Food Quality and Preference4 (1993) 133-l 39
INFLUENCEOFSUGARANDFATCONTENTS ONPREFERENCEFORCREAMCHEESE: APRELlMlNARYSTUDY* B. Daillant
& S. lssanchou
Laboratoire de Recherches sur les ArBmes, Institut National de la Recherche Agronomique, 17, rue Sully B.V 21034 Dijon Ckdex, France (Received November 1992; acqbted 3 June 1993)
is required for comprehensive testing of the sensory properties of these foods. In addition Kiister (1981) underlined the need for validation of results obtained in an artificial context with supplementary data obtained under more normal conditions. Some previous studies (Bellisle & Lucas, 198’7; Bellisle et al., 1988) have shown a discrepancy between brief sensory tests and consumption tests. However, other authors have found accordance between these tests (Shepherd & Farleigh, 1986; Stone SCPangborn, 1990). The main aim of this experiment was to study some methodological aspects of acceptability measurements. To study the effect of fat and sugar on acceptability, semi-solid French cream cheese was chosen as a food product model. Ratings after brief sensory stimulation were compared on the one hand with ratings after consumption of a normal portion and on the other hand with added quantity of sugar in ad libitum mixing. In addition, measures of atti-
ABSTRACT In order to assess the importance of fat and sugar in acceptability of semi-solid French cream cheese, 30 subjects gave a hedonic judgement, after consumption, fm three formulations (150 S;, varying in fat content which they were required to sweeten and fm nine small preoiously sweetened samples (30 g). Subjects participated in jive laboratory meals, cream cheese being the third part of a meal. After this period, the suljects$lkd out a questionnaire measuring attitudes, beliefs and intention of behaviour with regard to consumption of cream cheese. Sugar level had no signijicant effect on the hedonic judgement. However; the subjects added less sugar in the high fat cream cheese than in the low fat one. Concerning the influence offat level, measures on small samples and after consumption
were in accordance:
tudes and beliefs towards fat content
the low
were included
product was the farthest from ideal. Howe-vq users of high fat cream cheese differed in preference and attitudes from users of low fat product.
in the present
MATERIALS
KLywords: Preference; acceptability; attitudes; ideal; relative
AND
in cream
cheese
study.
METHODS
Subjects
rating; ad libitum mixing; hedonic judgement; fat; sugar:
Twelve French men and 18 French women, 18 to 48 years old, were recruited through an announcement in a local newspaper. None had previous experience with sensory testing and all of them usually ate their cream cheese with sugar. Subjects were classified as either low, medium or high fat cream cheese users.
INTRODUCTION Foods containing fat and sugar have high hedonic value and it appears that fat and sugar contribute sensory properties which seem to be affectively and cognitively loaded. Mela (1990) pointed out that ‘preferences for fats are almost certainly predominated by learned associations’. Thus, a psychological component
Samples French cream cheese with three different levels of fat-low, medium and high-were studied. French cream cheese resembles unsalted cottage cheese but has a smooth texture; the most common way to consume cream cheese in France is to add sugar. The percentages of fat for each level expressed as dry weight and total weight are given in Table 1.
*Presented at R.M. Pangborn Memorial Symposium: Advances in Sensory Science, JPrvenpG, Finland, August 4, 1992. 0 1994 Elsevier Science Publishers Ltd 0950-3293/94/$06.00 133
134
B. Da&ant, S. Issanchm
TABLE 1. Composition of the nine Cream Cheeses Used in the Sensory and Hedonic Testing a
Level of
Dry
fat
weight
Low Medium High
Concentration of sugar (g/100 g of cream cheese)
Total weight
10% 20% 40%
1.5% 3% 8%
6
8
10
A A B
B AB A
A B B
“For 15 subjects, samples marked A were tested during the fourth meal and samples marked B were tested during the fifth meal. The other 15 subjects tested samples marked B during the fourth meal and samples marked A during the fifth meal.
