Chem. Senses 27: 485-494,
2002
© Oxford University Press 2002
Sensory Evaluation of Character Impact Components in an Apple Model Mixture
1 Department of Agrotechnology and Food Sciences, Laboratory of Food Chemistry, Wageningen University and Research Centre, Wageningen, The Netherlands 2 Department of Industrial Design, Delft University of Technology, Delft, The Netherlands 3 Psychological Laboratory, Utrecht University, Utrecht, The Netherlands 4 Wageningen Taste and Smell Centre, Wageningen, The Netherlands
Correspondence to be sent to: A.G.J. Voragen, WUR, Department of Agrotechnology and Food Sciences, Laboratory of Food Chemistry, Biotechnion, PO Box 8129, 6700 EV, Wageningen, The Netherlands. e-mail: fons.voragen{at}chem.fdsci.wag-ur.nl
 |
Abstract |
|---|
 Top Abstract IntroductionPilot studyMain ExperimentResultsDiscussionReferences |
|---|
Food aromas generally are complex mixtures of volatiles. In the present study, we investigated the joint effects of hexyl acetate, trans-2-hexenal and 1-hexanol on the multi-attribute perception of an apple aroma. The first two substances were identified earlier as positive contributors to the apple aroma (high character impact), whereas the third component was identified as an irrelevant or negative contributor (low character impact). Aroma quality was quantified using a set of eight graphic rating scales. All three components had significant effects on the aroma profiles. These effects consist mainly of an effect of each component on the attribute that described its individual character and an effect of all three components on ratings on the main character attribute `apple'. As expected, the high impact components increased `apple' ratings, whereas the low character impact component decreased `apple' ratings. Furthermore, intensity ratings on the attribute that corresponded with the odour of the low impact component were suppressed by the presence of high impact components. These results indicate that the contributions of odorants to the mixture's aroma are not linear combinations of separate odour intensities, because sensory interactions were observed. In addition, humans detect components in complex mixtures more accurately than studies on identification performance have suggested. We conclude that for an adequate assessment of the effects of multiple mixture components on changes in aroma perception, it is sufficient to employ multiple response scales measuring intensities of attributes that are distinctive with respect to the expected qualitative changes. Results of this approach should be subjected to multivariate methods of statistical analysis.
|
Introduction |
|---|
|
TopAbstractIntroductionPilot studyMain ExperimentResultsDiscussionReferences |
|---|
Food aromas generally comprise extensive mixtures of volatile constituents. A large number of these constituents produce odours if presented alone at similar concentration levels. Others, however, might not produce noticeable odours at all. An important objective in aroma research is to minimize the number of components in a modelled aroma by selecting those volatile components that contribute most to the original aroma. In general, this selection is made on the basis of two sensory properties: the relative perceived intensity of that component presented in isolation and the extent to which its character resembles the quality of the particular food aroma. Both higher perceived intensity and higher typicality of a component's odour quality result in a higher character impact of the component on the mixture's aroma (Buttery and Ling, 1998
).
Character impact components (CICs) are usually identified by sensory
analysis of mixture constituents after decomposing the mixture by gas
chromatography (Dravnieks and O'Donnell,
1971). The constituents are evaluated with regard to their unmixed
odour qualities and intensities. In doing so, one disregards that the
contribution of an odorant to the mixture's aroma depends not only on its
sensory characteristics when presented in isolation, but also on sensory
interactions that occur when the odorant is perceived in the presence of other
components.
Perceptual mixture interactions
In olfaction, partial mutual masking of mixture components is the most
commonly observed interaction, even in mixtures consisting of as few as two
components (Cain, 1975;
Laing and Willcox, 1987;
Lawless, 1987;
Laing et al., 1994).
If one knows a component's psychophysical function, which relates component
concentration to its perceived intensity, the intensity of a `mixture' of that
component with itself can be predicted from the sum of the two respective
concentrations. However, the intensities of binary mixtures of two different
supra-threshold components are often lower than expected on the basis of their
respective psychophysical functions
(Moskowitz and Barbe, 1977;
Berglund and Olsson, 1993a,
b). On the other hand,
indications of synergetic effects were observed when mixing subthreshold
components (Guadagni et al.,
1963; Laska and Hudson,
1991).
The mutual perceptual suppression of odorant intensities in multi-component
mixtures is often observed (Moskowitz and
Barbe, 1977; Moskowitz,
1979). Cain hypothesized that the masking power of supra-threshold
odorants is positively related to either the chemical or the perceptual
complexity of the masker (Cain,
1975). In the taste modality, the masking power of two substances
in concert was indeed observed to be larger than the masking power of each of
the substances alone (Stevens and
Traverzo, 1997). In the case of olfaction, Laing and co-workers
demonstrated that humans perform increasingly worse with increasing numbers of
masking components when they are asked to identify odorous constituents in
mixtures (Laing and Francis,
1989; Laing and Glemarec,
1992; Livermore and Laing,
1996; Jinks and Laing,
1999). Even the seemingly easy task of identifying the qualities
of odorous constituents in binary mixtures yields probabilities of correct
detections far below perfection (Olsson,
1994). A similar relationship between mixture complexity and
masking power was observed when the mixture components themselves were
multicomponent mixtures, each mixture representing a familiar object odour
(Livermore and Laing,
1998).
