Food Quality and Preference 19 (2008) 565–573

Contents lists available at ScienceDirect

Food Quality and Preference

journal homepage: www.elsevier.com/locate/foodqual

The role of expectancy in sensory and hedonic evaluation: The case of

smoked salmon ice-cream
Martin R. Yeomans a,*, Lucy Chambers a, Heston Blumenthal b, Anthony Blake c
a
Department of Psychology, School of Life Sciences, University of Sussex, Falmer, Brighton BN1 9QH, UK
b
The Fat Duck Restaurant, Bray Oxford, UK
c
Firmenich Research, Geneva, Switzerland

a r t i c l e i n f o a b s t r a c t

Article history: Our experience of flavour involves integration of multiple sensory inputs, and the hedonic evaluation of
Received 15 August 2007 this complex flavour experience is important in determination of food choice. The appearance of food also
Received in revised form 7 February 2008 generates expectations about food flavour, and past work suggests that these expectations if confirmed
Accepted 24 February 2008
enhance the flavour experience. What is less clear is what happens when cues prior to ingestion predict
Available online 4 March 2008
a flavour which is in marked contrast to the actual flavour characteristics. To test this, we conducted three
experiments where expectations about food flavour were generated by plausible but inaccurate food
labels for a highly novel food, smoked-salmon ice-cream. In Experiment 1, the experience of the food
Keywords:
Flavour
in the mouth generated strong dislike when labelled as ice-cream, but acceptance when labelled as frozen
Hedonics savoury mousse. Labelling the food as ice-cream also resulted in stronger ratings of how salty and
Expectancy savoury the food was than when labelled as a savoury food. Experiment 2 confirmed these findings,
and also found that an uninformative label also resulted in acceptable liking ratings. Experiment 3 explic-
itly tested the effect of labels on flavour expectation, and confirmed that the ice-cream label generated
strong expectations of a sweet, fruity flavour, consistent with the visual appearance of the ice-cream,
but in marked contrast to the flavour of salty fish. As in Experiments 1 and 2, liking was minimal when
the food was tasted after the ice-cream label condition, but liking was acceptable in the other label con-
ditions. These data show that the contrast between expected and actual sensory qualities can result in a
strong negative affective response and enhancement of the unexpected sensory qualities.
Ó 2008 Elsevier Ltd. All rights reserved.

1. Introduction pounds released from food, mechanoreception contributing to

Although the senses are anatomically separate, they rarely
operate independently since the majority of stimuli in the environ-
ment stimulate multiple senses. The increased recognition that our
sensory experience reflects integration of these multiple sensory
inputs has been applied to many experiences, most notably in
the current context to our experience of food flavour (e.g., Delwi-
che, 2004; Keast, Dalton, & Breslin, 2004; Small, Jones-Gotman, Za-
torre, Petrides, & Evans, 1997). Although the experience of the
sensory qualities of a food are often described in terms of how it
‘‘tastes”, in practice this experience of flavour is a complex interac-
tion between multiple sensory experiences. Arguably, multi-sen-
sory integration may be at its most extreme in the case of
flavour perception since few other experiences offer the opportu-
nity for concomitant stimulation of all the major senses: gustation
through the five primary tastes, olfaction through both ortho- and
retronasal stimulation of olfactory receptors by volatile com-
* Corresponding author. Tel.: +44 1273 678617; fax: +44 1273 678058.
E-mail address: martin@sussex.ac.uk (M.R. Yeomans).
our perception of texture and providing information on tempera-
ture, pain arising from oral irritants and hearing that results from
sounds and vibrations coming from the mouth contributing to
our perception of aspects of texture. The focus of the present paper
is on a further aspect of the multi-sensory experience of flavour,
how expectations about food flavour arising from visual and cogni-
tive cues prior to ingestion modify our hedonic and sensory expe-
rience of the flavour of food in the mouth.
The visual appearance of a food is well known to influence fla-
vour recognition. Thus many studies have shown that the presence
of a congruent colour enhances the ability to identify food and
drink stimuli, relative to presentation of the same stimuli without
a colour cue or with an incongruent colour (Dubose, Cardello, &
Maller, 1980; Stillman, 1993; Teerling, 1992). Further evidence of
cross-modal associations within food-related stimuli involving vi-
sual cues comes from studies of interactions between visual and
olfactory stimuli. For example, when a white wine was coloured
red, the sensory descriptors applied to the odour of the wine were
consistently terms used normally for red rather than white wine
(Morrot, Brochet, & Dubourdieu, 2001). Thus, in the absence of