Sessions Subjects
participated
vidual booths. tween 12.00 starter
in five laboratory
meals in indi-
They were invited at lunch time, i.e. beand 13.00
(salad),
main
h. For each meal, subjects had course,
cream
cheese
and com-
pote. Cream cheese was thus the third part of the meal, which is common starter,
in France.
main course
Different
and compote
versions
of the
were served during
the study. For each meal, the starter, the main course and compote corresponded to the same caloric value (550-600 kcal). Subjects had to rate each component of the meal on a global ideal relative scale. This scale is an unstructured linear scale 10 cm long which is labelled at the left anchor ‘ideal for me’ and at the right anchor ‘very far from my ideal of cream cheese’ A zero score would thus indicate an ideal product. Usually the scale ‘relative to ideal’ is used for assessing the most preferred level of a factor in a food product. Booth et al. (1983) and Shepherd et al. (1984) used it for salt and Conner et al. (1986) for sugar. But in this study, it was used for a global assessment of a food product. Using this scale, the subject must refer to his own ideal of cream cheese (internal standard) and might be more involved with his rating than with usual hedonic scales. During the first three meals, a bowl containing 150 g of cream cheese (either low fat, medium fat or high fat) was presented unlabelled, and 150 g of sugar was available. The subjects were instructed to sweeten the cream cheese themselves (ad libitum), then to eat the quantity they required and give their hedonic judgement. During the next two meals, five sweetened unlabelled samples of 30 g were successively served. Subjects had to rate each sample on a global relativeto-ideal scale. Subjects swallowed the samples. Nine sweetened samples were prepared by adding sugar to the three types of cream cheese. Three levels of sugar were used: 6,8 and 10 g per 100 g: The levels used were
lower than the commercial level (between 9 and 10.5 g per 100 g) because a previous study (Daillant & Issanchou, 1991) reported that, in yoghurt, this commercial level seemed greater than the ideal level of sugar. The experimental design is presented in Table 1.
Measures of attitudes and beliefs A questionnaire based on the Fishbein and Ajzen model (Ajzen 8c Fishbein, 1980) has been elaborated in collaboration with the Institute of Food Research in Reading. This model was previously used in food research by R. Shepherd, (e.g. Shepherd, 1988) and H. Tuorila (e.g. Tuorila, 1987a). In this model, behaviour is predicted by behavioural intention, which is predicted by the attitude of the subject towards the behaviour and social pressure. The attitude is predicted by beliefs of the subject about the outcome of the behaviour modified by his or her evaluation of the outcome. Our questionnaire was assumed to measure beliefs, outcome evaluation, attitude and intention of behaviour. The measure of social pressure was not included in the questionnaire because many authors have reported that the attitude component was dominant over the norm component (Shepherd & Stockley, 1987; Tuorila, 1987b Shepherd, 1990) for meat and dairy products. The questionnaire thus consisted of the following measures:
(1) Two ‘belief’ items (e.g. ‘Eating high fat cream
cheese gives me a cream cheese which tastes good/helps me control my weight’) which were accompanied by nine-point rating scales leftanchored by ‘extremely unlikely’ and rightanchored by ‘extremely likely’. Two corresponding outcome evaluation items (2) (e.g. ‘Having food which tastes good/helps me control my weight’) which were accompanied by nine-point rating scales left-anchored by ‘extremely bad’ and right-anchored by ‘extremely good’). (3) Two nine-point scales designed to assess attitude via the use of the semantic differential (‘unpleasant/pleasant’; ‘harmful/beneficial’). One intention measure (e.g. ‘I intend to eat low (4) fat cream cheese the next time I eat cream cheese’) accompanied by nine-point rating scales left-anchored by ‘extremely unlikely’ and right-anchored by ‘extremely likely’. Attitudes and beliefs were measured concerning low, medium and high fat cream cheese. Items were presented in random order in the final questionnaire. Five different orders of the items were used. Two months after the study, the subjects were invited by letter to come into the laboratory to fill out this questionnaire. Only 20 subjects did so.