Although humans experience great difficulties in recognizing the
contribution of single components to the aroma of complex mixtures, they are
able to discriminate between complex mixtures of odorants that are identical
except for one component (Laska and
Hudson, 1992). This can be explained by assuming that some or all
of the odorants in a complex mixture blend perceptually into an intrinsically
new aroma (Livermore and Laing,
1998). The omission of components from a complex mixture may then
be detectable as a change in aroma quality, but not as an omission as such.
Food aromas as well as many other object-related aromas generally consist of
complex mixtures of odorants that, nonetheless, are perceived as homogeneous
aroma blends. It is, therefore, rather speculative to assume that omitting
components from a mixture would only affect the perceived intensity of their
respective characters in the aroma of the mixture.
In recent years, scholars at the Deutsche Forschungsanstalt für
Lebensmittelchemie have recognized the relevance of studying the contribution
of CICs in the mixture. In a number of studies they evaluated the impact of
components on mixture aroma by assessing the effect of omitting these from the
mixture (Blank et al.,
1992; Guth and Grosch,
1994; Schieberle and Hofmann,
1997; Reiners and Grosch,
1998). In a study on the aroma of french fries
(Wagner and Grosch, 1998), the
authors determined the components with high ratios of mixture concentration
versus detection threshold concentration, called odour activity values (OAVs).
Omitting these supra-threshold components from a model mixture often resulted
in a significant discrimination of the aromas of the reduced versus the
complete mixture. When reduced and complete mixtures were significantly
perceived as different, panellists characterized the aroma qualities of the
mixtures by rating intensities of attributes describing the odours of mixture
components. In this study, omitting the component with the second highest OAV
from the mixture was not detected in the discrimination task. After subsequent
omission of additional components, however, a number of panellists gave higher
ratings on the attribute describing the component with the second highest OAV.
This post hoc evaluation suggests that this component was perceived
only after it had been released from suppression due to a number of masking
components. Indications for `release from suppression' effects were also found
in similar studies on an Arabica coffee model
(Czerny et al., 1999)
and a white wine model (Grosch,
2000).
Evaluating aroma differences
The effect of component concentration on aroma quality can be quantified by measures of discriminability between mixtures, by similarity ratings, or by ratings on attributes describing the aromas. However, these methods vary with respect to their sensitivity to differences in aroma quality and their ability to characterize aroma quality.
If one only wishes to test hypotheses with respect to perceptual
discriminability of stimuli, discrimination tasks may suffice. Trials in
discrimination tasks yield binaryscaled results: a subject either does or does
not distinguish correctly between differing stimuli. Proportions of correct
stimulus discriminations can be calculated from repeated stimulus comparisons
and are tested against the expected chance proportion of a correct
discrimination. A psycho-physical application of this method conceives the
probability of correctly detecting a difference between stimuli as a measure
of the sensory difference between stimuli
(Thurstone, 1927;
Frijters, 1980). It expresses
sensory difference as the perceptual distance between stimuli on an arbitrary
sensory continuum. Such measures of perceptual distance have been used to
detect mixture interactions (Lawless and
Schlegel, 1984).
Similarity ratings are used to measure the degree of sensory similarity between stimuli using discrete or continuous rating scales. Hence, the rated dissimilarity of stimuli can be conceived of as a measure of the distance between stimulus representations on an arbitrary sensory continuum. This sensory continuum may either represent a quality continuum or an intensity continuum, depending on the nature of the difference between stimuli.
Although discrimination tasks and similarity ratings are sensitive in
detecting differences in both intensity and quality, they do not allow for
semantic interpretations of results in terms of odour quality
characterizations. Therefore, to study the qualitative nature of sensory
interactions, methods that directly address odour quality are needed.
Attribute ratings reflect perceived intensities of odour characteristics
indicated by odour quality descriptors. This makes attribute ratings an
adequate tool to study and describe mixture interactions in both qualitative
and quantitative terms. However, when the descriptor set is not distinctive
with respect to the characteristic on which the stimuli differ, attribute
ratings may have less discriminative power than similarity ratings
(Callegari et al.,
1997). This may explain why Lawless and Schlegel
(Lawless and Schlegel, 1984)
found a taste—odour interaction in mixtures with variable sucrose and
citral concentrations when using sensory distances calculated from
discrimination task results, whereas no interaction was observed when
attribute ratings were used. Intensity ratings on the attributes `sweetness'
and `lemon odour' were merely statistically additive for the used stimulus
set. In a meta-analysis, Callegari et al. showed that 25-30
distinctive descriptors are needed to cover the perceptual space for olfaction
alone. Therefore, in the cross-modal study of Lawless and Schlegel, a set of
two descriptors may have lacked the discriminative power needed to measure
interactions. Dravnieks and colleagues
(Dravnieks et al.,
1978) showed that over different panels, similarity ratings were
at least as consistent as measures derived from attribute ratings.
Summarizing, discrimination tasks and similarity ratings may be more sensitive
or reliable methods to detect mixture interactions than attribute ratings,
especially if selected descriptors are not distinctive. Nevertheless, the
latter are to be preferred if the qualitative nature of these interactions
should be assessed, provided that these are distinctive with respect to the
characteristics on which stimuli differ.