0950-3293/$ - see front matter Ó 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.foodqual.2008.02.009
566 M.R. Yeomans et al. / Food Quality and Preference 19 (2008) 565–573
appropriate visual cues, the actual olfactory quality of the wine had
little impact on the way in which the wine odour was described,
with instead the colour predominating. Likewise, explicit detection
of both food-related and non-food odours was enhanced when the
odour was presented alongside a congruent picture (e.g., ice-cream
picture presented with vanillin, see Gottfried, & Dolan, 2003), and
likewise speed and accuracy of odour detection was faster for con-
gruent odour-colour pairings (e.g., strawberry odour and pink col-
our) than for incongruent pairings (Luisa Dematte, Sanabria, &
Spence, 2006). One explanation for these effects is that the visual
cue sets up an expectation of the flavour to be experienced in the
mouth as a consequence of past associations between the visual
appearance and perceived flavour of similar food stimuli. Indeed,
it has been argued that such associations are likely to be memor-
ised without any explicit attention or learning (Koster, Prescott,
& Koster, 2004), highlighting further the key role of memory in
developing food-based expectancies (Mojet, & Koster, 2005).
An important methodology in examining the role of expecta-
tions in determining our experience of flavour has been to manip-
ulate the congruence between pre-ingestive visual and cognitive
cues and the actual sensory quality (taste and/or odour) once the
sample has been ingested, extrapolating the exploration of the ef-
fects of congruence in interactions between tastes and odours
(Frank, & Byram, 1988; Labbe, Damevin, Vaccher, Morgenegg, &
Martin, 2006). In most circumstances, visual cues will be a reliable
indicator of the actual flavour quality of a food, both in terms of
overall recognition of the nature of the food and also whether
the food is in an appropriate state to be ingested. However, when
there is a lack of congruence between the expected and actual sen-
sory quality of a food, this may lead to perceptual confusion and so
alter the sensory experience itself.
Alongside a clear literature on the extent to which visual
appearance may alter our ability to identify, and to some extent
modify the sensory quality of a food or drink, the extent to which
expectations about flavour also modify our hedonic evaluation of a
food has also received attention (Cardello, 2007; Deliza, & Macfie,
1996). Actual food choice often occurs based on written or verbal
description of a food, even before the actual food has been seen.
Thus, in restaurants our choice is based on expectations of liking
for flavours implicit in descriptions of the potential foods on offer,
with the expectation that the description and actual sensory qual-
ity will be congruent. In relation to food flavour, congruence has
been defined as ‘‘the extent to which two stimuli are appropriate
for combination in a food product” (Schifferstein, & Verlegh,
1996), and has been widely used to denote the impact of percep-
tual similarity between elements in food flavour on changes in sen-
sory quality. For example, perceptual similarity between an odour
and taste was a good predictor of taste intensity (Frank, Shaffer, &
Smith, 1991).
Expectations about the sensory quality of a stimulus can alter
liking and perception of that stimulus in two contrasting ways.
Firstly, the sensed and expected sensory qualities may combine,
so resulting in actual evaluations which are closer to the expected
evaluation than is seen when the same item is evaluated without
prior expectation. These outcome can be explained by assimilation
theory, first proposed in relation to attitudinal change in social
psychology (Hovland, Harvey, & Sherif, 1957), where attitudes
are adjusted by prior expectation. In relation to perception of the
qualities of food stimuli, many studies have reported assimilation
effects, both in relation to affective (liking) evaluations and sensory
evaluations (Cardello, & Sawyer, 1992; Deliza, & Macfie, 1996; Kah-
konen, Tuorila, & Rita, 1996; Lange, Rousseau, & Issanchou, 1999;
Schifferstein, Kole, & Mojet, 1999; Tuorila, Cardello, & Lesher,
1994). For example, verbal descriptions which implied that a prod-
uct (pomegranate juice) was very pleasant (e.g. the statement that
the product scored 8.1 on a 9 point liking scale) or very unpleasant
(e.g., that it scored 1.9 on a 9 point liking scale) generated expecta-
tions in line with these ratings (Cardello, & Sawyer, 1992). Accord-
ingly, a positive expectation lead to a small increase in actual rated
liking on tasting the product (assimilation) although an expecta-
tion of a disliked flavour had minimal effects on actual liking for
the tasted product. Effects on sensory qualities were clearer: being
told that a product had a bitter taste increased rated bitterness on
tasting, while expectations of low bitterness tended to decrease
bitterness evaluations. The authors concluded that the study pro-
vided evidence for the assimilation model: actual and expected
sensory experience combined to generate the overall flavour expe-
rience and liking. In relation to evaluation of liking, recent research
conducted under semi-naturalistic conditions in a cafeteria sup-
ports the idea of assimilation. Thus, the use of evocative descriptive
menu names resulted in stronger positive evaluations of the food
after it had been consumed than when the same food had been la-
belled by nutritionally accurate but non-evocative names (Wan-
sink, van Ittersum, & Painter, 2005). Likewise, labelling a tomato
soup with a name implying a higher quality (e.g., ‘‘Gastronome’s
Connoisseur’s Choice Cream of Tomato” relative to ‘‘McTaggart’s
Lean and Low Tomato”) resulted in significantly higher hedonic
ratings for the same soup regardless of actual nutrient content
(Yeomans, Lartamo, Procter, Lee, & Gray, 2001), and also resulted
in higher ratings of creaminess of the soup.
In the examples above, actual evaluations of foods generally
changed to be more in line with the expected quality even though
there was a discrepancy between the expected and actual qualities
of these stimuli. However, as discussed earlier, although most
studies of effects of expectations on evaluations of foods result in
assimilation, in some cases such discrepancies can lead to a de-
crease in the rated quality (contrast effect) rather than assimilation
For example, a strong expectation that an unusual breath freshener
(Jintan) had a pleasant taste (ie was a form of ‘‘Japanese candy”) re-
sulted in markedly lower liking ratings than when Jintan was as-
sessed without expectation (Zellner, Strickhouser, & Tornow,
2001).
A key question is then what determines whether information
about a product leads to an enhanced evaluation (assimilation)
or a decrease (contrast)? Recent reviews suggest a number of fac-
tors may be important (Cardello, 2007; Schifferstein, 2001). Firstly,
the size of the discrepancy: where the difference between actual
and expected qualities are small, the difference may not be noted,
and so assimilation takes place, whereas if the discrepancy is large,
contrast may occur. This effect is captured well by the affect expec-
tation model (Wilson, Lisle, Kraft, & Wetzel, 1989). Where the dis-
crepancy is not apparent, the expectation is no longer a point of
reference and so is not directly compared with the actual qualities.
In relation to food, several studies report findings consistent with
this idea (see Cardello, 2007 for review). A second factor is the
strength of the expectation: even where there is a large discrep-
ancy between expected and actual properties, assimilation may oc-
cur if the expectation is very strong. In relation to food, an
important test of these ideas was reported by Zellner et al.
(2001). Their participants’ evaluated two novel foods, Jintan and
guanabana nectar, with expectation about liking manipulated by
the information provided beforehand. Assimilation occurred where
expectation were based on specific information about the nature of
the food (e.g., where participants assessing Jintan were told that
other assessors had rated this as very disliked), even when the ex-
tent of the expected dislike was much greater than that seen when
the food was evaluated without prior expectations, but contrast
was seen where the expectation and actual experience were very
different. The findings by Zellner and colleagues are important
since they contrast with a larger literature suggesting assimilation
is the normal response to disconfirmed expectancies with food,
discussed earlier. One reason for this may be that most previous
 M.R. Yeomans et al. / Food Quality and Preference 19 (2008) 565–573
studies have concentrated on subtle differences in quality between
the expected and actual properties of foods, conditions which
would favour assimilation, whereas Zellner et al. (2001) explicitly
explored conditions where expectations could be disconfirmed.
Thus in hedonic evaluations, both assimilation and contrast are
possible outcomes of prior expectations, depending on the
strength, plausibility and nature of the expectations.
The affect expectation model can be applied usefully to under-
standing of expectations of liking, but since it is grounded in affect,
it is less directly applicable to evaluations of sensory rather than
affective quality. Notably, Zellner et al. (2001) only measured hedo-
nic evaluations, and the effect of large disconfirmed expectancies
on sensory rather than hedonic evaluations remains under-studied.
In one early study, participants rated sucrose or quinine solutions
which were either labelled correctly or incorrectly (Carlsmith, &
Aronson, 1963). Disconfirmed expectancies resulted in contrast ef-
fects: quinine was rated as more bitter when labelled as sucrose,
and vice versa. However, the authors interpreted these changes in
sweet and bitter dimensions as a reflection of hedonic changes. Re-
cent work, however, suggests that sweet and bitter ratings are dis-
sociable from hedonic evaluations (e.g., Yeomans, Mobini, Elliman,