Preferace fm Sugar and Fat As beliefs and outcome evaluation were measured on a nine-point scale (-4 to +4), each product (belief multiplied by outcome evaluation) took values from - 16 (negative attitudes) to + 16 (positive attitudes).
RESULTS
AND
DISCUSSION
Quantity of sugar added Two subjects
did not sweeten their cream cheese so they were removed from this analysis, leaving 28 subjects. The mean value of added quantity of sugar for these subjects was 7.9 g per 100 g (standard error of the mean = O-43). An analysis of variance on normalized added quantity of sugar was performed with the 28 subjects. Centring yields a similar level of consumption for each subject. Necessary attenuation of differences between the variances of consumption justified the reduction. A significant effect of fat level (p < O-03) was observed, and a multiple comparison of means (Duncan test) indicated that subjects added more sugar in low fat cream cheese than in high fat cream cheese. When the subjects were split into three groups according to their habits, the following trends were observed. For users of low fat cream cheese (seven sub jects), there was no significant effect of fat, denoting that they did not compensate for lack of fat with sugar. For users of medium fat cream cheese (eight subjects), a significant effect of fat at the 7% level was observed and a comparison of means showed that they added less sugar in high fat cream cheese than in medium and low fat cream cheese. Users of high fat cream cheese (10 subjects), tended to add more sugar in the low fat cream cheese, which would reveal a small compensatory effect of sugar ( p < O-25). Three subjects reported that they ate all three types of cream cheese so they were not included in any of the groups.
(low) and at the same time they also like the intrinsic properties of high fat cream cheese. For the users of medium fat cream cheese (nine subjects), the medium fat cream cheese was preferred, as expected, and sub jects agreed concerning their response to the medium fat product, as shown by the size of the box where 50% of the ratings are represented. Fat effect was significant at the 6% level. High fat cream cheese was also close to ideal. For the users of high fat cream cheese (10 sub jects), the high fat cream cheese seemed to be preferred and subjects agreed concerning their response to this product. However, fat effect was not significant, probably because subjects disagreed on their judgment of low fat cream cheese. In short, for the whole group, low fat cream cheese is the farthest from ideal. However, partitioning the consumers into groups shows some tendency to differ according to their habits of consumption. Subjects generally prefer high fat cream cheese and the cream cheese they are used to consuming. Tuorila (198’7a) and Tuorila et al. (1990) reported that subjects preferred the type of milk they normally used. Hellemann et al. (1990) reported similar trends for spreads. In the present study strong preference for high fat cream cheese seemed to compete with preference for the cream cheese that subjects are used to.
Hedonic
response after consumption
Results are presented in Fig. 1 in a box-plot format to show the dispersion. An analysis of variance and a multiple comparison of means (Duncan test) were performed. In each case the effect of subject was significant. For the whole group (30 subjects), a significant effect of fat ( p < 0.05) was observed: low fat cream cheese was the farthest from ideal. When the subjects were split into three groups according to their habits, preference for the three levels of fat were found to exhibit three different patterns. Users of low fat cream (eight subjects) tended to prefer high fat cream cheese and low fat cream cheese ( p < 0.09 for fat effect). One explanation could be that these subjects prefer the level of fat they normally ate
response to nine sweetened samples
Results are presented in Table 2. A three-way (subjects, fat, sugar) analysis of variance was performed. The effect of sugar was not significant overall, however, 8 g per 100 g (medium level) seemed to be the closest to ideal, and a separate analysis by the different user groups reveals that this trend was mainly due to the high fat users ( p < 0.05). The effect of varying fat level was significant ( p < O.Ol), with the low fat cream cheese scoring farthest from ideal. A second sidering
Hedonic
135
analysis of variance
the nine samples
was performed
as a factor.
