol 2,5 -dimethyl-4hydxoxy-3-(2H)furanone

3. Apple juice volatiles 3.1. A short review In 1964, KOCH and SCHILLER (13) first recognized the sensory value of trans-2-hexenal to apple juice aroma. This was achieved by sniffing GLCeffluents and chemical décomposition of aldehydes, followed by a reconstitution with synthetic trans-2-hexenal, where the original apple juice odour was nearly reestablished. FLATH etal ( 1 4 ) also separated the aroma with GLC and evaluated the effluents by sniffing. They determined the odour threshold of a number of volatiles and described hexanal, trans-2hexenal and ethyl-2-methylbutyrate as important contributers to apple juice aroma. During the same time, in the mid-sixties, DRAWERT etal (15) studied the biochemical formation of these C6-aldehydes from linolic and linolenic acid. The aldehydes are formed immediately after breaking the fruit tissue. Later, DRAWERT, SCHREIER and others investigated the aroma composition of apple and apple products (16,17,18,19). KOCH etal (20) correlated the sum of 2-hexenal and 2-hexenol to the intensity of apple juice essence. In a study on the quality of Mcintosh apples, PANASIUK etal

186 (2l) compared. GLC-data with odour description data of fresh juice. Overall intensity and the aroma notes ripe, aromatic and fruity correlated to the level of C6-aldehydes and the notes overripe, cheesy and butyric were related to the esters.

3.2. Commercial apple juice essence 3.2.1. The data. DUEER (22) analized 48 samples of commercial 150-fold apple juice essence. The concentration of 30 volatiles was measured and 9 sum variables were formed from this data. Sensory analysis included magnitude estimation of essence odour intensity (23) and verbal profiling of essence odour with 18 descriptive words and one hedonic word ( 2 4 ) . 3.2.2. Odour intensity. The data has been treated in various ways. A simpie correlation step indicated, that the unsaturated aldehydes and fermentation volatiles like isobutanol and the corresponding acetate are important to the odour intensity of apple juice essence (Table 2). Table 2:

Correlation between volatile levels and odour intensity of apple juice essences volatiles

correlation coeff.

total carbonyls trans-2-hexenal total unsaturated volatiles total volatiles i-butylacetate i-butanol

0.79 0.78 O.78 0.76 0.72 0.71

The relation between the concentration of trans-2-hexenal and perceived odour intensity is shown in Fig. 4- As the data is measured and plotted in ratio scales, the exponent n of Steven's power law (25) is shown to be 0,6: I = c0-6 Intensity = Concentration

0 6

187

The similar plotting for isobutanol, a fermentation product, gives Fig. 5The power exponent n is approx. 0.5, so this volatile seems to be less potent to the intensity than 2-hexenal.

2-hexenal

(0

c © 4-» c o> o

eorr.

0.78

log conc. Fig. 4s

trans-2-hexenal content v s odour intensity of apple juice essences

3.2.3. Hedonic value.

The volatiles can roughly be classified according

to their chemical nature in higher alcohols, esters and carbonyls. For each essence, the relative amounts of the 3 groups, together 100 fo, can be plotted twodimensionally in a triangel: Fig. 6 . T h e 4 5 essence samples are separated in 3 equally sized groups according to their hedonic value. B a d essences are characterized by a high level of alcohols and a low level of carbonyls, good essences by a low level of alcohols and a high level of carbonyls or in a few cases of esters.

188

i-butanol

corr. 0.71 e

o

log conc. Fig. 5:

isobutanol content vs odoux intensity of apple juice essences

Table 3 gives some correlation data of volatile levels to the hedonic values. Tab. 3:

Correlation between volatile levels and hedonic value of apple juiee essences

volatiles total unsaturated volatiles isobutylacetate cis-3-hexenal trans-2-hexenal trans-2-hexenol ethylbutyrate cis-3-hexenol

correi, coeff. 0.60 0.60 0•54 0.49 0.49 0.49 0.44

189

It should be noted that all volatiles mentioned in table 3 are C6-components and, exept the esters, unsaturated.