Walker, & Stevenson, 2006), which implies that the outcome of the
study by Carlsmith and Aronson (1963) may reflect actual sensory
rather than hedonic changes. Other studies report assimilation of
labelled sensory quality into actual product evaluations (e.g., Capo-
rale, Policastro, Carlucci, & Monteleone, 2006; e.g., Cardello, & Saw-
yer, 1992; Tuorila et al., 1994), but in the context of foods which had
qualities which were similar to expectations, and so may promote
assimilation since the difference between expected and actual qual-
ities may not have been noticeable. Thus there is a need for further
studies exploring the role of expectation in sensory and hedonic
evaluations of foods where expectations are disconfirmed, and this
difference is apparent to the consumer, and that was the aim of the
experiments reported here.
The aim of the three experiments described in this report was to
explore the importance of congruency between the expected fla-
vour (generated by a combination of a simple food label and visual
appearance) and actual sensory experience of a highly novel food
which had the capacity to generate extreme differences between
expectations and actual sensory qualities. To achieve this, we gen-
erated a highly novel food (smoked salmon ice-cream) whose vi-
sual appearance could be interpreted either as a fruit ice-cream
or savoury mousse. Critically, we wanted to determine the extent
to which hedonic and sensory evaluation of the food was influ-
enced by the congruence between the expected and actual flavour.
Where the label predicted the actual sensory experience (i.e., a la-
bel of ‘‘frozen savoury mousse”), we predicted that the food would
be perceived as acceptable. However, where the expectation gener-
ated by the label predicted a sweet food (i.e. when the food was la-
belled as ‘‘ice-cream”), we predicted that liking would be
significantly reduced when the actual food was experienced. More-
over the experience of unexpected sensory qualities (a savoury,
salty fish flavour) might lead to over-estimation of the key sensory
attributes of the salmon ice-cream due to the strong contrast be-
tween expected and perceived flavour.
2. Experiment 1
2.1. Method
2.1.2. Design
The study used a between-subjects design to contrast hedonic
and sensory evaluations of a novel food (smoked-salmon ice-
cream) depending on whether this was presented to assessors as
‘‘ice-cream” or ‘‘frozen savoury mousse”.
567
2.1.3. Participants
The participants were an untrained panel of 32 assessors (24
women and 8 men) selected from staff and students at the Univer-
sity of Sussex who had previously shown an interest in studies
relating to ingestion. Potential volunteers were contacted through
email, and were informed that the study simply involved evalua-
tion of a novel ice-cream. Volunteers who had diabetes, had any
known or suspected food allergy or had a prior diagnosis of an eat-
ing disorder were excluded, and the study was described as unsuit-
able for vegetarians. The participants were the first 32 respondents
who met the study criteria. Participants were assigned at random
to either an informed ICE-CREAM or informed SAVOURY group,
while ensuring the same gender ratio (there were 16 participants,
12 women and 4 men, in each condition). The two groups did not
differ significantly in age.
2.1.4. Test food
The food used in the test was a novel smoked-salmon ice-
cream, based on an original idea by Heston Blumenthal and then
developed further by Firmenich SA in Geneva. The ice-cream was
unusual in that it was not sweet; it contained the following
ingredients:
 565 ml: full fat UHT milk obtained from a local supermarket.
 200 ml: UHT cream with a 35% fat content obtained from a local
supermarket.
 235 g: maltodextrin with a dextrose equivalent of 10 (Star-Dri
10Ò from Tate & Lyle plc).
 230 g: scottish smoked salmon obtained from a local
supermarket.
 8 g: sodium chloride.
 3 g: MSG (monosodium glutamate from Ajinomoto plc).
The cream, milk, maltodextrin, salt and MSG were thoroughly
mixed to be free of any lumps and passed through a kitchen sieve.
900 ml of this mix were placed in the stainless steel beaker from a
Pacojet Machine (Pacojet AG, Bundesstraße 7, CH-6300 Zug/Swit-
zerland). A small aluminium bottle (5 cm diameter) was filled with
table salt and capped. This was placed in the centre of the Pacojet
beaker and was tall enough to protrude above the level of the li-
quid. The whole was covered and placed in a deep freezer over-
night. The next day the salt was poured out of the aluminium
bottle and replaced with warm water; this allowed the bottle to
be easily removed from the frozen mix leaving a central vertical
hole from top to bottom. This hole was packed with the smoked
salmon and the whole container was stored in the deep freezer un-
til needed.
The Pacojet machine is designed to forcibly grind and churn a
portion of a frozen mass into a finely divided purée. In this case
it thoroughly mixes the frozen cream mix with the salmon to a
homogeneous and smooth consistency; in this process ice crys-
tals are reduced to a size that is compatible with a smooth
creamy ice-cream. The overall colour was pink/peach with no
visible sign of salmon pieces. For the sensory tasting sessions,
a single scoop of ice-cream (c. 30 g) was served in a small plastic
tray.
2.1.5. Procedure
Participants attended for a single tasting session, which was
conducted between 1445 and 1600 h in small, air-conditioned
windowless cubicles in the Ingestive Behaviour Unit at Sussex
University, having been instructed to refrain from eating and to
drink only water for the two hours prior to testing. All data were
collected using Sussex Ingestion Pattern Monitor software (see
568 M.R. Yeomans et al. / Food Quality and Preference 19 (2008) 565–573
Yeomans, 2000) run on Apple G3 computers. Initially participants
completed a series of computerised mood and appetite ratings in
the form ‘‘How hwordi do you feel?”, where the adjectives rated
were: clear-headed, drowsy, thirsty, lively, calm, full, nervous, re-
laxed, hunger and nauseous. Ratings were made by positioning a
bar at the point of a horizontal line 500 pixels long which best re-
flected how participants currently felt, and ratings were scored
automatically on a scale from 0 (not at all) to 500 (extremely).
The polarity of the scale varied at random, as did the order in
which ratings were made, and the adjective describing the dimen-
sion to be rated was positioned centrally above the rating scale.
The appetite and key mood ratings (anxious, nauseous and ner-
vous) were used to test for spurious group differences which might
have confounded any differences in hedonic evaluations of the ice-
cream, and the remaining mood evaluations were included both to
ensure familiarity with the computerised ratings, and to disguise
the purpose of the study. After completion of the final mood rating,
the computer signalled the participant to call the experimenter,
who immediately served the ice-cream portion, along with fresh
mineral water for participants to use to cleanse their mouth after
tasting. The key manipulation in the study was the description of
the ice-cream on-screen at the time of evaluation. In the ICE-
CREAM condition, the food was labelled as ‘‘ice-cream”, but in
the SAVOURY condition the description given to participants was
‘‘frozen savoury mousse”. Evaluations of the ice-cream were all
made using the same style of ratings as with the mood and appe-
tite scales. Participants made one hedonic rating (‘‘How pleasant
is the taste?”), a rating of flavour strength (strong) and seven
flavour descriptors (fruity, savoury, sour, creamy, salty, bitter and
sweet). As with the mood/appetite ratings, the order in which
evaluations were made was randomised. Once these ratings had
been completed, the participant was debriefed, and rewarded
either with a small payment or course credits for research
participation.
Fig. 1. Rated pleasantness (A), saltiness (B), strength of flavour (C) and bitterness (D) of smoked salmon ice-cream depending on whether this was labelled as frozen savoury
mousse or ice-cream. Data are mean ± SEM.
2.1.6. Data analysis
Initial analyses were aimed at confirming that the between-
groups contrast of hedonic evaluation of the ice-cream was not
invalidated by spurious differences in appetite and mood at the
time of testing. Thus each mood and appetite rating was contrasted
between the ‘‘ice-cream” and ‘‘savoury mousse” conditions using
ANOVA, with gender of assessor as a controlled factor. The same
analysis was then used to assess the rated characteristics of the
ice-cream between the two groups.
2.2. Results
Assessors who were pre-informed that the food was a frozen
savoury mousse (condition SAVOURY) rated the flavour of the
smoked salmon ice-cream as significantly more pleasant than
did those who experienced the same food simply labelled as
ice-cream (F(1, 28) = 11.34, p < 0.005), with the latter condition
finding the flavour extremely aversive, compared to a mild dis-
like in the savoury condition (Fig. 1a). Thus expectancy generated
by the physical appearance of the food combined with a simple
written descriptor had a large effect on rated flavour pleasant-
ness. In addition, ratings of saltiness were significantly greater
in the ICE-CREAM than SAVOURY condition (F(1, 28) = 5.96,
p < 0.05: Fig. 1b), and the ICE-CREAM condition also perceived
the flavour as stronger overall (F(1, 28) = 7.63, p < 0.01: Fig. 1c).
Ratings of bitterness were variable, but again assessors in the
ICE-CREAM condition tended to rate the flavour as more bitter
than did those in the SAVOURY condition (F(1, 28) = 3.32,
p = 0.079: Fig. 1d). None of the other sensory ratings differed be-
tween conditions.
Evaluation of baseline ratings (Table 1) confirmed no spurious
differences in appetite (hunger, fullness or thirst) or mood (anxi-
ety, nervous and nausea) that could account for differences in eval-
uation of ice-cream flavour, and no analysis found any differences
 M.R. Yeomans et al. / Food Quality and Preference 19 (2008) 565–573 569