con-
This factor
was
TABLE 2 Mean Distances from Ideal Scored by 30 Consumers for three Levels of Sugar Content by three Fat Levels in Cream Cheese Level of fat
Concentration of sugar
Means
(g/ 100 $9 10
by fat level
6
8
Low Medium High
4+76”**
4.44w*
3.85u**
4.3.Fia*
3.66” 3.40”
2.95” 2.97”
3.31” 3.77”
3.3Ib 3.3Bb
Means by sugar level (NS)
3.94
3.45
3.65
* Means with different superscript letters are significantly different at the 5% level. ** Means by sugar and fat level. Means wit.h different superscript letters are significantly different at the 5% level.
136 B. Daillunt, S. Zssanchou
I
9
1
WHOLE GROUP, N&IO
IO
lSERS OF MEDIUM FAT CREAM CHEESE, N=
9
.9-
8
7-
7
6-
6
5-
5
4
4
3-
3
2-
2
l-
1
OJ
0
FAT CONTENT
FAT CONTENT
-
10
107
1 USERS OF LOW FAT CREAM CHEESE, N=8 94
S-
8
e-
7-
7-
6-
6-
USERS OF HIGH FAT CREAM CHEESE. N=l(
i
*1 *
s-
5r
4-
4-
3-
3-
2-
zz-
1-
1 . 1
1
1
01
medium
high
medium
IOW
high
FAT CONTENT
FAT CONTENT
FIG. 1. Hedonic response after consumption. Box-plot representations of distances from ideal. The asterisk (*) represents the mean, the horizontal bar is the median; 50% of the ratings are found within the box. The values represented by the ‘whiskers’ (vertical bars) are mean plus the differences between upper and lower quartile multiplied by plus or minus 1.5. Outliers are represented by points. Means with different letters are significantly different at the 5% level.
significant
( p < O-01) and the multiple
means
(Duncan
test)
groups,
as indicated
Sample farthest
split
the
by the
superscripts
‘low fat/low
sugar’
(superscript
from
Samples
ideal.
comparison
samples
‘medium
into
of
three
in Table
2.
u) was the fat/medium
sugar’ (superscript w) sugar’ and ‘high fat/medium were the closest to ideal. The six other samples (superscript v) were intermediate. A compensatory be observed
effect
of sugar
in low fat cream
cheese
on
ratings
could
in the sense that
distance from ideal decreased with increasing sugar level. A similar compensatory effect was found with salt in bread and butter by Hellemann
et al. (1990).
Ratings after consumption agreed with the sensory testing in finding the low fat cream cheese to be the
farthest
from
ideal. The
scale may account to refer
to their
have decreased testing
own internal Another
standards
reason
subjects
behaviour
than
had
and this may unnatural
may be that the sen-
were told to swallow the samples.
born and Giovanni (1984) in a swallowing condition sumer
ideal relative
Indeed,
the bias due to the rather
situation.
sory assessors
use of a global
for this result.
Pang-
found that scores obtained were more typical of con-
the
no-swallowing
generally used for brief sensory testing. Concerning the degree of agreement
condition between
be-
haviour and rating, the mean value for added sugar was 7.9 g per 100 g, and the closest level to ideal seemed to be 8 g per 100 g. Similar hedonic
rating
and
good agreement
ad &turn
mixing
between
were found
by
Preference for Sugar and Fat
Stone and Pangborn (1990) with NaCl in broth and sugar in lemonade. Likewise, Shepherd and Farleigh (1986) measured total intake and table salt use and ob served that subjects with low intake showed preference (on an ideal relative scale) for lower concentrations of salt in the particular food tested. In contrast, Bellisle and Lucas (1987) and Bellisle et al. (1988) reported the sugar or salt level to be higher in the preferred sample than in the most consumed sample. Factors accounting for the differences reported by these authors could be as follows:
(1) The use of a nine-point pleasantness scale for
measuring preference. With this scale, the sub jects may have made assessments without reference to their own internal standard. (2) The large range of concentrations, 0 g to 40 g of sucrose per 100 g of yoghurt. Range with high concentrations may have led subjects to prefer high concentrations (McBride, 1985) (3) Expectorated sample ratings may lead to higher preferred sucrose concentrations than the swallowed one (Mattes & Mela, 1986). These authors compared expectorated versus swallowed sample ratings and reported the latter to be more reflective of dietary intake behaviour.