Carbonyls °/
/(20)a)

a

) highest c o n c e n t r a t i o n

tested

Fig. 6: Taste of amino acids w i t h b r a n c h e d side chains

(7, 9)

RR1C(NH3)+COOquality

Me Me Et Pr a

Me Et Et Pr

sweet sweet not sweet not sweet

) highest concentration

h y d r o p h o b i c pocket pound

threshold (mmol /I) 8 18 (100)a) (100)a)

tested

in a d i f f e r e n t p o s i t i o n . T h e

I sits s i d e w a y s

ring of

in the p o c k e t but for c o m p o u n d

d e s p i t e the fact that the n s / e g - S y s t e m i n t e r a c t s same way the ring can also sit on its other axis.

in the This

a l t e r n a t i v e p o s i t i o n allows the m a x i m u m , and t h e r e f o r e g e t i c a l l y m o s t favourable,

h y d r o p h o b i c c o n t a c t . Thus,

h a v e to c l a s s i f y the a m i n o s u l p h o n i c a c i d s as type 2 in f i g . l ; this type s t i m u l a t e s e x c l u s i v e l y

sweet

com-

III,

taste.

enerwe

748 For the 2-amino acids with a chain branch at C

3

both sweet

arid bitter tastes are possible. For example, D-valine and L-valine

(fig. 8: XlVa, XlVb). However, a third methyl

stituent

in the 3-position, as for example in tert.

implies the complete loss of stimulant activity

sub-

leucine,

(fig. 8: Xa,

Xb).

Amino acids with bridged ring systems in their side chains allow a more detailed

description of the

hydrophobic

pocket

of the schematic receptor. The adamantane derivative, (fig.9: IX) is neither sweet nor bitter but the derivatives of bicyclo

(3.2.1.) octane

(3.3.1) nonane

(fig. 9: XIa, Xlb) and bicyclo

(fig. 9: XII) although not sweet are bitter.

Fig. 9

Fig. 8

m j

55 23 1

tested

11: Charge d i s t r i b u t i o n ~ PhX (9,14) charge X

a

sweet not sweet not sweet not sweet

halides,

threshold (mmol/l) 25

(fig. 1 2 ) the n s / e g - s y s t e m

an effective

r e c e p t o r . H o w e v e r , the fact that

interaction iodoalkanes

with

is

the

are m o r e

hydro-

751

phobic than the chloroalkanes

implies that where the alkyl

group is the same an iodoalkane has a much lower sweet threshold

taste

value.

Fig. 12: Sweet taste threshold values (1, mmol/1) and partition coefficients (2, log P, octanol/water) of alkyl halides,(-: not sweet, cH: cyclohexyl) (9, 14) X/R 1 2 Cl 1 2 Br 1 2 I l 2

Me

Et

Pr

Bu

Pe

50 1.3

18 ]. 8

10 1.8

8 2. 3

12 2.3 13 2.5 5 2.9

He

cH

F

50 1. 3

6 2.9 -

Finally, some AMADORI- and HEYNS-compounds are

interesting

from a sensory point of view because they are per se a combination of a sugar and an amino acid

in the same molecule.

From fig. 13 it can be seen that all the N-alkyl derivatives of 2-desoxy-2-amino-D-glucose are sweet independent of the configuration of the amino acid moiety. The threshold

values

are at least a factor 2-3 less then that for D-glucose. In contrast the N-alkyl derivatives of 1-desoxy-1-amino-D-frue tose are not sweet.

From the work of BIRCH (6) it is known that the n s / e g system of glucose

(fig. 14: XVII) is made up of the proton

of the hydroxy group at C^ and the oxygen of the hydroxy group at C . It is also known that the n g / e g - s y s t e m of fructose

(fig. 14: XIX) operates through the proton of the 2

hydroxy group at C C1.

and the oxygen of the hydroxy group at

752

The t a s t e

of

the AMADORI- and HEYNS-compounds

the above model. b)

it

is clear

With the HEYNS-compounds

that

as in g l u c o s e .

the b i p o l a r

The amino a c i d

both the D- and L - i s o m e r s , hydrophobic threshold relative

interaction

value f o r

it

(fig.

is destroyed

as e x p e c t e d ,

As i s

shown in f i g .

xvii

1:

/ «» H

•» H OH

HO

H

and

bitter

of

14

14:

at XX).

taste.

the AMADORI-

w i t h our model.