Where no perceptual blending occurs in a mixture of odorants, we expect
that the effect of changing a constituent's concentration in that mixture is
best reflected by ratings on the component's corresponding descriptor.
Panellists can generate these descriptors on presentation of the unmixed
odorants, in which case these odorants can be used as standards to train
panellists on the use of descriptors. This helps to align panellists'
conceptual representations of attributes and, hence, may improve consistency
of panel responses (O'Mahony,
1991; Lesschaeve and Sulmont,
1996). A descriptor set so designed may include as many
descriptors as there are components in the mixture, which will often be less
than the number of 25-30 recommended by Callegari et al.
(Callegari et al.,
1997). Although a small selection of attributes covers merely a
part of the olfactory universe, we argue that it is still a sensitive tool for
describing the interactions in the mixture under investigation if
component-derived descriptors are used.
Statistical interaction versus sensory interaction
When systematically manipulating the presence of a number of components in
a mixture according to a factorial mixing design, one faces the task of
deriving sensory mixture interactions from factorial plots of intensity
measures. Note that this is not identical to identifying statistical
interactions in the factorial plot. In general, statistical methods assume
linear models relating dependent variables to independent variables.
Psychophysical studies, however, have shown that the relationship between
stimulus concentration and its perceived intensity rarely approaches
linearity, but generally yields negatively accelerating curves fitting power
functions with exponents ranging from 0.1 to 1.0
(Stevens, 1961;
Cain, 1969;
Baird et al., 1996).
It can be shown that, due to the non-linearity of psychophysical functions,
factorial mixing plots of two-component mixture intensities will generally
show converging lines, even when the `constituents' are the same substance
(De Graaf and Frijters, 1988).
This may lead to a statistically significant interaction, while no sensory
interaction is present. Only in those rare cases where the factorial mixing
plot shows a set of diverging lines does a statistical interaction effect
support a sensory interaction: a case of extremely strong synergism
(De Graaf and Frijters, 1988;
Schifferstein and Frijters,
1990; Schifferstein,
1995). In addition, sensory interactions are evident when a
component suppresses an aspect of the mixture's aroma to which the component
does not itself contribute. This will appear as a significant, negative
statistical main effect and/or interaction effect of the suppressing component
on intensity ratings. In the latter case, the suppressive effect does not
necessarily coincide with a significant statistical interaction, although it
does concern a sensory interaction. In this paper we will first report the
outcomes of statistical tests and, subsequently, discuss these outcomes in
terms of their implications with respect to sensory interactions.
The present study
In the present study, we investigated whether and how sensory interactions affect the perception of CICs in a complex mixture of odorants that observers recognize as a natural food aroma. To study the contribution of different CICs, we omit CICs systematically from the mixture. If a component's odour does not blend into an aroma at all, its impact can be measured using intensity ratings on its respective quality descriptor. Suppression of this component's intensity by other components can be measured accordingly. If, at the other extreme, all components contribute to one aroma blend, the main character descriptor of the aroma can be used to measure the impact of constituent components. The omission of a CIC should then reduce intensity ratings on the main character descriptor. In order to be able to describe aromas with different degrees of blending, we used a detailed aroma-profiling task involving both single component descriptors and a main character descriptor. If components contribute to the main aroma character and also remain individually distinguishable, effects on both the main character descriptor and on specific quality descriptors will be observed.
Generally, food aromas are elicited by odorous components of varying odour intensities. Intensity is likely to be an important factor influencing the impact of an odorant on the mixture's aroma. In the present study, however, we wished to study processes involving odour quality only. To eliminate the effect of odour intensity, we matched the intensities of the three unmixed components under investigation before evaluating their effects in a multi-component model solution. These three components were selected according to their expected character impact: two components rated high on the target quality and one component rated low on this target quality.
|
Pilot study |
|---|
|
TopAbstractIntroductionPilot studyMain ExperimentResultsDiscussionReferences |
|---|
Materials and methods
Subjects
Eighteen paid volunteers, five men and 13 women, served as subjects. They
were recruited from the local Wageningen community and were selected on the
basis of their ability to generate and use refined odour attributes. In
addition, they showed high inter-subject coherency in the use of graphic
rating scales. This implied that subjects generated inter-subject-correlating
profiles when they rated various aroma intensities. All subjects were
experienced olfactory panellists (Bult
et al., 2001), but they were naïve with respect to
the objectives of the experiment. Their ages ranged from 19 to 51 (average 29
years). All were non-smokers and none had any history of olfactory
dysfunction. Subjects were in good health during the experiments. All gave
written informed consent.
Stimuli
The aroma model that is used in this study was derived from a headspace
sample of fresh apple juice earlier at this laboratory. Although the model
consists of a limited number of components, it was recognized and described as
apple by 13 out of 23 subjects upon presentation of the olfactory stimulus and
without any extra information being given
(Bult et al., 2001).
As identification performance for many common odours is 
50%
(Cain, 1979), this model was
deemed appropriate for the present study. Although we expect that more
authentic apple aromas can be made, the aroma model in the present study may
validly be assumed to represent a recognizable food aroma.