137
CONTROL MY WEIGHT
16
-8 1
-16
L”
L
M
H
Low fat usage (n = 6)
L
M
H
High fat usage (n = 8) TASTE GOOD
16 12
a
Measures of attitudes and beliefs
4
The stated intentions of behaviour were in agreement with habits declared two months before. That is to say, subjects intended to consume the type of cream cheese that they had previously declared they use. Analysis of variance was performed on the data from the measured beliefs multiplied by the outcome evaluation for ‘help me control my weight’ and ‘taste good’, for low fat users and high fat users. Means and confidence intervals are presented in Fig. 2. The effect of subject was not significant. The results for the variable ‘Control my weight’ showed that fat was a significant source of variation for low fat users but not for high fat users. Users of low fat cream cheese have very negative attitudes towards high fat cream cheese and very positive attitudes towards low fat cream cheese. But high fat users have slightly positive attitudes towards low fat cream cheese and slightly negative attitudes towards high fat cream cheese. The results for the variable ‘Taste good’ showed that the fat was a significant source of variation for high fat users but this effect was less significant for low fat users. Users of high fat cream cheese have very negative attitudes towards low fat cream cheese and very positive attitudes towards high fat cream cheese. Low fat users have very positive attitudes towards high fat cream cheese but do not have negative attitudes towards low fat cream cheese.
0 -4 -8
-201 L
M
Low fat usage (n = 6)
H
L
M
H
High fat usage (n = 8)
FIG. 2. Beliefs X Outcome evaluation scores, means and 95% confidence interval. Fat level: L = low fat, M = medium fat, H = high fat. Means with different letters are significantly different at the 5% level.
Therefore, subjects tended to underestimate the possible negative properties of their own type of cream cheese and overestimate the possible positive properties, as previously found by Tuorila (1987a). For each fat usage group the correlations between scores of each item and scores for ‘beneficial’ were calculated in order to understand the meaning of beneficial/harmful. In spite of the small number of observations (three levels of fat times the number of sub jects in each group), there were interesting differences among the user groups presented in Table 3. For low
138 B. Dailht,
S. Issanchou
3. Correlations of Attitudes and Beliefs Scores for 30 Subjects Grouped by their Low, Medium and High Usage of Cream Cheese with a ‘Harmful-Beneficial’ Scale
TABLE
Attitude, belief item
Cream cheese usage LOW
Control my weight” Taste good” Pleasantb
0.68++ 0.07 0.36
Medium
0.13 o-53+ 0.51 t
High
0.15 0.49+ 0.62t+
‘Belief X Evaluation score. ‘Attitude measure. t p< 0.05. tt p< 0.01.