Table 1
Baseline ratings of appetite and mood prior to flavour evaluation in Experiment 1
Attribute rated ICE-CREAM condition SAVOURY condition
Full 239.1 ± 18.9 231.7 ± 27.1
Hunger 192.4 ± 25.9 203.5 ± 30.7
Thirst 242.5 ± 26.9 234.3 ± 26.5
Nauseous 65.7 ± 27.6 65.5 ± 24.5
Nervous 157.2. ± 28.3 152.9 ± 29.4
between female and male assessors either in baseline mood and
appetite or in assessment of the ice-cream.
2.3. Conclusion
These data suggest that the sensory qualities implied by a sim-
ple food label affected the perceived pleasantness of flavour of a
novel, and also modified actual perception of flavour components.
More importantly, these data imply that inaccurate expectations
about food flavour can lead to strongly aversive responses when
the food was tasted, in line with a limited existing literature also
showing hedonic contrast effects with food stimuli (Cardello, &
Sawyer, 1992; Zellner et al., 2001). These data thus consistent with
the idea that flavour perception is an integration of sensory infor-
mation with past memory of similar stimuli predicted by the visual
qualities, and accompanying written descriptor, of the rated food.
However, two methodological issues limit interpretation of Exper-
iment 1. Firstly, the recruitment strategy identified the study as an
‘‘ice-cream” tasting study during the recruitment of participants.
Consequently, the specific information that the product was a ‘‘fro-
zen savoury mousse” may have had less impact than if participants
had not had the initial expectation of ice-cream. Secondly, while
Experiment 1 established large differences in evaluation of the
smoked salmon ice-cream depending on the associated label, it is
impossible to determine whether this resulted from the negative
impact of incongruous ice-cream label, or a positive effect of con-
gruous frozen savoury mousse label. Experiment 2 was designed
to counter these limitations.
3. Experiment 2
3.1. Method
3.1.1. Design
A between-subjects design was again used to contrast hedonic
and sensory evaluations of a novel food (smoked salmon ice-
cream) with the same two conditions (‘‘ice-cream” or ‘‘frozen
savoury mousse”) as in Experiment 1, and a third control condition
which provided no descriptive information (‘‘Food 386”).
3.1.2. Participants
The participants were 44 untrained assessors. The recruitment
was the same as in Experiment 1 except that the description of
the study in all communications prior to testing, and in the infor-
mation sheet for participants, described the study as ‘‘assessing a
novel food”. Participants were assigned at random to the three con-
ditions, with the original aim of using 15 participants in each con-
dition. However, one participant failed to attend their test session,
and the final composition of the three panels (Table 2) had 15 par-
ticipants in the SAVOURY and CONTROL conditions, but 14 in the
ICE-CREAM condition.
3.1.3. Procedure and test food
The same test food was used as in Experiment 1, and the proce-
dure was identical with the sole exception that the on-screen
descriptor of the food to be assessed in the CONTROL condition
was ‘‘Food 386”.
3.1.4. Data analysis
Each rating was contrasted between the three conditions using
one-way ANOVA, with gender of assessor as a controlled factor.
Where significant effects of condition were found, post-hoc con-
trasts were used to confirm which differences between the three
conditions were significant, with significance adjusted for multi-
ple-contrasts using the Bonferroni adjustment.
3.2. Results
Analysis of pleasantness ratings found an overall effect of con-
dition [F(2, 41) = 5.04, p < 0.05]. Inspection of the pleasantness
data, however, revealed that one participant was a significant out-
lier (data more than 2SD from the group mean) in the ICE-CREAM
condition. This participant gave a pleasantness rating of 475 to the
smoked salmon ice-cream labelled as ice-cream, compared to a
group mean of 40.1 ± 36.4. Notably, this participant, who was
Spanish, commented that he had experience of savoury ice-cream,
which he liked. Exclusion of these data increased the significance
of the effect of condition on pleasantness [F(2, 40) = 9.68,
p < 0.001]. Protected contrasts revealed that ratings in the ice-
cream condition were significantly less than in the other two con-
ditions, but ratings in the control and savoury condition did not
differ significantly (Fig. 2a).
The effects of labelling on sensory evaluation seen in Experi-
ment 1 were partly replicated. Saltiness also varied with condition
[F(2, 40) = 5.08, p < 0.05], and as in Experiment 1 rated saltiness
was higher in the ice-cream than savoury condition (Fig. 2b), but
savoury and control conditions did not differ significantly. As in
Experiment 1, perceived flavour strength varied with condition
[F(2, 40) = 5.69, p < 0.001], with higher ratings in the ICE-CREAM
than savoury or control conditions (Fig. 2c). There was also a trend
for higher savoury ratings in the ice-cream condition, but this did
not reach overall significance [F(2, 40) = 1.94, NS]. Although the
mean ratings for bitterness in Experiment 2 were in the same
direction as Experiment, with higher bitterness ratings in the ice-
cream than savoury condition, the overall effect of condition was
not significant [F(2, 40) = 0.88, NS].
No significant effects of condition were found on creaminess,
fruitiness, sourness or sweetness evaluations, although in many
ratings there were trends for ratings in the ice-cream condition
standing out as different to the other conditions (tending to be
more sour, less sweet, etc.).
3.3. Conclusion