Fig.

SWM XX

by RNH ( f i g .

qualities

in accord

XV

n

the

lowered

In the AMADOR I-compounds the

14 the t a s t e

OH

a,

for

type 2) and t h e r e f o r e is considerably

because t h e hydroxy group

XVIIIQ

XVIII

t o have a g r e a t e r

t h e s e compounds have no sweet

and HEYNS-compounds a r e

XIX

14:

s i d e c h a i n means t h a t , is possible

the carbon atom 1 has been r e p l a c e d Thus,

(fig.

i n t e r a c t i o n may be t h e same

t h e compound

to the sugar a l o n e .

ng/eg-system

corroborate

753

Fig. 13: Taste of a)

a) -NH-R

AMADORI-

(a) and HEYNS-compounds (b)

CHo-NH-R

I C = 0 • 1

D-Val L-Val

CHO

b)

I H-C -NHR 1 1 threshold (mmol/1)

quality

tasteless tastele ss

b) -NH-R D-Val L-Val D-Leu L-Leu D-Phe L-Phe

sweet sweet sweet sweet sweet sweet

20-40 20-30 20-40 20-30 50-60 50-80

D-glucose sucrose

sweet sweet

70-80 10-12

Fig. 16

XV

XVII OH

XXI

754

The example of D-glucose clearly shows that polar groups above the plane of the cyclohexane ring, for example compound XV, do not inhibit sweet taste in contrast to the ef f ects of the apolar groups in compounds

XI and XII which

have already been discussed.

Fig. 15: Taste of glucosides and (BIRCH, 1976) OC-Glcp-R sweet bitter

R

+ + +

H OMe OEt OPr OBu OPh 0CH 2 Ph a

)

+ ++ ++ + ++

fructosides a)

R-Glcp-R sweet bitter + +

(+)

+

ß-Frup-R sweet bitter

+

+

(+) +

++ ++ + ++

+ ++

(+) : weak response, +: normal response, ++: intensely response, -: no response

Introduction of apolar groups above the plane of the cyclohexane ring converts a sweet tasting molecule

into a bitter

one. This was shown by BIRCH (6) with the 0^-and 13-alkyl glucosides: B-methylglucoside

is sweet and bitter, B-ethyl-

glucoside is bitter and slightly sweet,

B-propylglucoside

has a strong bitter taste (fig. 15). Using our

sweet-bitter

receptor model it can be suggested that introduction of a hydrophobic group at C^ causes the molecule to twist, such that the extra hydrophobic chain can enter the hydrophobic pocket

(fig. 16: XXI). In this way the bipolar contact is

converted

to a monopolar one which functions through one of

the oxygen atoms. The taste is thus converted bitter .

from sweet to

755

References : 1) Shallenberger, R.S., Acree, T.E.: Chemical structure of compounds and their sweet and bitter taste. In: Beidler, L.M.(Ed.): Handbook of Sensory Physiology 4/2. Springer-Verlag : Berlin-Heidelberg-New York, 1971, p. 221 2) Kier, L.B. : J.Pharm.Sei.

61, 1394, 1972

3) Beidler, L.M. : Biophysics of Sweetness. In: Inglett, G.E. (Ed.): Symposium Sweeteners. AVI Publ.Co. : Westport, Conn. 1974, p. 10 4) Belitz, H.-D., Chen, W., Jugel, H., Treleano, R., Wieser, H., Gasteiger, J., Marsiii, M.: Sweet and bitter compounds: Structure and taste relationship. In: Boudreau, J.C. (Ed.).: Food Taste Chemistry, ACS Symp. Series No. 115, 1979, p. 93 5) Pautet, F., Nofre, C.: Z.Lebensm.Unters.Forsch. (1978) 6) Birch, G.G.: Crit.Rev.Food