fat users, ‘beneficial’ was correlated with ‘control weight’. On the contrary, for medium and high fat users, ‘beneficial’ was correlated with ‘taste good’ and ‘pleasant’. This means that beneficial has a health meaning for low fat users and a sensory meaning for medium and high fat users. It is important to note that medium fat users respond like high fat users. The correlations between scores of belief X evaluation, attitudes, intention and hedonic rating for each user group are presented in Table 4. Scores for intention to consume were positively correlated with ‘control my weight’ and ‘beneficial’ scores for the users of low fat cream cheese, and with ‘beneficial’, ‘pleasant’ and ‘taste good’ scores for the users of medium and high fat cream cheese. These results help to clarify the reasons why a particular sub ject intended to consume a particular type of cream cheese. This use of the model is in agreement with Shepherd (1990): ‘The attitudes model developed by Fishbein and Ajzen can be used to determine the relative importance of different factors in influencing food choice.’ Distance from ideal was not correlated with any item in the model for users of low fat cream cheese, and we
conclude that there are discrepancies between attitudes and hedonic judgements for this group. For users of medium fat cream cheese, distance was negatively correlated with ‘taste good’, ‘pleasant’ and ‘intention to consume’. Thus the smaller the distance from ideal, the more they intended to consume the product. For users of high fat cream cheese, less significant correlations were observed with ‘taste good’ and ‘intention to consume’. Correlations between hedonic judgement and attitudes observed with medium and high fat users were reported by Tuorila (1987a,@. But such correlations may not lead to conclusions about the relationships between attitudes and hedonic judgement. Indeed, attitudes could influence hedonic judgement and/or hedonic judgement could give rise to appropriate attitudes. Even though the samples were presented unlabelled, presumably subjects could still identify them as low, medium or high fat to some extent, particularly since all subjects were regular consumers of one type of cream cheese and knew their usual product very well.
CONCLUSION Fat had a significant effect on hedonic judgment, with sensory measures and scores after consumption being in accordance: low fat cream cheese was the farthest from ideal. Sugar had no significant effect on the distance from ideal. However, low fat cream cheese becomes closer to ideal as sugar content increases. Moreover, this partial compensatory effect was observed on behavioural measures as well as sensory. Hedonic responses are not only affectively loaded, and measures of attitudes help to understand some interactions between these affects and cognitions. Our results provide evidence to support the behaviour described by Drewnowski (1987): ‘Negative attitudes to
4. Correlations between Score of Belief X Evaluation, Attitudes, Intention and Hedonic Bating for 30 Subjects Grouped by their Low, Medium and High Usage of Cream Cheese
TABLE
Intention to consume: Cream cheese usage
Attitude, belief item
LOW
Control weight” Beneficialb Taste good” Pleasantb Intention to consume “Belief X Evaluation score. bAttitude score.
tp
Medium
0.59+++
0.13
0.78+++
0.59++
-0.05 0.22
Distance after consumption: Cream cheese usage High
-0.13
0.62+++
0.83tt+
0.76+++
0.84+++
0.75+++
LOW
MediWl
Hi@
0.16 0.15 0.11 -0.30 0.04
-0.38 -0.35 -0.57++
0.12 -0.04 -0.38+ -0.17
-o-54++ -0.57tt
-0.37+
Preference for Sugar and Fat
dietary fats may outweigh the positive sensory response, leading consumers to reject good tasting but “unhealthy” foods.’ However, experiments using actual measures of consumption need to be conducted to confirm
this phenomenon.
It appeared ferent jects
that partitioning
usage groups are
needed
the population
gave useful insights, if different
into dif-
but more sub-
populations
are
to be
observed. This work was a preliminary study, first in order to compare hedonic rating after tasting small samples and after consumption of a normal portion, then to underline the importance of attitudes towards a product and ‘the usefulness of using the model proposed by Fishbein and Ajzen’ (Shepherd & Farleigh, 1986). Subsequently, an experiment with more subjects will be necessary with the same types of measurements, i.e., laboratory tests and attitudes and beliefs questionnaire. Moreover, measure of consumption at home will be useful in observing what people do in a natural context.
ACKNOWLEDGEMENTS The authors would like to thank P. Schlich for the help in planning the experiment and analysing the data. The authors are grateful to Dr R. Shepherd and Dr P. Sparks from the Institute of Food Research for help in elaborating the attitude questionnaire and the European Community for providing travel funds through a FLAIR exchange.
REFERENCES Ajzen, I. & Fishbein, M. (1980). Utitunding Attitudes and Predicting Social Behavim Prentice-Hall, Englewood Cliffs,
NJ.
