Sunteți pe pagina 1din 10

Food Quality and Preference 12 (2001) 327336

www.elsevier.com/locate/foodqual

Sensory perception of fat in milk


Michael Bom Frst *, Garmt Dijksterhuis, Magni Martens
Sensory Science Group, Department of Dairy and Food Science, Royal Veterinary and Agricultural University, Rolighedsvej 305,
DK-1958 Frederiksberg C, Denmark
Received 15 September 2000; received in revised form 4 December 2000; accepted 22 December 2000

Abstract
The sensory properties of fat in milk were examined by sensory descriptive analysis. To date, no single food additive has been
completely successful in mimicking the sensory properties of fat in milk. This experiment investigated the eects of various factors
and combinations thereof on sensory properties and perceived fattiness of milk, and compared them to the actual fat content (0.1;
1.3 and 3.5% fat milk was used). The other factors studied were the addition of thickener, whitener, cream aroma and homogenisation. Multivariate data analytical methods (Partial Least Squares Regression) were applied for analysis of the data. The three
former additional factors contributed signicantly to perceived fattiness of the milk, and homogenisation had a small but not signicant eect. It was shown that a combination of thickener, whitener and cream aroma in 0.1% fat milk was approximately successful in mimicking sensory properties of 1.3% fat milk. # 2001 Elsevier Science Ltd. All rights reserved.
Keywords: Milk; Fats; Lipids; Sensory analysis; Multivariate data analysis; Partial least square regression (PLSR); Fattiness perception; Fat substitutes; Thickening agent; Whitening agent; Cream aroma; Homogenisation

1. Introduction
The sensory properties of milk are inuenced by the
fat content in milk (Phillips, McGi, Barbano, & Lawless, 1995a; Tuorila, 1986). In order to produce milk
and milk-based drinks with salient sensory properties, it
is necessary to understand how the dierent sensory
modalities are aected by the fat content. This also will
lead to a better understanding of what the perceived
fattiness in milk is comprised of. Previous research has
shown that much of the sensory dierences between
non-fat and other types of milk are mainly in found
appearance, texture and mouthfeel (Phillips et al.,
1995a; Tuorila, 1986). Various food additives have been
examined as a substitute of fat in milk. Phillips, McGi,
Barbano, and Lawless (1995b) observed some eect of
protein content on the sensory properties. The addition
of 2% non-fat dry milk to skimmed milk gave a similar
eect on the relative physical viscosity (measured by
capillary viscometer) as addition of 2% fat. Still, the
perceived texture (Thickness, Mouthcoating and Residual

* Corresponding author. Fax: +45-35-28-31-90.


E-mail address: mbf@kvl.dk (M.B. Frst).

mouthfeel) did not change as much as with an increase


in the actual fat content. Also, there was a signicant
increase in cooked avour and both the perceived and
physically measured (by MacBeth Colour-Eye spectrophotometer) colour of the milk with added non-fat dry
milk were not the same as that of milk with 2% fat. This
result suggests that a fat substitute in milk, to be successful, should change the appearance characteristics of
the milk more than the texture properties.
Other food additives have been examined as well.
Phillips and Barbano (1997) tested a series of food
additives by sensory descriptive analysis with trained
panellists. They found that a combination of sodium
caseinate and titanium dioxide in skimmed milk was
best at mimicking the colour of 2% fat milk. However,
the combination did not fully succeed in improving the
perceived texture. Some of the other visual attributes,
like the one descriptor visual hang up, a measure of
how much milk is clinging to the inner surface after
swirling of the glass, was still much lower than 2% fat
milk.
Other experiments substituted the addition of non-fat
dry milk with a protein standardisation by ultraltration (Quinones, Barbano, & Phillips, 1997, 1998). By
mixing retentate and permeate from ultraltration they

0950-3293/01/$ - see front matter # 2001 Elsevier Science Ltd. All rights reserved.
PII: S0950-3293(01)00018-0