Table 2 Experiment 2 suggests that it was the lack of congruity between

Characteristics of the participants in Experiment 2 the label ‘‘ice-cream” and the experienced sensory quality of the
smoked salmon ice-cream that resulted in the strong aversive re-
Condition
sponse in the ICE-CREAM condition in both Experiment 1 and 2
Control Ice-cream Savoury
since reactions in the conditions where the label was informative
Age 21.3 ± 1.0 21.4 ± 0.7 21.1 ± 0.5 and accurate (‘‘frozen savoury mousse”) and uninformative (CON-
BMI 22.7 ± 0.6 22.1 ± 0.6 22.6 ± 0.5
TROL condition labelled ‘‘Food 386”) were very similar. The differ-
Gender (F/M) 12/3 11/3 11/4
ences in hedonic evaluation between the ICE-CREAM and
570 M.R. Yeomans et al. / Food Quality and Preference 19 (2008) 565–573

Fig. 2. Rated pleasantness (A), saltiness (B) and strength of flavour (C) of smoked salmon ice-cream depending on whether this was labelled as ice-cream, frozen savoury
mousse or a neutral food (Control). Data are mean ± SEM.

SAVOURY conditions in Experiments 1 and 2 were very similar, this phase. After completion of the final rating of expected flavour,
with the food rated as having a stronger and more salty flavour an additional rating was made based on how confident participants
in the ice-cream label condition, although the hedonic differences were about their evaluation of the food by label and appearance.
were much greater than sensory differences. The question was phrased ‘‘How confident are you that you have
The similarity in assessment of the salmon ice-cream in the accurately described the flavour of this food?”, rated form not at
SAVOURY and CONTROL conditions was surprising since the all (0) to totally (500). Once completed, the food was placed in
appearance of the food might have been predicted to generate front of the participants to allow them to consume a small portion
expectations of a sweet ice-cream. However, a shortcoming in and make a second set of evaluations, using the same procedure as
Experiments 1 and 2 was that differences in expectation between in Experiments 1 and 2.
conditions were implied but not tested explicitly. Previous re-
search has assessed the role of expectation by obtaining separate 4.1.4. Data analysis
ratings based on appearance alone prior to sensory testing (e.g., The expected and actual evaluations of the ice-cream were con-
Cardello, & Sawyer, 1992). The same approach was discounted here trasted using ANOVA, with label between-subjects factor and type
because of concerns that the inspection of the food during the eval- of evaluation (expected or actual) within-subjects. Gender of asses-
uation of the appearance may have led to participants detecting sor was again included as a controlled factor. Where significant ef-
the faint odour of fish from the ice-cream, so compromising the la- fects of condition were found, post-hoc contrasts were used to
bel manipulation. However, to test whether the manipulations confirm which differences between the three conditions were sig-
used in the present study did generate the inferred expectations, nificant, with significance adjusted for multiple-contrasts using the
a final study examined contrasted the effects of the label manipu- Bonferroni adjustment.
lations on evaluations of both the expected and actual sensory and
hedonic evaluations of the ice-cream in the three conditions used 5. Results
in Experiment 2.
5.1. Expected flavour evaluations
4. Experiment 3
Average sensory and hedonic evaluations of the expected fla-
4.1. Method vour of the salmon ice-cream in the three conditions are summa-
rised in Table 3. These data confirm that the label applied to the
4.1.1. Design
A between-subjects design was used to contrast anticipatory Table 3
and actual hedonic and sensory evaluations of the same novel food Expected sensory and hedonic evaluations of the salmon ice-cream in the three label
(smoked salmon ice-cream) in the three label conditions used in conditions in Experiment 3
Experiment 2 (ICE-CREAM, SAVOURY and CONTROL). Evaluation Label condition Main effect of
condition1
Ice-cream Savoury Food 386
4.1.2. Participants mousse
The participants were 60 untrained assessors, 48 women and 12
Pleasant 375a ± 13 241b ± 23 258b ± 18 p < 0.001
men. The recruitment was the same as that in Experiment 2, and
Savoury 50a ± 8 482c ± 5 121b ± 19 p < 0.001
participants were assigned at random to the three conditions, with Salty 31a ± 6 323b ± 30 69a ± 17 p < 0.001
20 participants in each condition. Sweet 340a ± 28 89c ± 13 275 b ± 19 p < 0.001
Creamy 389a ± 13 314 b ± 18 320b ± 16 p < 0.01
Fruity 322a ± 15 76c ± 13 225b ± 26 p < 0.001
4.1.3. Procedure and test food
Bitter 87 ± 17 104 ± 17 118 ± 20 NS
The same test food was used as in Experiments 1 and 2. How- Sour 56 ± 13 62 ± 11 47 ± 9 NS
ever, in this case when the food was first presented, participants Strong 256a ± 15 387b ± 11 300a ± 20 p < 0.001
firstly made the same set of evaluations as in the previous studies Confidence 329a ± 17 284a ± 19 146b ± 25 p < 0.001
but based purely on the appearance of the food. The food was Data are mean ± SEM, n = 20 in each group
placed towards the back of the test table and participants were in- 1
Value is significance for main effect of condition from one-way ANOVA. Means
structed to base their evaluations on the appearance alone during with different superscripts differ significantly (bonferroni contrast).
 M.R. Yeomans et al. / Food Quality and Preference 19 (2008) 565–573 571