An apple reference stimulus was prepared from 10 components. Its composition was largely identical to that of the original apple mixture (Table 1). However, ethanol was excluded from the original mixture because it was of subthreshold concentration, even after substantially increasing its concentration. Pre-testing revealed that the ethanol component did not induce any consistent olfactory sensations. Furthermore, the propyl propanoate concentration in the mixture was raised by a factor of five to enable a more accurate stimulus preparation, thus improving stimulus reliability. Since this component still had a low intensity in the given concentration, we assumed that this alteration did not have a significant impact on the character of the mixture aroma.
|
In addition to the mixture, 10 one-component stimuli were prepared from each of the 10 components in the apple model mixture. To obtain equi-intense stimuli, the concentration of each singular solution as well as the concentrations in the complete apple mixture were raised, so that all intensities matched the sensory intensity of an 80 µl/l solution of trans-2-hexenal. This was done in a preliminary study employing four faculty members. The resulting composition of the apple reference stimulus is given in the sixth column of Table 1. The concentrations of the singular solutions are given in the fourth column of Table 1. Mixtures and single-component dilutions were prepared using distilled water. All stimuli were prepared at least 2 h and not earlier than 26 h before presentation. Stimulus solutions were stored in the dark at 4°C and were presented at ambient temperature (21 ± 1°C).
Procedure
Stimuli were presented in 200 ml glass jars, closed by low-odour plastic
screw caps, which could be opened by one simple twist. Each jar contained 10
ml of solution. To prevent volatile components from migrating from the screw
caps to the headspace, these two phases were separated with aluminium foil.
The subjects had to open each stimulus jar by unscrewing the cap, while
keeping the jar just underneath their noses. Responses were to be given after
taking a few short sniffs.
In the first session, the subjects generated odour attributes individually for all 10 unmixed stimuli. In the second part of this session, these attributes were discussed in a plenary meeting of all subjects. Consensus on the use of attributes was reached after plenary consultation in the second session.
At the start of the third session, the model mixture with the apple aroma was presented as the reference stimulus for `apple' aroma. Subsequently, the 10 singular component solutions were presented. Of each solution, subjects rated the intensity of its `apple' character on 150 mm scales printed on paper, labelled `no apple' at the left end and `very much apple' at the right end. The two components that scored highest and the one that scored lowest on `apple' were selected as, respectively, two CICs and one non-character impact component (non-CIC). Note that the definition of character impact used here is based on quality rather than perceived intensity, since all stimuli were approximately equally intense.
Statistical analysis
Ratings for all 10 stimuli were collected within subjects. To compensate
for idiosyncratic scale usage, attribute ratings were normalized to obtain
equal means and standard deviations for each subject. For convenience, the
complete data set for the group of subjects was transformed linearly in order
to obtain a group score range from 0 to 100. This resulted in an average
`apple' value of 39.1 (SD = 22.4) for every subject, over 10 evaluated
samples. Throughout the paper we used SPSS, version 7.5.2 (1997; SPSS,
Chicago, IL) for data analyses and 0.05 as the level of significance.
Results
Averaged normalized `apple' scores (±SE) are plotted for each component in Figure 1. Analysis of variance revealed a significant effect of `component' on `apple' responses [F(9,179) = 10.9, P < 0.01]. Subsequent post hoc testing, using Duncan's multiple range statistic, showed that the 10 `apple' scores could be grouped in four clusters of not significantly different means (Figure 1). From the cluster of components rated highest on `apple', hexyl acetate and trans-2-hexenal were selected as CICs, whereas 1-hexanol was selected from the lowest ranking cluster as a non-CIC.
|
|
Main Experiment |
|---|
|
TopAbstractIntroductionPilot studyMain ExperimentResultsDiscussionReferences |
|---|
Materials and methods
Subjects
Eighteen paid volunteers, five men and 13 women, served as subjects for the
main experiment. This group was identical to the group described in the pilot
study, except for one female subject who was replaced by another female
subject. The new subject met the criteria for admission to the panel as
described for the pilot study. Ages ranged from 19 to 51 years and the average
age was 30 years.
Stimuli
Ten different stimulus mixtures were prepared from the 10 selected
components by systematically adding combinations of the two CICs and the
non-CIC to a base solution of seven components in distilled water.
Concentrations of the base solution components and the additives are given in
the last column of Table 1. The
three additives—trans-2-hexenal, hexyl acetate and
1-hexanol—were added in singular, binary, or ternary combinations, thus
resulting in 2 (presence of hexyl acetate = HYL) x 2 (presence of
1-hexanol = HOL) x 2 (presence of trans-2-hexenal = HAL) = 8
mixtures (Table 2). To match
the unmixed intensities of hexyl acetate and trans-2-hexenal, the
1-hexanol component concentration had been increased to 300 ml/l relative to
its original 20 ml/l concentration in the apple mixture. In addition, two
mixtures containing the original—lower—concentrations of 1-hexanol
were included in the stimulus set. One mixture consisted of the base mixture
with only the low concentration 1-hexanol added (No. 9). The other also
contained the two CICs hexyl acetate and trans-2-hexenal (No. 10, the
original apple aroma).
|
Procedure
Subjects performed a descriptive analysis of the aromas of the 10 different
stimuli. The attributes that were generated for the 10 unmixed substances were
reduced to seven by letting each subject select three attributes that they
considered the least appropriate descriptors for the full apple mixture. The
attributes that were selected most frequently were discarded. In addition, an
`apple' attribute was included. Consequently, the eight attributes used in the
descriptive analysis were: `sour hard-boiled candy—glue';
`macaroon—hedge'; `sweet—lacquer'; `fruity—acetone';
`nuts—musty'; `pear—apple'; `bittersweet—rum' and `apple'.
