Content uploaded by Rachel Herz
Author content
All content in this area was uploaded by Rachel Herz on Jul 22, 2014
Content may be subject to copyright.
1 Introduction
``A rose by any other name would not smell as sweet.''
According to Rozin and Fallon (1987, page 24) with regard to odor: ``It is the subject's
conc epti o n of the object, rather than the sensory properties of the object, that primar i ly
determine its hedonic value''. Mark Twain's The Invalid's Story (Clemens 1882) clearly
illustrates this point. In this story, the protagonist is a stowaway on a railway car and is
lying beside a large sack from which a s mell is emanating. T he character is paranoid
about being caught and starts to b elieve that the sack beside him contains a dead body.
As the carr iage warms and the smell intensifies his fears are ever more confirmed.
The story climaxes at the point where the stowaway can bear the smell no longer and
jumps from the moving railway car, ultimately leading to his own death
ö
never to
find out that the sack contained just a lot of ``innocent cheese''. This story shows how
the p erceiver's context produced an expectation of what the source of the smell was
that made his perceptual interpretation unbearably negative. Had the same individual
be en at a dinner party and smelled the same stimulus, thoughts of appetizing food
would have likely entered his head, and his response to the same odor would have
be en entirely different.
It has been shown that visual and verbal context effects can influence odor quality
perception. Zellner and Kautz (1990) reported that the perceived intensity of food
odors was increased when the odor extracts were p resented in colored as compared
with colourless liquid. Likewise, when panelists rated fragrances with or without brand
labels, differences in ratings of both sweetness and lik ing were found (Moskowitz
1979). Comparable context effects have been shown in other sensory systems. In v ision,
Duncker (1939) found that the shape of an object altered its perceived color. In his
study, a leaf-shaped object was judged as more green than a donkey composed of
the same color. Similarly, Delk and Fillenbaum (1965) found that red-associated objects
such as hearts or lip s were perceived as more red than identically c olored circles
or mushroom s. These examples illustrate that congruen cy between conc ept and
physical attribute can alter qualitative aspects of perception. However, the dramatic
The influence of verbal labeling on the perception of odors:
E v ide nce for olfactory illusions?
Perception, 2001 , volume 30, pages 381 ^ 391
Rachel S Herz, Julia von Clef
Department of Psychology, Box 1853, Brown University, Providence, RI 02912, USA;
e-mail: Rachel
Herz@Brown.edu
Received 17 July 2000, in revised form 25 October 2000
Abstract. Using the defin ition that an illusion is observed when a stimulus is invariant but
context alters its perception, we examined whether verbal context could produce olfactory illu-
sions. To te st this effect, we chose five odors with minimally fixed sources and that could be
interpreted with various hedonic connotations. The odors were violet leaf, patchouli, pine oil,
menthol, and a 1 : 1 mixture of isovaleric and bu tyric acids. Subjects individually sniffed each
odor at two di fferent sessions separated by one week. At each session an odor was given a
different verbal label (either positive or negative) and subjects rated the odors on several hedonic
scales and provided perceptual and interpretative responses to them. Results showed that the
perception of an odor could be significantly influenced by the label provided for it. We propose
that the c ases where verbal labels inverted odor perce ption are the first empirical demo nstrations
of olfactory illusions.
DOI:10.1068/p3179
misperceptio ns demonstrate d by illusions are perhap s a b etter analogy for the olfactory
example of mistaking cheese for a dead body.
Most generally, an illusio n is a p e rception that does not correspo nd to reality (see
Gregory 1997). Most visual illusions are created by the visual context that surrounds
a target stimulus (Luo and Wang 1997). The Mu
«
ller-Lyer illusion is the most definitive
example, where the direction of the t ails of the figure, either feathering (making the
line look longer) or arrowhead (making the line look shorter), influence perceived
length. A number of other experimental examinations support the influence of visual
context in creating visual illusions (Ku
«
nn apas 1955; Schiffman and Thompson 1978;
Spivey-Knowlton and Bridgeman 1993). However, illusory effects are not limited to the
situation where aspects within the same sensory system induce the illusion (ie visual
context and a visual stimulus).