Bellisle, F. & Lucas, F. (1987). La measure des preferences alimentaires: Les tests brefs d’evaluation sensorielle permettent-iis de pridire la consommation? Sci. Aliments, 7 (hors irie) (8)) 33-7. Bellisle, F., Giachetti, I. & Toumier, A. (1988). Comment determiner la salinite priferee d’un aliment: Comparison de l’evaluation sensorielle et des tests de consommation. Sti. Alim.ent.r,8,.X57-64. Booth, D. A., Thompson, A. & Shahedian, B. (1983). A robust brief measure of individuals most preferred level of salt in an ordinary foodstuff. Appetiti, 4, 301-12. Conner, M. T., Haddon, V. & Booth, D. A. (1986). Very rapid, precise assessment of effects of constituent variation on
139
product acceptability: consumer sweetness preferences in a lime drink. L_.ebensm. Wiss. u. Technol., 19, 486-90. Daillant, B. & Issanchou, S. (1991). Most preferred level of sugar: Rapid measure and consumption test. J Sensory Studies, 6, 131-44. Drewnowski, A. (1987). Fats and food acceptance: Sensory, hedonic and attitudinal aspects. In Food Acceptance and Nutrition, ed. Solms, J., Booth, D. A., Pangborn, R. M. & Raunhardt, 0. Academic Press, London, pp. 189-204. Hellemann, U., Barylko-Pikielna, N., Matuzeuska, I. (1990). Interaction between bread and butter with varying NaCl contents: hedonic responses and sensory characteristics. Food Quality and A-eferace, 2, 167-76. Koster, E. P. (1981) Sensory evaluation in a ‘natural environment’. In Fluvour 81, ed. Schreier, P. Walter de Gruyter, Berlin, New York, pp. 93-100. MacBride, R. L. (1985). Stimulus range influences intensity and hedonic ratings of flavour. Appetite, 6, 125-31. Mattes, R. D. & Mela, D. J. (1986). Relationships between and among selected measures of sweet-taste preference and dietary intake. Chemical Senses, 11(4), 523-39. Mela, D. J. (1990). Sensory preferences for fats: what, who, why? Food Quality and Preferace, 2,95-101. Pangborn, R. M. & Giovanni, M. E. (1984). Dietary intake of sweet foods and of dairy fats and resultant gustatory responses to sugar in lemonade and to fat in milk. Appetite, 5, 317-27. Shepherd, R. (1988). Belief structure in relation to low-fat milk consumpti0n.J. Human Nutrition and Dietetics, 1,421-8. Shepherd, R. (1990). Behavioural modelling of fat consump tion. Food Quality and Preference, 2,89-94. Shepherd, R. & Farleigh, C. A. (1986). Preferences, attitudes and personality as determinants of salt intake. Human Nutrition: Applied Nutrition, 4OA, 195-208. Shepherd, R. & Stockley, L. (1987). Nutrition knowledge, attitudes, and fat consumption. J American Dietetic Association, 5(87), 615-19. Shepherd, R., & Farleigh, C. A. & Land, D. G. (1984). The relationship between salt intake and preferences for different salt levels in soup. Appetite, 5, 281-90. Stone, L. J. & Pangborn, R. M. (1990). Preferences and intake measures of salt and sugar, and their relation to personality traits. Appetite, 15,63-79. Tuorila, H. (1987a). Hedonic response and attitudes in the acceptance of sweetness, saltiness and fattiness of foods. In Food Acceptance and Nutrition, ed. Solms, J., Booth, D. A., Pangbom, R. M. & Raunhardt, 0. Academic Press, London, pp. 337-51. Tuorila, H. (19876), Selection of milks with varying fat contents and related overall liking, attitudes, norms and intentions. ApPetite, 8, 1-14. Tuorila, H., Lehtovaara, A. & Matuszewska, I. (1990). Sandwiches and milk with varying fat and sodium contents: What is the best combination? Food Quality and. preference, 2, 223-31.