The attribute names are translations of the Dutch terms used. Reference
stimuli for the eight attributes (the apple mixture plus the respective
components at concentrations identical to the pilot study) were presented
prior to every experimental session in order to refresh odour-attribute
associations. The use of the attributes `apple' and `pear—apple' may
seem confounding because of their similarity. Subjects, however, perceived the
respective qualities differently (see also
Figures 2 and 3) and they
considered these attributes the most appropriate for these aromas.
|
Stimulus preparation and presentation proceeded as described in the pilot
study. One session lasted 40-50 min. Laboratory conditions conformed to the
ISO 8589 standard (International
Organization for Standardization, 1987). During a session,
subjects were seated in separate booths. The uniform stimulus jars were coded
with randomly generated three-digit codes and they were presented in random
order, each individual receiving a separate order. Subjects were instructed to
rate attribute intensities on eight linear 150 mm graphic rating scales that
were presented on a laptop computer screen (Compaq Contura 80386 DX 25 MHz
with monochrome display), using the left button of a two-button computer
mouse. Between two stimuli, they waited for at least 1 min, which was
computer-paced. After completing two training series in the first session,
subjects completed nine experimental series of 10 stimuli each in three
separate sessions. Consequently, nine evaluations of every distinct stimulus
were recorded for every subject.
Statistical analysis
Data from the training sessions were discarded. Since no significant
systematic changes in responses over sessions were observed, ratings were
averaged over the nine repeated experimental sessions. Thus, data analyses
were performed on averaged intensity scores on eight attribute variables for
10 different stimuli per subject.
Perceived aroma quality is reflected in the aroma profile. This does not
imply that specific alterations of single component concentrations are
reflected exclusively in attribute ratings of their respective accompanying
attributes. Therefore, we initially tested for differences between complete
profiles due to stimulus composition by doubly multivariate repeated measures
analyses of variance—ANOVA (Stevens,
1996). These analyses permit simultaneous multivariate analyses of
results on a set of dependent variables according to a repeated-measures
design. The approach of initially performing a multivariate analysis also
guards against spurious effects due to the increased overall significance
level that results from performing successive univariate tests. Since the
experiment had a fractional factorial design (10 categories from a 12 category
full factorial design), the analysis was split into two consecutive
multivariate analyses. First, CIC and non-CIC effects were tested in a 2
x 2 x 2, HOL (not present versus high concentration) x HYL
x HAL design. Subsequently, the influence of all three 1-hexanol
concentrations (not present, low concentration, high concentration) and the
presence or the absence of both trans-2-hexenal and hexyl acetate was
tested in a 3 (HOL) x 2 (HYLHAL) design.
Any effects in the multivariate space indicate that aroma changes are perceived due to the addition or the omission of the CICs and the non-CIC. However, they do not give insight into the qualitative nature of the differences perceived. Because results on single attribute ratings may elucidate this in part, we proceeded with an analysis of single attribute data by univariate repeated measures ANOVA, as a post hoc test after significant multivariate effects are found. Multivariate F-values were calculated according to Pillai's trace criterion.
|
Results |
|---|
|
TopAbstractIntroductionPilot studyMain ExperimentResultsDiscussionReferences |
|---|
The multivariate effects of CICs and the non-CIC on the eight attribute ratings in the HOL x HYL x HAL analysis were found significant for HOL [F(8,10) = 4.80, P = 0.012] and HAL [F(8,10) = 13.12, P < 0.001], whereas HYL [F(8,10) = 2.48, P = 0.090] failed to reach significance. A significant multivariate HOL x HAL interaction was also observed [F(8,10) = 4.73, P = 0.013]. No three-way interaction was found. The HOL3 x HYLHAL analysis of all three 1-hexanol levels yielded significant multivariate effects for HOL3 [F(16,56) = 2.07, P = 0.023] and HYLHAL [F(8,10) = 3.97, P = 0.023]. The HOL3 x HYLHAL interaction was also significant [F(16,56) = 2.16, P = 0.017].
Univariate repeated measures ANOVAs were performed for the main and the two-way interaction effects that were significant in the multivariate analyses. Effects were found on four out of eight dependent variables. All effects of component presence on aroma that were found in the multivariate analysis had counterparts in one or several of these univariate effects. These univariate effects, therefore, appear to explain the multivariate effects. Hence, further discussion of results will be restricted to the four dependent variables that showed significant effects.