Social context has been shown to produce visual illusions in the autokinesis effect
(a fixed light in a dark room appears to move). What is interesting about the auto-
kinesis effe ct is that the degree of light movement is different for e ach observer, and
is very susceptible to suggestion and group pres sure. Sherif (1935, 1937) demonstrated
that if ins tructions were given as to the direction the light should move, subjects
tended to see movement in that direction. Even more striking, Rechtschaffen and
Mednich (1955) told subjects that the light would spell out spe cific words and the
subjects reported that it did. Moreover, the words were relevant to each subject's per-
sonal life and were embarassing to some of them.
These instances show how verbal context (ie the experimenter's instructions) can
induce visual illusions. Verbal labels as manipulators for olfactory p e rception may be
even more potent than they are in vision be c ause we are so visually and verbally
or iented and automatically search for visual ^ verbal referents to our olfactory experi-
ence s (Herz 2000). Moreover, for olfaction, despite the contrast of two modalities, verbal
labels may be more reliable perceptual frames fo r olfactory perception than other odors
are b e cause of the sensory confusion arising from odor mi xing
ö
for example smelling
ö
``cheese'' in a restaurant filled with other food aromas. Thus, the word ``cheese'' is an
ec ologically relevant and uncontaminated perceptual reference frame for an olfactory
experience. Verbal labels for odors and the visual arrowheads and feathers in the Mu
«
l-
ler-Lyer illus ion can thus be considered analogou s in that each provides perceptually
expected and `reliable' information regarding the frame of reference for the stimulus.
It is well acknowledged that verbal context regul arly causes olfactory misperc eptio ns;
however, they are usually minor and extensions of real ity. For example, smelling garlic
while walking past a pizza restaurant and misperceiving the smell to be pizza itself
(see Engen 1987). However, in some instanc es the effect of verbal context can be very
dramatic
ö
for example, misperceiving the same stimulus to b e either cheese or a dead
body depending on the verbal context the odor is perceived through. Importantly,
although all odors are susceptible to verbal cueing, some may only be susceptible to
verbal-context misperceptions within close approximation to the ir object category. For
example, telling some one that the smell of lime is actually lemon may be acceptable, but
telling them that it is coconut is less b eli evable, and th at it is pizza even less so (Cain
and Potts 1996). For the present research we were interested in odors whi ch could
show dramatic verbal context effects and which would be believed to be produc ts of
both very pleasant and unpleasant odor sources. By definition such odors should not
have a fixed source but at least two possible anchors with large differences in hedonic
connotation from each other. We call ed these odors ``ambiguous'' and develop ed a set of
five odors to test in the present experiment. We did not expect al l odors to be equally
affected by the verbal-label manipulation, but did not know to what degree they would
vary and which odors would be most strongly or weakly affected. Therefore, the individ-
ual odors used were separately assessed as an independent factor i n the present study.
38 2 R S Herz, J von Clef
The aim of our study was to evaluate, as potential evidence for olfactory illusions,
the effect of providing verbal labels with opposing hedonic connotations on the percep-
tion of odors. Using the definition that an illusion occurs when a stimulus is invariant but
context alter s its pe rception, we assessed whether verbal context in the form of labeling
could cause olfactory illusions. In the present scenario, the odorant is the invariant
physical stimulus, the context is the verbal label used to describe the chemical stimulus,
and the illusion is demonstrated by perception being altered as a function of the verbal
label applied. We propose that cases where perception of a specific odor is inverted as
a function of the verbal label it is presented with demonstrate an olfactory illusion.