328

M.B. Frst et al. / Food Quality and Preference 12 (2001) 327336

produced milks with a true protein content in the range


of 0.94.8%. The normal range for American milk is
3.133.38% (Quinones et al., 1998). In the experiments,
the milks varied in fat content from 0.13 to 3.3%. Their
results showed that the sensory texture and appearance
descriptors were aected by the protein standardisation
(as well as by the fat content), where increase in protein
content gave a whiter appearance and texture properties
like Thickness, Mouthcoating and Residual mouthcoating increased with a higher protein content. Larger
dierences in sensory properties between low and high
protein content were observed in samples with low fat
content. The main objective of those two studies was to
see the eect of a decrease in protein content. The economical value of milk protein as a separate food additive has increased substantially over recent decades
(Quinones et al., 1998), thus, there would be an increase
in outcome if part of the protein in milk could be
separated and sold for other purposes.
Tepper and Kuang (1996) examined the eect of avour addition in milk model systems by multidimensional scaling. They showed that products with
high levels of added aroma compounds were perceived
similar to products with a higher fat content, which then
indicates an eect of avour on perception of fattiness.
However, in their experiment they operated with a
rather large concentration range of aroma compounds
(0, 0.5 and 1% w/v powdered cream avour) and a large
span of fat content (0, 5 and 10% w/v added bland
vegetable oil). Still, the authors claim that their pilot
studies conrm that the chosen concentrations provided
small but distinguishable dierences between samples.
Richardson, Booth, and Stanley (1993) have theorised
that the fat particle size distribution makes an essential
contribution to perceived creaminess in milk, but only
in the presence of adequate viscosity. They suggested
that in order for a uid dairy product to be perceived
like dairy cream, it should provide a smooth but viscous
uid layer between the tongue and the palate. Homogenisation of the milk only had an eect on perceived
creaminess when the milk was also thickened to the
viscosity of double cream (47.5% fat). They used a nonfat thickener (1 part 3% w/v of sodium carboxymethylcellulose solution to nine parts of milk).
All of the mentioned experiments show an eect of
dierent factors on the sensory properties of milk.
However, none of these experiments examined more
than two factors simultaneously in relation to perceived
fattiness of milk. Also, some of the experiments used
stimuli that are extreme compared to the range of milk
normally consumed in western countries. Thus, the
objectives of the present experiment was to examine the
eect of numerous factors related to perception of fattiness in milk within a realistic product range. In this
study a meta-descriptor Total fattiness was introduced
to the sensory panel, to evaluate the perceived fattiness

in milk. By doing so it was possible to see how the other


sensory descriptors correlated with perceived fattiness.

2. Materials and methods


2.1. Milks
Standard Danish milk with three dierent levels of fat
were used (0.1, 1.3 and 3.5%). Organically produced
milk was chosen, since this is commercially available as
non-homogenised milk in Denmark, and it thus allowed
for testing of homogenisation without interference with
the production method (conventionally produced milk
is only sold homogenised in Denmark). On the basis of
previous literature in the eld, it was decided to test four
factors in a reduced design. The factors varied in the
experiment are listed in Table 1.
In pilot studies, a series of four thickeners were tested
in selected concentrations within the range recommended by the producer. Some other thickeners were
considered for the experiment, but for these, pasteurisation was necessary before it was fully dissolved.
Those were avoided since the additional heat treatment
might have an unwanted eect on the sensory properties
(Nursten, 1997). Three dierent cream avours were
tested in several concentrations, and the one agreed
upon by the experimenters as most similar to real cream
aroma (13% fat cream) was selected. Finally, titanium
dioxide was tested in a range of concentrations and the
concentration in Table 1 was selected. The authors
philosophy of the addition of food additives was that it
should have a perceivable eect, but not too much of an
eect i.e. keeping it within a realistic product range. In
order to suspend the whitener in the milk, it was necessary to homogenise the samples. Likewise it was necessary to homogenise samples with thickener in order to
avoid lumps of thickener in the milk. This resulted in
some restrictions on the design, so that whitener and
thickener could not be tested independently of homogenisation. So, the full design of 48 samples was reduced
to 30 possible combinations. From these, 16 were selected for descriptive analysis (listed in Table 3, together
with results from the descriptive analysis). The selection
criteria for the samples for sensory evaluation were
based on three basic considerations: (1) achieving a
large sensory space (from 0.1% fat without any additives to 3.5% fat with all additives); (2) obtaining a
representation of all combinations of additions; (3)
combinations of additions that were expected to have
large eects on sensory properties.
2.2. Descriptive analysis
Sensory descriptive analysis was performed under
normal light with milk, (approximaley 100 ml) in clear