Table 3 shows the ANOVA results for the HOL x HYL x HAL design grouped for each separate dependent variable and specified for separate sources of variance. Apple ratings, that reflect character impact of the three components in the mixture, show significant main effects of HOL, HAL and HYL. No interactions were found with respect to `apple' ratings. In Figure 2A, the effects on aroma quality are illustrated. As may be expected on the basis of the nature of character impact components, addition of the CICs (HYL, HAL) to the base mixture increased `apple' ratings. The addition of the non-CIC (HOL) decreased `apple' ratings.
|
Most pronounced were the effects on the `nuts—musty' ratings, which were significantly affected by all three components. The `nuts—musty' attribute describes the character of the 1-hexanol component, which is reflected by significant higher `nuts—musty' ratings at high HOL levels (Table 3, Figure 2B). Furthermore, a significant HOL x HAL interaction is found for `nuts—musty' ratings. This interaction appears to be responsible for the multivariate HOL x HAL interaction found, since it is the only univariate interaction effect. It can be attributed to a masking influence of HAL on the `nuts—musty' character introduced by HOL. The presence of trans-2-hexenal does not affect `nuts—musty' ratings when 1-hexanol is not present. If 1-hexanol is present, however, the addition of trans-2-hexenal to the mixture suppresses the `nuts-musty' character drastically (Figure 2B). Likewise, HYL, the other CIC, appears to exhibit a masking effect on the `nuts—musty' character. Although the HOL x HYL interaction was not statistically significant, HYL had a significant main effect (Table 3) on `nuts—musty' ratings comprising a decrease in `nuts—musty' ratings due to HYL (Figure 2B).
HYL had a significant effect on its character descriptor `pear—apple'. As may be expected, this effect comprised an increase of `pear—apple' ratings after adding hexyl acetate to the mixture. Furthermore, `pear—apple' ratings decreased significantly when 1-hexanol was added to the mixture (Figure 2C).
Ratings on `bittersweet—rum' also increased significantly when its characteristic component, trans-2-hexenal, was added to the mixture (Table 3, Figure 2D). No other effects were found for this descriptor.
Table 4 shows the ANOVA results for the HOL3 x HYLHAL 3 x 2 design, grouped for each separate dependent variable and specified for separate sources of variance. The results are similar to those presented in Table 3. High HOL levels suppress `apple' and `pear—apple' ratings and increase `nuts—musty' ratings (Figure 2E,G,F, respectively). Ratings on `apple' and `pear—apple' increased when CICs (HYLHAL) were added (Figure 2E,G, respectively). The contribution of HOL3 to `nuts—musty' is suppressed by HYLHAL (Figure 2F).
|
Applying orthogonal simple contrasts on HOL3 levels (none, low, high), comparing the levels `none' to `low', respectively `none' to `high', revealed significant effects for `high' versus `none' on ratings for `apple' [F(1,17) = 14.48, P = 0.001], `pear—apple' [F(1,17) = 4.84, P = 0.042] and `nuts—musty' [F(1,17) = 36.16, P < 0.001]. No significant effects of `low' versus `none' were observed. Likewise, the HOL3 x HYLHAL interaction could be attributed to the interaction between `none' versus `high' (HOL) and HYLHAL. Therefore, all main and interaction effects of HOL3 were due to the influence of the highest 1-hexanol concentration level.
|
Discussion |
|---|
|
TopAbstractIntroductionPilot studyMain ExperimentResultsDiscussionReferences |
|---|
If only `apple' ratings are taken into account, the CICs and non-CIC investigated here show statistical additivity. Adding HYL or HAL to the apple base mixture increases apple ratings. Whether HYL and HAL produce sensory hypo- or hyper-additivity cannot be concluded on the basis of the present data, since this requires more information on the form of the psychophysical functions of these two substances (De Graaf and Frijters, 1988
;
Schifferstein, 1995). The
presence of 1-hexanol suppresses `apple' ratings for all mixtures, reflecting
a marked sensory interaction. Correspondingly, 1-hexanol suppresses mixture
ratings on the `pear—apple' attribute. When the multivariate HOL x HAL interaction was investigated further by studying single attribute effects, this interaction could be attributed to trans-2-hexenal suppressing the `nuts—musty' intensity at high 1-hexanol levels. In other words, the CIC trans-2-hexenal suppresses the contribution of 1-hexanol to the ratings on its corresponding attribute. This adds to the status of trans-2-hexenal as `character impact component' since it suppresses part of the effect of a non-character impact component on the mixture's aroma. Summarized, the observed interactions show that sensory interactions among mixture components are present and that these interactions pertain to ratings on a number of attributes.
In experimental investigations of mixture aroma quality, a single attribute
describing the main character of the aroma cannot sufficiently reflect
contributions of all components to the aroma. Had, for instance, only `apple'
ratings been used in this study, then the important trans-2-hexenal
x 1-hexanol interaction would have gone unnoticed. Therefore, we argue
that the use of multiple attribute ratings should be preferred to
one-dimensional measures in food aroma studies. However, a limited set of
eight attributes is rather small according to recommended 25-30
(Callegari et al.,
1997). Therefore, some concern is justified with respect to the
validity of the operationalization of aroma quality in the present study. When
using the common technique of odour profiling as an operationalization of
aroma quality, one assumes a linear additive model for contributions of each
attribute to the overall aroma. Therefore, an observed mutual suppression of
components can not result from the chosen operationalization technique. Hence,
we attribute the sensitivity for mixture interactions of our characterization
method to the use of descriptors generated on the basis of the constituent
odours. This enabled the training of attribute usage and allowed for direct
measurement of mutual suppression of odorants in the mixture. Although the
operationalization used here may not allow full representation of perceived
aromas, it proved sufficient for the assessment of mixture interactions.