food had a large impact on expected sensory and hedonic quality. Table 4
The expected flavour was rated to be significantly more pleasant, Sensory and hedonic evaluations of the salmon ice-cream in the three label conditions
in Experiment 3
sweet, fruity and creamy, and significantly less savoury and salty,
in the ICE-CREAM than SAVOURY condition. Expected ratings in Evaluation Label condition Main effect of
the CONTROL condition, where no informative label was given, condition1
Ice-cream Savoury Food 386
were less consistent: ratings for pleasant and creamy were similar mousse
in CONTROL and SAVOURY conditions, but ratings of salty and Pleasant 21a ± 8 145b ± 24 140b ± 29 p < 0.001
strong were similar in CONTROL and ICE-CREAM conditions. On Savoury 434 ± 9 407 ± 16 393 ± 11 NS
three ratings, sweet, fruity and savoury, the ratings in the CON- Salty 412a ± 11 361b ± 16 388ab ± 11 p < 0.05
Sweet 28a ± 9 86b ± 19 83b ± 20 p < 0.05
TROL condition fell between those in the other conditions, suggest-
Creamy 363 ± 11 355 ± 9 351 ± 13 NS
ing that the absence of an informative label made the food Fruity 16 ± 5 32 ± 13 24 ± 8 NS
ambiguous. This was further reflected in the confidence ratings Bitter 135 ± 23 84 ± 13 121 ± 20 NS
for the expected flavour, where confidence ratings were signifi- Sour 171 ± 23 113 ± 15 121 ± 24 NS
Strong 439 ± 11 410 ± 13 405 ± 12 NS
cantly lower in the CONTROL than ICE-CREAM or SAVOURY condi-
tions (Table 3). Data are mean ± SEM, n = 20 in each group.
1
Value is significance for main effect of condition from one-way ANOVA. Means
5.2. Actual flavour evaluations with different superscripts differ significantly (bonferroni contrast).

The evaluations of the flavour in the three conditions were very

similar to those in Experiment 2 (Table 4). As before, flavour pleas-
antness was significantly affected by food label [F(2, 56) = 10.65,
p < 0.001], with significantly lower ratings in the ICE-CREAM than
CONTROL and SAVOURY conditions. Actual sensory evaluations
were much less affected by label, although both ratings of sweet
[F(2, 56) = 4.16, p < 0.05] and salty [F(2, 56) = 4.21, p < 0.05] did differ
between conditions, with the test food rated as less sweet and more
salty in the ICE-CREAM than CONTROL or SAVOURY conditions.
5.3. Relationship between expected and actual flavour
To test whether expectations were assimilated into flavour
evaluations, expected and actual flavours were contrasted by 2-
way ANOVA, with condition and rating type (expected or actual
flavour) as factors. For pleasantness, there was a significant inter-
action between condition and rating type [F(2, 54) = 18.46,
p < 0.001], and these data are shown in Fig. 3a. In all cases actual
pleasantness was lower than expected, but where the label ICE-
CREAM was used, expected pleasantness was the largest, but actual
pleasantness extremely low. Thus rather than assimilation, the lack
of congruence between the expected and actual flavour generated
an aversive response. This was not seen where the flavour was pre-
dicted by the label ‘‘frozen savour mousse”. Expected flavour rat-
ings in the CONTROL condition also suggested participants had
some expectation of a sweet, fruity taste, although less so than in
the ICE-CREAM condition, but this did not result in such a large
reduction in pleasantness on tasting the food. A similar, but smal-
ler, effect of labelling was seen with sweetness ratings (Fig. 3b),
where again the interaction between rating type (expected or ac-
tual) and condition was significant [F(2, 54 = 18.46, p < 0.001].
Here, although expected sweetness was markedly different be-
tween SAVOURY and CONTROL conditions, actual ratings were
very similar. In contrast, the expected sweetness was very high
in the ICE-CREAM condition, but actual ratings were significantly
lower than in the other conditions.
6. Discussion
Rated pleasantness of the flavour of a novel savoury ice-cream
was highly dependent on the label presented when the food was
served. The label ‘‘ice-cream” reliably resulted in lower pleasant-
ness ratings than did the label ‘‘frozen savoury mousse” (Experi-
ments 1 and 2). However, when a neutral control label was
included (Experiment 2), responses to the savoury and control
conditions were very similar. These effects were seen both with
hedonic and sensory changes, with consistent finding of higher rat-