329

M.B. Frst et al. / Food Quality and Preference 12 (2001) 327336

glasses, in the sensory laboratory at the University. A


panel consisting of seven external paid panellists was
used for the evaluation. All panellists had much experience with sensory evaluation. In ve training sessions
panellists were trained on the products, and descriptors
were chosen after suggestions from the panel leader on
the basis of consensus among the panellists. Each training session had a duration of approximately 1 1/2 h. (In
the fth training session panellists evaluated a subset of
the samples for sensory evaluation in the sensory evaluation booths, this session took only approximately 50
min). A total of 15 descriptors were used for the
descriptive analysis. Those are listed in Table 2, together
with their denitions and original Danish words. A

reference sample (1.3% fat, +thickener, +whitener,


+aroma, +homogenisation), was chosen and the
intensity of this reference sample was scored by the
panellists in the training sessions. The reference sample
was presented to the panellists and tasted prior to each
evaluation, alongside with a score card marked with the
average scores for the panel for this sample. References
for descriptors creamy smell and boiled milk smell
were also presented and smelled by the panellists prior
to evaluation. For reference materials were used 13%
fat cream and boiled skimmed milk respectively. For the
latter reference material, skimmed milk was brought to
boiling point and cooled down again prior to sessions.
All samples and references were kept at 12 C for 1 h

Table 1
Factors varied in the experiment; levels and origin of material used
Experimental factor

Value of levels

Origin of material or processequipment

Fat content

0.1% (w/w)
1.3% (w/w)
3.5% (w/w)
0
1 g/l
0
1 g/l
0
0.75 g/l
0
150 bar

Organically produced non-homogenised milk (MD-Foods, Slagelse, Denmark)

Thickener
Whitener
Aroma
Homogenisation

Alginate FD 155 (Danisco Cultor, Arhus, Denmark)


Titanium(IV)dioxide reagent grade (Lancaster, Eastgate, UK)
Cream Flavouring U33162 (Danisco Cultor, Arhus, Denmark)
Pilot Scale homogeniser (Rannie, Denmark)

Table 2
Sensory descriptors, their denitions and original words in Danish
Descriptors

Denition (reference material)

Original words in Danish

Aroma
Creamy aroma
Boiled milk aroma

Intensity of raw cream aroma (13% fat cream)


Intensity of boiled milk aroma (boiled skim milk)

Lugt
Fldeagtig lugt
Kogt mlk lugt

Appearance/colour
Whiteness
Yellowness
Blueness
Transparency
Glass coating
Thickness visual

Degree of intensity of the colour white in the centre of the glass


Degree of intensity of the colour yellow in the centre of the glass
Degree of intensity of the colour blue in the centre of the glass
Degree of transparency of the sample at the edge of the glass tilted approximately 30
Amount of milk clinging to the inner surface of the serving glass after swirling the sample
Degree of thickness measured during swirling of glass

Udseende
Hvidhed
Gullighed
Blalighed
Gennemsigtighed
Glasvedhftning
Tykhed/viskositet

Flavour/Tastea
Creamy avour
Boiled milk avour
Sweet taste

Intensity of cream avour


Intensity of boiled milk avour
Intensity of sweet taste

Smag
Fldeagtig smag
Kogt mlk smag
Sd smag

Texture/Mouthfeel
Thickness oral
Creaminess oral
Residual mouth ll

Perceived thickness of the sample evaluated in the mouth


Perceived creaminess of the sample evaluated in the mouth
Degree of residual mouth coating after expectoration of the sample

Konsistens
Tykhed/viskositet
Cremethed
Eftermundfylde

Meta descriptor
Total fattiness

Overall perception of fat content in the sample

Samlet fedhed

In Danish, no word for avour exists. Flavour is expressed through the Danish term for taste, smag.