To eliminate the effect of odour intensity, the intensities of the two CICs and the non-CIC were matched before their contribution to a multi-component model solution was investigated. This involved raising the concentration of the non-CIC component. Although this may have altered the quality of the aroma, it was necessary to do so in order to be able to attribute effects exclusively to the influence of odour quality. Interestingly, the highest rating on the apple attribute was given for the original apple model (Figure 2E), in which the non-CIC was present in low concentration.
The two CICs and the non-CIC showed main effects on their corresponding
odour attributes. From this, it can be concluded that these components did
indeed influence the perceived aromas. More specifically, when the three
studied components were added to the mixture, ratings for the three respective
character descriptors increased significantly. This suggests that panellists
were able to recognize the unique contribution of each of the three
manipulated components to the mixture's aroma. This is surprising given that
Laing and colleagues showed that humans have great difficulty recognizing as
few as three or four components in mixtures containing up to five or six
odorants (Laing and Francis,
1989; Laing and Glemarec,
1992; Livermore and Laing,
1996). This discrepancy in results is unlikely to be due to
differing similarities between the odorants used in each study. Laing et
al. employed dissimilar odorants, which should have maximized the number
of correctly identified components, whereas at least one odorant in the
present study was both structurally and perceptually similar to a CIC (i.e.
hexanal and trans-2-hexenal). Also, differing mixture complexities
cannot have caused this outcome since the mixture in the present study is more
complex, creating a more difficult task.
An explanation for the seemingly enhanced performance of subjects in the present experiment can be found in the methodology employed. In the present study, the subjects were aided by being provided with specific descriptors that directed them in rating specific feature intensities. Subjects were not requested to focus on physical components, as was the case in the study by Laing et al. Furthermore, Laing et al. gave their subjects dichotomous decision options: an odorant is present in the mixture or it is not. Under this regime a subject has a complex task: he or she has to assess the intensity of component-specific contributions to the mixture and has to decide on the relevance of the perceived intensities to the question of whether components are present or not. In contrast, the present study employed continuous attribute scales that enabled subjects to express the intensity of sensations. No absolute decisions on presence had to be made.
|
Acknowledgments |
|---|
This research was funded by the Netherlands Organisation for Scientific Research (NWO), by grant NWO-SGW-575-24.002 awarded to J.H.A.K.
|
References |
|---|
|
TopAbstractIntroductionPilot studyMain ExperimentResultsDiscussionReferences |
|---|
Baird, J.C., Berglund, B. and Olsson, M.J. (1996) Magnitude estimation of perceived odor intensity: empirical and theoretical properties. J. Exp. Psychol. Hum. Percept. Perform., 22,244 -255.[Web of Science][Medline]
Berglund, B. and Olsson, M.J. (1993a) Odor-intensity interaction in binary mixtures. J. Exp. Psychol. Human Percept. Perform., 19,302 -314.[Web of Science][Medline]
Berglund, B. and Olsson, M.J. (1993b) Odor-intensity interaction in binary and ternary mixtures.Percept. Psychophys. , 53,475 -482.[Web of Science][Medline]
Blank, I., Sen, A. and Grosch, W. (1992) Sensory study on the character-impact flavour compounds of dill herb (Anethum graveolens L.). Food Chem.,43 , 337-343.
Bult, J.H.F., Schifferstein, H.N.J., Roozen, J.P., Voragen,
A.G.J. and Kroeze, J.H.A. (2001) The influence of
olfactory concept on the probability of detecting sub- and peri-threshold
components in a mixture of odorants. Chem. Senses,26
, 459-469.
Buttery, R.G. and Ling, L.C. (1998) Additional studies on flavor components of tortilla chips. J. Agric. Food Chem., 46,2764 -2769.
Cain, W.S. (1969) Odor intensity: differences in the exponent of the psychophysical function.Percept. Psychophys. , 6,349 -354.
Cain, W.S. (1975) Odor intensity: mixtures
and masking. Chem. Senses Flav., 1,339
-352.
Cain, W.S. (1979) To know with the nose:
keys to odor identification. Science,203
, 467-470.
Callegari, P., Rouault, J. and Laffort, P.
(1997) Olfactory quality: from descriptor profiles to
similarities. Chem. Senses, 22,1
-8.
Czerny, M., Mayer, F. and Grosch, W. (1999) Sensory study on the character impact odorants of roasted arabica coffee. J. Agric. Food Chem.,47 , 695-699.[Web of Science][Medline]
De Graaf, C. and Frijters, J.E.R. (1988) Assessment of the taste interaction between qualitatively similar-tasting substances: a comparison between comparison rules. J. Exp. Psychol. Human Percept. Perform., 14,526 -538.[Web of Science][Medline]
Dravnieks, A. and O'Donnell, A. (1971) Principles and some techniques of high-resolution headspace analysis.J. Agric. Food Chem. , 19,1049 -1056.
Dravnieks, A., Bock, F.C., Powers, J.J., Tibbetts, M. and
Ford, M. (1978) Comparison of odors directly and
through profiling. Chem. Senses Flav.,3
, 191-225.
Frijters, J.E.R. (1980) Psychophysical and psychometrical models for the triangular method. Doctoral thesis, Utrecht University, Utrecht, The Netherlands.
Grosch, W. (2000) Specificity of the human nose in perceiving food odorants. In Schieberle, P. and Engel, K.H. (eds), The Proceedings of the 9th Weurman Flavour Research Symposium, Freising, Germany, 22-25 June 1999. Deutsche Forschungsanstalt für Lebensmittelchemie, Garching, pp.213 -219.