2 Method
2. 1 Subjects
Eighty undergraduates (forty m ale, forty female; m ean age 21:25 years) from the
University of Pennsylvania served as subjects. Subje cts were i ndividually tes ted and
paid for their participation. Subjects were pres creene d in a teleph one interview prior to
participation, and only nonsmokers without chemical allergies were selected. On th e
days of testing, subjects were asked not to wear any fragrance (other than their usual
soap and shampo o). All subjects had a self-reported normal sense of smell and were
free from respiratory infections when they participated.
2.2 Odorants and verbal labels
The following chemical odorants were used: menthol, patchouli, violet leaf, pine oil,
and a 1 : 1 combination of isovaleric and butyric acids (I ^ B acid). Pine oil, menthol,
and patchouli were prepared as 100% solutions, violet leaf was diluted to a 50% sol-
ution, and I ^ B acid to a 1% solution. The odorless solvent used to prepare the dilu-
tions was diethyl phthalate. All odors were supplied by Haarman n and Reimer Corp.
(300 North Street, Teterboro, NJ 07608, USA), except for isovaleric and butyr ic acids
which were purchased from Aldrich Chemical Company, Inc. (PO Box 2060, Milwaukee,
WI 53201, USA). To produce the olfactory exper ien ce, one diethyl phthalate pellet
saturated with each odorant was prepared and placed into a white opaque plastic jar
and covered with pure cotton. To smell an odorant, subjects unscrewed the lid of the
jar when instructed and sniffed at the cotton inside. There were no visual cues by
which the odo rs ( jars) could be discriminated. Table 1 shows each odor with its two
alternate verbal labels. The alternate verbal labels for each odorant were chosen on
the basis of pretesting with a group of volunteers who were demographically similar to
the study participants. Volunteers sniffed each odorant and then generated as m any
name s as they could for what they thought the odor stimuli might be. The most
Tab le 1. Odor labels and hedonic order by group and session.
Odorant Label, session 1 Label, session 2 Hedonic order
Group 1
I ± B acid parmesan cheese vomit positive, negative
Menthol chest medicine breath mint negative, positive
Patchouli musty basement incense negative, positive
Violet leaf fresh cucumber mildew positive, negative
Pine oil spray disinfectant Christmas tree negative, positive
Group 2
I ± B acid vomit parmesan cheese negative, positive
Menthol breath mint chest medicine positive, negative
Patchouli incense musty basement negative, positive
Violet leaf mildew fresh cucumber positive, negative
Pine oil Christmas tree spray disinfectant positive, negative
Olfactory illusions 383
consistently applied positive and negative labels were then retested with another set of
volunteers to determine validity. The second group of volunteers was given each odor
with alternate verbal labels and asked to assess how much they agreed with the verbal
descriptors. The present alternate labels were then selected from the most agreed-
upon matches for the specific odor in question.
2.3 Design
A randomized block design was followed to produce two label-order groups of twenty
subjects within each gender. The two label-order g roups are referred to as group 1
and group 2. The expe rien ce of subjects w ithin group 1 and group 2 was identical
except that the order of labels provided at session 1 and session 2 was reversed. This
allowed us to examine what effec t the order of labels would have on p erception of a
specific odorant. T he labels given to each odor at each session by grou p are shown in
table 1. Subje cts assessed the same five odors one at a time, at two different session s
separated by one week. Odors were presented in different orders at each session. At
the first session, the odorant was given either a positive or negative label, and at the
second session the labels were reversed. Subjects were not told that they were smelling
the same odors at both sessions. Twenty different random orders of o dor presentation
were prepared and then repeated four times (once for each of m ales and females in
group 1 and group 2). Gender was considered i n the present design because it has b een
shown that under certain conditions women show greater olfactory sensitivity to odors,
as well as greater verbal fluency with odor naming, than do men ( Cain 1982; Doty
et al 1981).