330

M.B. Frst et al. / Food Quality and Preference 12 (2001) 327336

before sessions. Only one sample at a time was served to


panellists and were taken out 12 min before serving.
For all evaluation sessions a computerised score collection software (FIZZ, Biosystemes, France) was used. A
horizontal 15-cm unstructured line scale anchored at the
left end with a little or none (in Danish: lidt or
ingen) and at the right end with a lot or very
intense (Danish: meget or meget intens) was used.
Sensory analysis of the 16 products was carried out in
triplicate, and in randomised order within each replicate. In each session, only eight products were evaluated, so a total of six sessions were necessary to
complete the experiment.
2.3. Data analysis
Data analysis was performed using descriptive univariate analyses (mean and ANOVA for each descriptor). ANOVAs for the individual descriptors were performed using panellists as a random factor.
Multivariate data analysis (Partial Least Squares
Regression [PLSR]) was applied to investigate relationships between sensory data and the experimental design.
Initially, data were analysed to correct for irrelevant
dierences between panellists by PLSR, and level corrected data were used for analysis of eects of dierent
treatments (cf. Martens, Wede, Bredie, & Martens,
1999). After level correction, data was averaged over
panellists, and those data were used for analysis. For all
the multivariate analyses, cross validation was performed, leaving each replicate out at a time (Martens &
Naes, 1989). The analyses were performed in standard
statistical software packages (SPSS 9.0.0, SPSS Inc.
Chicago, IL, USA, for univariate statistics and
Unscrambler 7.51a, Camo ASA, Trondheim, Norway,
for multivariate data analysis).

3. Results and discussion


The results from ANOVA showed signicant dierences among the samples with regard to all sensory
descriptors. Mean values and least signicant dierences at 5% level for all samples over all panellists and
replicates are presented in Table 3. It shows that the
experimental design used produced dierences in all
sensory modalities (aroma, appearance, avour/taste
and texture/mouthfeel).
3.1. Sensory descriptors and eect of experimental
factors
After pre-processing of the data (level correction to
account for panellists dierent use of scale), ANOVAPLSR (APLSR) was performed (design as X-variables
and sensory data as Y-variable). Figs. 1a and 2a show

correlation loading plots from four signicant dimensions (explaining in average 60, 15, 4.2 and 4.1% of the
variation in sensory data, respectively). The observed
eects become much more apparent when looking at
score plots for the four dimensions, when the dierent
factors are labelled separately and products are grouped
by factors or combinations thereof. These are shown in
Figs. 1b,c and 2b,c, and are referred to along with the
explanations in the text. All other combinations of
dimensions were explored during data analysis, but the
combinations shown, 12 and 34, proved sucient for
interpretation of the data.
In dimension 1, two major clusters of descriptors are
seen. One group consisting of Boiled milk smell and
avour, Transparency and Blueness relates to 0.1% fat
milk. At the opposite end of the rst dimension is a
group of descriptors all relating to 3.5% fat milk, consisting of Sweet taste, Creamy smell and avour,
Thickness visual and oral, Glass coating, Creaminess,
Residual mouth ll, Total fattiness and Yellowness.
This indicates a clear fat-level-direction in the results
(Fig. 1b). The distances between groups with dierent
fat levels show that there is a much larger dierence in
fattiness between 0.1 fat and 1.3% fat than there is
between 1.3 and 3.5%. This indicates that perceived
fattiness is not a linear function of actual fat content in
the range spanned by the milk in our experiment,
which is the range of milk fat content normally sold in
Denmark.
The combined eects of thickener, whitener and
homogenisation result in a direction orthogonal to that
of fat level (Fig. 1c). This direction is mainly a dierence
in perceived whiteness. Samples with the addition of
those three factors mainly score higher in whiteness,
whereas those without, score lower. In Fig. 1c the factors are grouped, so the eect of the individual factors
can be seen. From this, it is evident that whitener and
thickener contribute to fattiness. There seems to be a
small eect of homogenisation as well. Signicant differences in a few sensory descriptors were observed.
Homogenisation gives an increase in Creamy avour
and a decrease in Blueness, both of these dierences
were only observed in 1.3% fat milk (Table 3). The
eect of whitener and thickener is an increased perceived whiteness of the milk, as well as some increase in
the group of high fat related descriptors. In this direction, no separation of thickener and whitener is seen,
this subject will be returned to in dimension 4 below.
The third dimension spans the variation between
samples with and without added aroma compounds, so
this is a avour dimension (Fig. 2b). Samples with
added aroma compounds have a higher intensity of
Creamy smell and avour. Notice also that sweet taste
correlates with the creamy descriptors (Fig. 2a). However, the absolute dierences in sweetness are smaller
than the dierences in creamy smell and avour (cf.