Guadagni, D.G., Buttery, R.G., Okano, S. and Burr, H.K. (1963) Additive effect of sub-threshold concentrations of some organic compounds associated with food aromas.Nature , 200,1288 -1289.[Medline]
Guth, H. and Grosch, W. (1994) Identification of the character impact odorants of stewed beef juice by instrumental analyses and sensory studies. J. Agric. Food Chem., 42,2862 -2866.
International Organization for Standardization (ISO) (1987) Sensory Analysis—General Guidance for the Design of Test Rooms, ISO 8589, 1-9. ISO, Geneva.
Jinks, A. and Laing, D.G. (1999) A limit in the processing of components in odour mixtures.Perception , 28,395 -404.[Web of Science][Medline]
Laing, D.G. and Francis, G.W. (1989) The capacity of humans to identify odors in mixtures. Physiol. Behav., 46,809 -814.[Medline]
Laing, D.G. and Glemarec, A. (1992) Selective attention and the perceptual analysis of odor mixtures.Physiol. Behav. , 52,1047 -1053.[Medline]
Laing, D.G. and Willcox, M.E. (1987) An investigation of the mechanisms of odor suppression using physical and dichorhinic mixtures. Behav. Brain Res.,26 , 79-87.[Web of Science][Medline]
Laing, D.G., Eddy, A., Francis, G.W. and Stephens, L. (1994) Evidence for the temporal processing of odor mixtures in humans. Brain Res., 651,317 -328.[Web of Science][Medline]
Laska, M. and Hudson, R. (1991) A
comparison of the detection thresholds of odour mixtures and their
components. Chem. Senses, 16,651
-662.
Laska, M. and Hudson, R. (1992)
Ability to discriminate between related odor mixtures. Chem.
Senses, 17,403
-415.
Lawless, H.T. (1987) An olfactory analogy to release from mixture suppression in taste. Bull. Psychonom. Soc., 25,266 -268.
Lawless, H.T. and Schlegel, M.P. (1984) Direct and indirect scaling of sensory differences in simple taste and odor mixtures. J. Food Sci., 49,44 -46, 51.
Lesschaeve, I. and Sulmont, C. (1996) Defining odour descriptors with one or three standards. 12th International Symposium on Olfaction and Taste, 25-31 August 1996, Zürich, p. 78.
Livermore, A. and Laing, D.G. (1996) Influence of training and experience on the perception of multicomponent odor mixtures. J. Exp. Psychol. Human Percept. Perform.,22 , 267-277.[Web of Science][Medline]
Livermore, A. and Laing, D.G. (1998) The influence of chemical complexity on the perception of multicomponent odor mixtures. Percept. Psychophys.,60 , 650-661.[Web of Science][Medline]
Moskowitz, H.R. (1979) Utility of the vector model for higher-order mixtures: a correction. Sensory Processes, 3,366 -369.[Medline]
Moskowitz, H.R. and Barbe, C.D. (1977) Profiling of odor components and their mixtures. Sensory Processes, 1,212 -226.[Web of Science][Medline]
O'Mahony, M. (1991) Descriptive analysis and concept alignment. In Lawless, H.T. and Klein, B.P. (eds),Sensory Science Theory and Applications in Foods . Marcel Dekker, New York, pp. 223-268.
Olsson, M.J. (1994) An interaction model for odor quality and intensity. Percept. Psychophys.,55 , 363-372.[Web of Science][Medline]
Reiners, J. and Grosch, W. (1998) Odorants of virgin olive oils with different flavor profiles.J. Agric. Food Chem. , 46,2754 -2763.
Schieberle, P. and Hofmann, T. (1997) Evaluation of the character impact odorants in fresh strawberry juice by quantitative measurements and sensory studies on model mixtures.J. Agric. Food Chem. , 45,227 -232.
Schifferstein, H.N.J. (1995) Perception of taste mixtures. In Doty, R.L. (ed.), Handbook of Olfaction and Gustation. Marcel Dekker, New York, pp.689 -713.
Schifferstein, H.N.J. and Frijters, J.E.R.
(1990) Sensory integration in citric acid/sucrose
mixtures. Chem. Senses, 15,87
-109.
Stevens, J. (1996) Applied Multivariate Statistics for the Social Sciences. Lawrence Erlbaum, Mahwah, NJ.
Stevens, J.C. and Traverzo, A. (1997)
Detection of a target taste in a complex masker. Chem.
Senses, 22,529
-534.
Stevens, S.S. (1961) Toward a resolution of the Fechner—Thurstone legacy. Psychometrika,26 , 35-47.
Thurstone, L.L. (1927) A law of comparative judgement. Psychol. Rev., 34,273 -286.
Wagner, R.K. and Grosch, W. (1998) Key odorants of french fries. J. Am. Oil Chem. Soc.,75 , 1385-1392.
Accepted February 26, 2002
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  E. Le Berre, N. Beno, A. Ishii, C. Chabanet, P. Etievant, and T. Thomas-Danguin Just Noticeable Differences in Component Concentrations Modify the Odor Quality of a Blending Mixture Chem Senses, April 1, 2008; 33(4): 389 - 395. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||