Fig. 3. Expected and actual pleasantness (A) and sweetness (B) of smoked salmon ice-cream depending on whether this was labelled as ice-cream, frozen savoury mousse or a
neutral food (Control). Data are mean ± SEM.
572 M.R. Yeomans et al. / Food Quality and Preference 19 (2008) 565–573
ings of flavour strength and saltiness for the food when labelled as
‘‘ice-cream”.The implication is that the ‘‘ice-cream” label generated
a strong expectation of a sweet, fruity flavour, and the surprise of a
strong, salty fish-flavoured food resulted in very low pleasantness
ratings (many participants verbally described the food as disgust-
ing). Experiment 3 explicitly measured expected and actual evalu-
ations, and confirmed the interpretation of Experiments 1 and 2
that the ‘‘ice-cream” label did generate an expectation of a sweet,
fruity flavour, which then contrasted strongly with the actual
experience of salty fish.
Although past research has shown that expectations generated
by food labels can modify liking for food flavours, the majority of
these studies have found assimilation of the expected and actual
experience (Cardello, Maller, Masor, Dubose, & Edelman, 1985;
Cardello, & Sawyer, 1992; Wansink et al., 2005). However the pres-
ent study demonstrates that where the expected and actual sen-
sory experience are very different (i.e., the contrast of an
expectation for sweet and fruity relative to actual salty fish), the
disconfirmed expectation leads to a strong contrast effect, and con-
sequent dislike and rejection of the test food, adding to the very
few studies reporting contrast effects with food stimuli (e.g., Car-
dello, & Sawyer, 1992; Zellner et al., 2001). Thus where the differ-
ence between expectation and actual sensory quality was very
large, the experience was highly unpleasant even for a food that
could be perceived as pleasant on its own (the pleasantness ratings
for the salmon ice-cream in the conditions other than ‘‘ice-cream”
were close to neutral, 250 on the hedonic scale used here). Thus
where expectations were disconfirmed and the actual sensory
experience was highly novel and surprising, the reaction was one
of extreme dislike. Although these conditions would rarely be
experienced by a consumer in real-life, the extent to which expec-
tations influenced liking in the present context demonstrate the
potential for expectations to moderate hedonic evaluation where
there is a clear mis-match between expected and perceived sen-
sory quality. It was also notable that positive expectations about
the novel food generated by the label ‘‘frozen savoury mousse”
did not lead to assimilation of the expected flavour since no differ-
ences emerged between the congruent and neutral food label con-
ditions in Experiment 3. A possible explanation for this might be
that the expectations generated by the congruent label were not
sufficiently strong to facilitate assimilation when the actual food
was tasted, perhaps driven by the unique novelty of the food
tested. This conclusion is supported in part by the finding that con-
fidence ratings in the SAVOURY condition were less than in ICE-
CREAM, but notably both were greater than in the CONTROL con-
dition. However, the large difference in confidence in ratings be-
tween SAVOURY and CONTROL conditions might have been
expected to promote assimilation in the SAVOURY condition. An-
other possible explanation for the lack of assimilation in the
SAVOURY condition was that the description of the study as eval-
uation of a novel food generated an expectation of novelty, which
alerted reduced the surprise when salmon ice-cream was tasted in
the CONTROL condition without giving any confidence of what
sensory quality they would experience. This possibility could be
assessed in future studies.
The present data also confirm and extend previous findings on
the role of expectation on sensory evaluation, particularly in rela-
tion to evidence for a contrast effect when expectations about sen-
sory quality were disconfirmed, a phenomena which has been
rarely reported for the effects of expectations generated by infor-
mation prior to tasting for food stimuli, where assimilation is again
the normal response to disconfirmed expectations. Thus where a
sweet fruity flavour was expected but a salty fish flavour experi-
enced (the ICE-CREAM condition), the unexpected qualities tended
to be rated as stronger sensory experiences than in conditions
where either the expectation was for a similar product (the
SAVOURY condition), or where information gave no expectation
(the uninformative CONTROL food label). These data are important,
as they provide a rare example of contrast effects in sensory eval-
uation generated by expectations about flavour in a food context.
Although interpreted at the time as evidence of hedonic change,
data in the study by Carlsmith and Aronson (1963) also show con-
trast effects, with sweet solutions rated as less sweet when bitter-
ness was predicted and vice versa. Contrast effects are more readily
seen in perceptual studies where evaluation of a context stimulus
of higher or lower intensity alters evaluation of subsequent target
stimuli (e.g., Conner, Land, & Booth, 1987; Riskey, Parducci, & Beau-
champ, 1979; see Schifferstein, 2001 for review), but no such con-
text stimuli preceded evaluation of the test stimulus in the present
study, suggesting that adaptation or range-frequency effects (Par-
ducci, 1965) cannot explain the current findings of exaggerated
sensory experience. Thus the present data suggest that disconfir-
mation about sensory expectation can lead to increased intensity
of the unexpected sensory quality, suggesting some top-down con-
trol influence on sensory experience in this context.
The current studies also suggest that beliefs about the nature of a
food generated by the preceding food label were more powerful
than were the visual cues alone. Thus the differences in sensory
and hedonic evaluation of the salmon ice-cream between the three
test conditions all involved the same visual cues, yet responses to
visual cues alone (the CONTROL condition where the food label
was uninformative in Experiments 2 and 3) and visual cues with
an accurate food label (the SAVOURY condition) were the same.
Only where the food label was congruent with the visual cues but
misleading about the actual sensory quality (the ICE-CREAM condi-
tion) did differences in sensory and hedonic evaluation emerge.
A limitation of the first two experiments in the current study
was that no actual evaluation of the expected pleasantness or fla-
vour quality of the food was made before the food was consumed.
Although explicit data regarding anticipated flavour would have
some value here, we were concerned that asking participants to
evaluate their expectations prior to tasting may have led to them
detect the aroma of fish, and so compromised the lack of congruity
between the label and sensory qualities. Experiment 3, however,