Table 3
Mean values (over panellists and replicates), Least Signicant Dierences (LSD) values (P<0.05) for all products and descriptors, product specications and product codesa
Product names

Product description

Sensory descriptors
Aroma

Fat level (%)

Thickener

Whitener

Aroma

Creamy smell

0.1
1.3
3.5

0twAh
0tWaH
0TwaH
0tWAH
0TWAH
1twaH
1twAh
1tWAH
1TwAH
1TWaH
3twaH
3TwaH
3TWAH

0.1
0.1
0.1
0.1
0.1
1.3
1.3
1.3
1.3
1.3
3.5
3.5
3.5

+
+
+
+

+
+

+
+

+
+
+

+
+

+
+

+
+
+
+
+
+
+
+
+
+
+

Product description

Boiled milk smell

Whiteness

Yellowness

1.12
3.56
5.24
6.11

0.97
6.20
5.79
4.43

0.78
6.02
6.39
6.27

0.85
4.27
5.84
8.05

0.76
6.45
4.18
2.69

0.74
10.40
7.99
5.24

0.95
2.53
6.55
9.65

0.90
3.31
6.56
8.79

5.58
4.19
3.76
6.53
6.42
6.16
7.89
8.44
8.36
6.36
7.21
7.04
9.07

5.88
6.05
6.38
5.34
5.50
4.86
4.42
3.86
4.35
3.91
4.14
4.47
3.59

6.02
8.26
5.75
7.64
7.73
6.36
6.34
7.43
6.71
8.13
6.38
6.89
7.63

4.79
4.16
4.18
5.12
4.89
6.17
6.56
6.41
6.68
5.72
7.67
7.47
6.64

6.04
4.09
6.96
3.86
4.14
3.37
3.65
2.96
3.09
2.68
3.04
2.54
2.47

10.08
7.93
10.48
8.14
7.95
7.36
7.91
6.44
7.21
5.81
5.85
5.07
3.99

3.06
4.08
2.41
4.76
5.04
6.77
7.10
8.41
7.83
8.62
9.46
10.46
10.41

3.90
5.47
3.65
5.41
6.49
7.14
6.67
8.19
7.45
8.31
8.45
9.51
9.84

Thickener

Whitener

Aroma

Homogenisation

LSD (5%)
0twah
1twah
3twah

0.1
1.3
3.5

0twAh
0tWaH
0TwaH
0tWAH
0TWAH
1twaH
1twAh
1tWAH
1TwAH
1TwaH
3twaH
3TwaH
3TWAH

0.1
0.1
0.1
0.1
0.1
1.3
1.3
1.3
1.3
1.3
3.5
3.5
3.5

+
+
+
+

Transparency

Glass
coating

Thickness
visual

Residual mouth ll

Total fattiness

Sensory descriptors
Flavour/Taste

Fat level (%)