166, 167

Sci.Nutr. 8, 57 (1976)

7) Wieser, H., Jugel, H., Belitz, H.-D.: Z.Lebensm.Unters.Forsch. 164, 277 (1977) 8) Wieser, H., Belitz, H.-D.: Z.Lebensm.Unters.Forsch. 65 (1975)

159,

9) Treleano, R., Belitz, H.-D.: unpubl.results 10) Treleano, R, Belitz, H.-D., Jugel, H., Wieser, H.: Z.Lebensm.Unters.Forsch. 167, 320 (1978) 11) Jugel, H., Belitz, H.-D.:

unpubl.results

12) Chen, W., Belitz, H.-D.: unpubl.results 13) Stempfl, W., Belitz, H.-D.:

unpubl.resu1ts

14) Belitz, H.-D., Chen, W., Jugel, H., Treleano, R., Wieser, H.: Comparative studies in the taste of amino-, hydroxy-, halogen- and nitro-compounds: relations between taste and structure. In: van der Starre, H. (Ed.): Proc. VI Ith. Intern. Symp. Olfaction Taste, IV. Congr. Europ. Chemorec.Res.Org. IRL Press Ltd.: London, 1980, p.45

BIFUNCTIONAL UNIT CONCEPT IN FLAVOUR CHEMISTRY

G. Ohloff Firmenich SA CH-1211 Geneva 8, Switzerland

The aroma compounds referred to as 'browned flavours' (1) are part of that class of flavour substances which has been the only one so far to show a marked relation between molecular structure and odour. o

H

1

Indeed, Hodge postulated that the caramel-like odour

was

a

function of a specific molecular feature, which is a planar enol-carbonyl sub-structure of cyclic dicarbonyl compounds, which are capable of forming a strong hydrogen bond (2). Typical examples are maltol [J_] and isomaltol [2^. The a-diketone sub-structure of the pyranoid maltol J_ consists of a single tautomer whose hydrogen-bond is arranged in a five-membered ring, whereas in the g-diketone isomaltol [2J the hydrogenbond is part of a six-membered ring. In this paper I will try to investigate the frequency of naturally occurring flavour compounds whose particular molecular feature is a bifunctional unit consisting of

a proton-donor

and a proton-acceptor group. Several such compounds formed as a result of non-enzymatic browning reactions during food processing were shown to be base-catalysed degradation products of sugars. There is, however, an increasing number

of

aroma

compounds becoming known that contain a bifunctional unit iso-

© 1981 by W a l t e r d e G r u y t e r &. C o , Berlin • N e w Y o r k Flavour '81

758

lated from unprocessed flavour systems. This would imply that they are generated by a biogenetic pathway.

XX XX XX OC W

"

w

W

Xl

OH

uC

~

0

SH

OH

10

AX/

Ji^^OH

HO^

11

0K

12

"

.0

^OH

O^ ^OH

/—\

13

14

15

The aroma compounds isolated from browned flavours have a rather complex odour profile and are mainly associated with a caramel-like, sweet, nutty and burnt-sugar odour. For every single compound of this type known so far the odour profile proved to vary, although within limits. This variation is obviously caused by the hydrophobic part of the molecule, a fact which becomes evident in compounds

-

(3) . Substitution of

the oxygen by sulphur - as in compounds 1_ ~ 2. (4a, 5, 6) does not lead to a significant modification of the odour except for the burnt note becoming more pronounced. Compound 1£, however, is virtually odourless at room temperature (4k). The fact that the odour strength increases if a methyl group is substituted by an ethyl group - as in compounds T_l_ - J_5 - is accounted for by the increased lipophilic character of the homologues.

o

16

OH

o

0>

17

Or J i ^ ^OH

18

0S

OH

19

In unsubstituted a-diketones the pronounced caramel odour disappears - as in pyromeconic acid 16 (2) and 4-hydroxy-2#-furan-

759 3-one [J_7] (7) , or weakens considerably - as in 2-hydroxy-2cyclohexen-l-one one [19] (8).