did measure expected and actual ratings and the similarity in sen-
sory evaluations for the actual flavour of the ice-cream in Experi-
ments 2 and 3 suggest that completion of expected ratings did
not impact adversely on subsequent hedonic and sensory evalua-
tions. The inclusion of expected and actual ratings in Experiment
3 did demonstrate the extent of the difference between expected
and actual pleasantness.
In summary, these three experiments provide striking evidence
that expectations play a major role in generating hedonic re-
sponses to food stimuli, such that the same food was rated as
acceptable when the flavour was expected but close to disgusting
when the expected and actual flavour were very different. These
data thus add to a small set of studies showing hedonic and sen-
sory contrast effects in flavour perception.
References
Caporale, G., Policastro, S., Carlucci, A., & Monteleone, E. (2006). Consumer
expectations for sensory properties in virgin olive oils. Food Quality and
Preference, 17, 116–125.
Cardello, A. V. (2007). Measuring consumer expectations to improve food product
development. In H. J. H. Macfie (Ed.), Consumer-led food product development
(pp. 223–261). Cambridge: Woodhead publishing.
Cardello, A. V., Maller, O., Masor, H. B., Dubose, C., & Edelman, B. (1985). Role of
consumer expectancies in the acceptance of novel foods. Journal of Food Science,
50, 1707–1718.
Cardello, A. V., & Sawyer, F. M. (1992). Effects of disconfirmed consumer
expectations on food acceptability. Journal of Sensory Studies, 7.
Carlsmith, J. M., & Aronson, E. (1963). Some hedonic consequences of the
confirmation and disconfirmation of expectancies. Journal of Abnormal and
Social Psychology, 66, 151–156.
 M.R. Yeomans et al. / Food Quality and Preference 19 (2008) 565–573
Conner, M. T., Land, D. G., & Booth, D. A. (1987). Effect of stimulus range on
judgements of sweetness intensity in a lime drink. British Journal of Psychology,
78, 357–364.
Deliza, R., & Macfie, H. J. H. (1996). The generation of sensory expectation by
external cues and its effect on sensory perception and hedonic ratings – A
review. Journal of Sensory Studies, 11, 103–128.
Delwiche, J. (2004). The impact of perceptual interactions on perceived flavor. Food
Quality and Preference, 15, 137–146.
Dubose, C., Cardello, A. V., & Maller, O. (1980). Effects of colorants and flavorants on
identifiaiton, perceived flavor intensity, and hedonic quality of fruit-flavored
beverages and cake. Journal of Food Science, 45, 1393–1399.
Frank, R. A., & Byram, J. (1988). Taste-smell interactions are tastant and odorant
dependent. Chemical Senses, 13, 445–455.
Frank, R. A., Shaffer, G., & Smith, D. V. (1991). Taste-odor similarities predict
taste enhancement and suppression in taste-odor mixture. Chemical Senses, 16,
523.
Gottfried, J. A., & Dolan, R. J. (2003). The nose smells what the eye sees: Crossmodal
visual facilitation of human olfactory perception. Neuron, 39, 375–386.
Hovland, C. I., Harvey, O. H., & Sherif, M. (1957). Assimilation and contrast effects in
reactions to communication and attitude change. Journal of Abnormal and Social
Psychology, 55, 244–252.
Kahkonen, P., Tuorila, H., & Rita, H. (1996). How information enhances acceptability
of a low-fat spread. Food Quality and Preference, 7, 87–94.
Keast, R. S. J., Dalton, P. H., & Breslin, P. A. S. (2004). Flavour interactions at the
sensory level. In A. J. Taylor & D. D. Roberts (Eds.), Flavour perception
(pp. 228–255). Oxford: Blackwell Scientific.
Koster, M. A., Prescott, J., & Koster, E. P. (2004). Incidental learning and memory for
three basic tastes in food. Chemical Senses, 29, 441–453.
Labbe, D., Damevin, L., Vaccher, C., Morgenegg, C., & Martin, N. (2006). Modulation
of perceived taste by olfaction in familiar and unfamiliar beverages. Food Quality
and Preference, 17, 582–589.
Lange, C., Rousseau, F., & Issanchou, S. (1999). Expectations, liking and purchase
behaviour under economical constraint. Food Quality and Preference, 10, 31–39.
Luisa Dematte, M., Sanabria, D., & Spence, C. (2006). Cross-modal associations
between odors and colors. Chemical Senses, 31, 531–538.
Mojet, J., & Koster, E. P. (2005). Sensory memory and food texture. Food Quality and
Preference, 16, 251–266.
Morrot, G., Brochet, F., & Dubourdieu, D. (2001). The color of odors. Brain and
Language, 79, 309–320.
573
Parducci, A. (1965). Category judgement: A range-frequency model. Psychological
Review, 72, 407–418.
Riskey, D. R., Parducci, A., & Beauchamp, G. K. (1979). Effects of context in
judgements of sweetness and pleasantness. Perception and Psychophysics, 26,
171–176.
Schifferstein, H. N., & Verlegh, P. W. (1996). The role of congruency and
pleasantness in odor-induced taste enrichment. Acta Psychologica, 94, 87–105.
Schifferstein, H. N. J. (2001). Effects of product beliefs on product perception and
liking. In L. J. Frewer, E. Risvik, & H. N. J. Schifferstein (Eds.), Food, people and
society: A european perspective on consumer’s food choices (pp. 73–96). London:
Springer-verlag.
Schifferstein, H. N. J., Kole, A. P. W., & Mojet, J. (1999). Asymmetry in the
disconfirmation of expectations for natural yogurt. Appetite, 32, 307–329.
Small, D. M., Jones-Gotman, M., Zatorre, R. J., Petrides, M., & Evans, A. C. (1997).
Flavor processing: More than the sum of its parts. Neuroreport, 8, 3913–3917.
Stillman, J. A. (1993). Color influences flavor identification in fruit-flavoured
beverages. Journal of Food Science, 58, 810–812.
Teerling, A. (1992). The colour of taste. Chemical Senses, 17, 886.
Tuorila, H., Cardello, A. V., & Lesher, L. L. (1994). Antecedents and consequences of
expectations related to fat-free and regular-fat foods. Appetite, 23, 247–263.
Wansink, B., van Ittersum, K., & Painter, J. E. (2005). How descriptive food names
bias sensory perceptions in restaurants. Food Quality and Preference, 16,
393–400.
Wilson, T. D., Lisle, D. J., Kraft, D., & Wetzel, C. G. (1989). Preferences as expectation-
driven inferences: Effects of affective expectations on affective experience.
Journal of Personality and Social Psychology, 56, 519–530.
Yeomans, M. R. (2000). Rating changes over the course of meals: What do they tell
us about motivation to eat? Neuroscience and Biobehavioral Reviews, 24,
249–259.
Yeomans, M. R., Lartamo, S., Procter, E. L., Lee, M. D., & Gray, R. W. (2001). The actual,
but not labelled, fat content of a soup preload alters short-term appetite in
healthy men. Physiology and Behavior, 73, 533–540.
Yeomans, M. R., Mobini, S., Elliman, T. D., Walker, H. C., & Stevenson, R. J. (2006).
Hedonic and sensory characteristics of odors conditioned by pairing with
tastants in humans. Journal of Experimental Psychology: Animal Behavior
Processes, 32, 215–228.
Zellner, D. A., Strickhouser, D., & Tornow, C. E. (2001). Disconfirmed hedonic
expectations produce perceptual contrast, not assimilation. American Journal of
Psychology, 117, 363–387.