Blueness

+
+

+
+

+
+
+

+
+

+
+

+
+
+
+
+
+
+
+
+
+
+

Creamy avour

Texture/Mouthfeel
Boiled milk avour

Sweet taste

Thickness oral

Creaminess oral

1.08

1.01

0.72

0.88

0.99

1.00

1.03

3.46
5.21
8.05

7.06
6.08
5.07

5.04
5.91
6.47

3.23
6.43
8.81

3.14
6.14
8.85

3.34
5.99
8.24

3.26
5.94
8.97

5.73
4.58
3.36
6.12
6.61
6.86
8.14
8.60
8.86
7.56
8.20
8.93
10.05

6.58
6.49
7.05
6.09
6.25
5.34
4.67
4.80
4.58
4.74
4.53
4.19
3.57

6.85
5.53
4.97
6.38
6.31
6.41
7.39
7.14
7.46
6.39
6.89
7.22
8.07

4.09
5.04
3.41
5.04
5.90
6.57
6.56
8.08
7.87
8.27
8.59
9.62
9.56

4.89
4.50
3.18
4.90
6.51
6.77
7.30
8.47
8.81
8.11
8.66
9.76
10.39

4.88
4.13
3.05
5.00
5.97
6.09
6.95
8.26
8.01
7.67
8.29
9.11
9.69

4.73
4.44
2.89
5.24
6.76
6.87
7.49
8.59
8.96
7.94
8.76
9.64
10.69

331

a
Product abbreviations refer to the levels of the factors (0.1 and 3, 0.1, 1.3 and 3.5% fat, respectively; t and T, none and addition of thickener; w and W, none and addition of whitener; a and A, none and addition of aroma
compounds; h and H, no homogenisation and homogenisation).

M.B. Frst et al. / Food Quality and Preference 12 (2001) 327336

LSD (5%)
0twah
1twah
3twah

Appearance

Homogenisation

332

M.B. Frst et al. / Food Quality and Preference 12 (2001) 327336

Table 3). Still, there is some transference of scoring the


attribute creamy to the scoring of sweetness. This can be
concluded since the latter did not change with the different factors in the experiment. Samples with added
aroma score higher in Total Fattiness, indicating that an
aroma component contributes to perceived fattiness in
the milks.
Finally the fourth dimension shows the sensory differences between samples with added whitener versus
those with added thickener (Fig. 2c). At the 0.1% fat
level, the sample with only added whitener scores higher
in Whiteness, Glass coating, Thickness visual and oral,

Creaminess, Residual mouth ll and Total fattiness


(Table 3). At the same fat level the sample with only
added thickener has a higher blueness and transparency.
At the 1.3% fat level, the sample with only added
thickener still scores signicantly higher in transparency, compared to the sample with only added whitener
(Table 3). The signicant dierences in sensory descriptors indicate that the addition of whitener has a higher
impact on sensory properties than addition of thickener
has. However, the chosen levels of additions for the two
factors inuence this conclusion. The grouping of the
descriptors in Fig. 2a indicates that Glass coating and

Fig. 1. (a) APLSR correlation loadings for the two rst dimensions showing dierences among the 16 products. ~ Sensory Descriptors, & Factors
and * Products. For clarity of this gure product names are not shown. The inner and outer circles represent 50 and 100% explained variance,
respectively. (b) Score plot from APLSR. Dimension I and 2. Indicating dierences in fat levels (0=0.1%, 1=1.3% and 3=3.5%), and showing fat
level direction based on the loadings. (c) Score plot from APLSR. Dimension I and 2. Indicating dierences in addition of thickener (t/T), whitener
(w/W) and homogenisation (h/H). Capital letters indicate addition of the factor. Showing whitener and thickener level direction and homogenisation
level direction based on the loadings.

M.B. Frst et al. / Food Quality and Preference 12 (2001) 327336

Thickness visual/oral may relate more to the colour of


the milk, than to the actual thickness (Fig. 2a). Samples
with only added whitener have higher intensities in these
descriptors, compared to samples with only added
thickener.
3.2. Total fattiness
To evaluate the perception of fat in the products, the
panel used the meta-descriptor Total fattiness. It can
be seen conceptually as a projection of all sensory
descriptors onto a fattiness percept. However, to rate

333

this meta-descriptor may be a task of high cognitive


character, and can be inuenced by the panellists
interpretations of how the individual descriptors contribute to perceived fattiness. So, what is then really
investigated is how the dierent sensory attributes contribute to believed perceived fattiness as a result from
the previous scoring of all other attributes. From the
correlation loadings plot for rst and second dimension
(Fig. 1a) it can be seen that there is a whole group of
descriptors that are highly positively correlated with
Total fattiness, (Creaminess oral, Creamy smell and
avour, Sweet taste, Thickness visual and oral, Glass