[JJ8] (8, 9) or in 2-hydroxy-2-cyclopenten-l0 OH

^CT

20

22

23

If the methyl group in the y-pyrone ring is not situated next to the enolic hydroxyl group - as in allo-maltol 2_0 (2) - the caramel odour is also lost. This does not apply to 4-methyl-2hydroxy-2-cyclopenten-l-one

[2_1_] (4b) which smells strongly of

roasted walnuts, fenugreek, maple and lovage. Adding a double bond in the five-membered ring - as in 2_2 (4a) - considerably weakens the odour. Although

3-methyl-2-hydroxy-2,4-cyclopenta-

dien-l-one [22] has a caramel and maple-like odour, it does not attain the intensity of its dihydro derivative Cyclotene [23] (10).

o

o

24

25

There are no experimental data concerning the necessity for and the function of the enol-carbonyl sub-structure in a-diketones. We have found that 2 , 2 , 5 ,5-tetramethyl-3# , 5ff-furan-3 , 4-dione [2*t] is odourless

(11). In contrast, the known

(+)-camphorquinone

[25] has a pleasant but weak flowery-spicy odour with a celluloid base, which would be expected from oxygen-containing monoterpene derivatives, but none of the usual burnt notes. Blocking of the hydrogen bridge in a-diketones transforming the enol into an ether or an ester entails the disappearance of the typical burnt flavour. O-Methylmaltol [26^] has a weak and slightly fruity odour

(4c), whereas 2,5-dimethyl-4-meth-

7 60

oxy-2#-furan-3-one o

[^7] has an aroma characteristic of sherry

OC

XX tx 0

26

wine

0CH 3

0.

27

X 0R

28 29

R = Me

R.Ac

(12). The odour of methyl ether 28 and acetate 29 differs (r) — —

completely from that of C y c l o t e n e w

[23_] (4a). Its pentyl ether

50 even has jasmonoid odour properties (3). Indeed, the number of carbon atoms and the shape of the molecule of 3_0 are the same as dihydro-jasmone

[31]

(4d).

R 32 33 34

R a Me 35 R - Et 36 R =• Prop 37

Little is known about the olfactory properties of such as isomaltol

[2] (2), whose planar

g-diketones

alkyl—enol—carbonyl

arrangement can form a six-membered hydrogen-bridged ring. Both 2-acetyl compounds derived from cyclopentanone hexanone

[32_] and cyclo-

[ 3_5 ] have been described as Cyclotene-like

[2_3] odo-

rants whose minty, herbaceous character becomes more pronounced as the number of carbon atoms increases

[33_ and 3_4,

and 57 ]

(13). A burnt, phenolic odour was attributed to the keto-aldehyde 3_8 (4e) . In contrast p-menthane-2,6-dione less

is odour-

(14), a fact which could probably be accounted for by the

impossibility of forming an intramolecular hydrogen bond. 'Browned' flavour components are not confined to singly

substi-

tuted cycloalkane-1,2-diones such as 6. Derivatives with higher substitutions can also fit into this system, as shown by 40 -

761

^-^OH

O ^ f ^ OH

40

41

HO-^^

&

O

42

43

The odour of 3,5,5-trimethyl-2-hydroxy-2-cyclopenten-l-one ['tO ] (15) resembles that of the singly substituted compound 25, whereas that of

3,5,5-trimethyl-2-hydroxy-2-cyclohexen-l-one

[^1] (16) was described as straw-like, kerosene another double-bond to

(13). By adding

(4f) the phenolic character becomes

stronger, and maltol- and Cyclotene-like odour qualities appear. The diosphenols

and ^jk (17) existing as two stable tautomers

have a strong phenolic note of herbal-minty tonality. The dominant flavour characteristics of compounds 4_0 -

are conse-

quently the burnt notes, even though some of the odour qualities of monoterpenoid compounds are unmistakably present. All the above examples stay within relatively narrow limits as far as their molecular structures are concerned. Indeed, in accordance with Hodge 's postulate we only investigated diketones having the basic skeleton of either cyclopentane or cyclohexane, as well as derivatives of oxa-analogues.

c6