Fig. 2. (a) APLSR correlation loadings for components 3 and 4 showing dierences among the 16 products. ~ Sensory Descriptors, & Factors and
* Products. For clarity of this gure product names are not shown, and only relevant descriptor names are shown. The inner and outer circles
represent 50% and 100% explained variance, respectively. (b) Score plot from APLSR. Dimension 3 and 4. Indicating addition of aroma compounds (a/A). Capital letters indicate addition of the factor, and showing aroma level direction based on the loadings. (c) Score plot from APLSR.
Dimension 3 and 4. Indicating dierences in addition of thickener (t/T) and whitener (w/W). Capital letters indicate addition of the factor.

334

M.B. Frst et al. / Food Quality and Preference 12 (2001) 327336

coating and Residual mouth ll). Yellowness is also


correlated to Total fattiness, but not as highly as the rest
of this group. Further Fig. 1a shows that Whiteness is
not correlated to Total fattiness, and that the group of
descriptors in the opposite end of the rst dimension is
negatively correlated to Total fattiness. Closer examination of the grouping of the fattiness related descriptors in the subsequent dimensions show that in
dimension 3 and 4 (Fig. 2a) Total fattiness is still closely
grouped with Residual mouth ll and Creaminess, indicating that these two descriptors are most fully reecting perceived fattiness in milk. However, there is not
much Total fattiness left to explain in these two dimensions, since it lies relatively close to the origin in Fig. 2a.
The conclusion that Creaminess and Residual mouth ll
most fully reect perceived fattiness in milk was conrmed by a PLSR-analysis with Total fattiness as the Yvariable and all other sensory descriptors as the X-variables. Variation in Creaminess and Residual mouth ll
contributed the most to explained variance in Total
fattiness (both of them higher than 96%).
3.3. Dierences between products
Cobweb-plots can be used to focus on the dierences
between a few products. Fig. 3 shows the sensory properties of the three dierent fat levels. The gure shows

the same dierences that were apparent from the correlation loading plots (Fig. 1a) and the score plot with
labels for fat content (Fig. 1b). A rather large group of
descriptors is positively correlated with fattiness (Residual mouth ll, Creaminess, Thickness visual/oral,
Creamy smell and avour, Glass coating and Yellowness). Another group of descriptors is inversely correlated with a high fat content (Boiled milk smell and
avour, Blueness and Transparency). All of these
descriptors vary signicantly (P<0.05) across the three
products. Whiteness is not aected signicantly by the
changes in fat content. It is also, as shown in APLSR,
evident that the sensory dierences between 0.1 and
1.3% fat are larger than between 1.3 and 3.5% fat. This
is conrmed by the signicant dierences between products observed by ANOVA. Fig. 4 shows the sensory
prole of 0.1% fat with all added additives and 1.3%
fat, homogenised. There are only signicant dierences
in two descriptors. The 0.1% fat milk with additions
have a higher Whiteness, but a lower Glass coating. The
Total fattiness is the same, so the sensory properties of
1.3% fat have been mimicked quite successfully by the
addition of all the food additives used in the experiment.
This was not the case with only one food additive added
(Fig. 5). Addition of whitener and aroma were not sufcient either to mimic the increase in true fat content
from 0.1 to 1.3%, since there is a signicant dierence

Fig. 3. Cobweb plot of sensory proles from the three dierent fat levels.

M.B. Frst et al. / Food Quality and Preference 12 (2001) 327336

Fig. 4. Cobweb plot of sensory proles from 1.3% fat, homogenised and 0.1% fat added all additives.

Fig. 5. Cobweb plot of sensory proles from 1.3% fat, homogenised and 0.1% fat added only one of the food ingredients.

335

336

M.B. Frst et al. / Food Quality and Preference 12 (2001) 327336

in Total fattiness. Nor were combinations of two food


additives in 1.3% fat milk completely capable of
mimicking the increase from 1.3% fat to 3.5% fat. But
there were no signicant dierences in Total fattiness
(Table 3). The conclusions from this experiment might
not have a high relevance with regard to milk for consumption, since there (in the European Union) is a very
restrictive legislation about food additives. However,
for other milk-based beverages, where there is more
room for improvements by adding food additives, the
conclusions about how to imitate fattiness are valuable.

Acknowledgements
This work is part of the FTEK programme supported by the Danish Dairy Research Foundation
(Danish Dairy Board) and the Danish Government.
Nina Ahn, Judith Henning and Anne Marie Laustsen
are thanked for technical assistance. Ditte Marie Folkenberg is acknowledged for inspiring the rst author to
using the meta-descriptor Total fattiness. The donation of food additives from Danisco Cultor is highly
appreciated.

References
4. Conclusions
The results from the experiment show that the sensory
properties of fat in milk are comprised of texture/
mouthfeel, appearance and avour attributes. Analysis
of the contribution of the individual sensory descriptors
to perceived fattiness in milk shows that Creaminess
and Residual mouth ll most accurately reect fattiness
of milk in this fat range. It is unknown to what extent
the results about perceived fattiness from a sensory
panel can be extended to the general population. However, the assessment of fattiness obtained from a trained
sensory panel gives a very detailed picture of human
sensory perception of fattiness. The sensory dierences
between 0.1% fat milk with added thickener, whitener
and aroma are very small as compared to those in
homogenised 1.3% fat milk. They only dier signicantly in two descriptors: Whiteness and Glass coating. It was not the case when only using one food
additive, or combinations of two. This clearly shows
that the sensory properties of fat in milk are not easily
substituted with food additives seeking to imitate the
sensory properties of fat. The experiment showed that
there are larger sensory dierences between milk with
0.1% and 1.3% fat, than there are between 1.3% and
3.5% fat. This shows that the fat does not aect the
sensory properties of milk in a linear fashion. It is very
likely that sensory properties of 0.1% fat milk can be
changed drastically by adding only a little extra fat to
the milk.

Martens, H., & Naes, T. (1989). Multivariate calibration. Chichester:


John Wiley.
Martens, H., Wede, S., Bredie, W. L. P., & Martens, M. (1999).
Manual for sensory data analysis. Copenhagen: Department of
Dairy and Food Science, KVL.
Nursten, H. E. (1997). The avour of milk and dairy products. I. Milk
of dierent kinds, milk powder, butter and cream. International
Journal of Dairy Technology, 50(2), 4856.
Phillips, L. G., & Barbano, D. M. (1997). The inuence of fat substitutes based on protein and titanium dioxide on the sensory properties of lowfat milks. Journal of Dairy Science, 80(11), 27262731.
Phillips, L. G., McGi, M. L., Barbano, D. M., & Lawless, H. T.
(1995a). The inuence of fat on the sensory properties, viscosity,
and color of lowfat milk. Journal of Dairy Science, 78(6), 1258
1266.
Phillips, L. G., McGi, M. L., Barbano, D. M., & Lawless, H. T.
(1995b). The inuence of nonfat dry milk on the sensory properties,
viscosity, and color of Iowfat milks. Journal of Dairy Science,
78(10), 21132118.
Quinones, H. J., Barbano, D. M., & Phillips, L. G. (1997). Inuence of
protein standardization by ultraltration on the viscosity, color, and
sensory properties of skim and 1% milk. Journal of Dairy Science,
80(12), 31423151.
Quinones, H. J., Barbano, D. M., & Phillips, L. G. (1998). Inuence of
protein standardization by ultraltration on the viscosity, color, and
sensory properties of 2 and 3.3% milks. Journal of Dairy Science,
81(4), 884894.
Richardson, N. J., Booth, D. A., & Stanley, N. L. (1993). Eect of
homogenization and fat content on oral perception of low and high
viscosity model creams. Journal of Sensory Studies, 8(2), 133143.
Tepper, B. J., & Kuang, T. (1996). Perception of fat in a milk model
system using multidimensional scaling. Journal of Sensory Studies,
11(3), 175190.
Tuorila, H. (1986). Sensory proles of milks with varying fat contents.
Lebensmittel Wissenschaft und Technologie, 19(4), 344345.

S-ar putea să vă placă și