2.4 Procedures
Subjects were told that the purpose of the experi m e nt was a general investigation of odor
pe rception. At the start of session 1, subjects were familiari zed w ith the odor-smelling
and rating procedures, and any questions were addressed. The experiment began with
the experime nter handing the first jar to the subject and saying the designated odor
name for that subject and session (eg ``this is chest me dicine'', for menthol). The subject
then unscrewed the lid of the jar and sniffed at the cotton inside for a few seconds.
Subjects were allowed to sniff the odor as many times as they needed to make their
evaluations. After sniffing the odor, the subjects filled in a questionnaire which asked
for their ratings of the odorant on three 9-point hedonic scales: pleasantness, familiarity,
and intensity (1 extremely unpleasant, unfamiliar, weak; 9 extremely pleasant,
familiar, strong). The subjects were then asked to provide written answers for: (i) what
the odor made them want to do /how they would use it, (ii) whether it evoked a
memory (and if so to describe it bri efly), and (iii) what they would call the odorant.
The order of question items was the same for all subjects at both sessions.
The three written descriptive measures were evaluated together by two independent
judges who were blind to the exp erim ental conditions to determine whether the subjects
thought they were smelling a different odor (or not) at each session. From thi s assess-
ment, the measure of perceptual interpretation was der ived. If the responses to these
questions were generally the sam e at each session, perceptual interpretation was scored
as `same'. If the resp onses to these questions were generally different at each session,
then perceptual i nterpretation was scored as `different'. An example for `same' interpre-
tation would be if, for I ^ B acid, a subject wrote different food-related memories at
each session, and at both sessions said that the odorant would go well with pasta an d
called it ``cheese''. An example for `different' interpretation would be if, for I ^ B acid,
the subje ct at the first session recalled a memory and at the second did not, and at
session 1 said he or she would like to eat, and at session 2 that he or she would like to run
out of the room, and if he or she indicated that the odorant should be called ``par m esan
cheese'' at session 1, and `` vomit'' at session 2. Obv iously the re were many permutations
384 R S Herz, J von Clef
of respon ses possible for each trial. When the judges disagreed on an interpretation,
the subject's re sponses were discussed until consensus was achieved.
After completing a trial for o ne odor, the experimenter handed the next odor to the
subject until assessm ents for the five odors were completed. Subjects were not under any
time constraints to make their evaluations. After rating the five odors, subjects were
dismissed and asked to return in one week for further testing. When subjects returned
a week later they were presented with the s ame five odors in a different order, and
this time told the alternate name for each odor (eg ``this is breath mint'', for menthol).
Subjects completed the same questionnaires (rating scales, written responses) as before
for each odor. At the end of session 2, subjects were fully debriefed and probed to
see if they had guessed the exper imental hypothesis. No subjects corre ctly ascertained
the experimental aims.
3Results
3.1 Perceptual interpretation
This measure assessed whether subjects p e rceived an odor as b eing the sam e or different
at the two sessions as a function of verbal context (label) and was key to our deter mi-
nation of olfactory illusion effe cts. The numbers of subjects who perceived the odors
to be different at each session are shown in table 2.
Because the order of labels given may have influenced perceptual interpretation,
group 1 and group 2 were considered separately (see table 2). As c an be seen, 83% of
subjects in group 1 perceived both I ^ B acid and violet leaf as being different odors at
each session. Patchouli followed closely behind and was perc e ived as a different odor
at each session by 80% of subjects. Perc entage comparison tests (nonparametric test for
significan ce between two p roportions, Statistica
Õ
) showed that there were no differences
between these three groups. Pine oil and menthol were less affected by verbal context.
Percentage comparison tests showed that the number of subjects who perc eived pine
oil and menthol to be different was significantly less ( p 5 0:01) than the number of
subjects who perceived I ^ B acid, violet leaf, and patchouli to be different. In partic-
ular, m e nthol was the l e ast i nfluenced by verbal context and was considered to be a
different odor at the two sessions by only half (50%) of the subjects.
Tab le 2. The number of subjects who perceived the odorants as different as a function of label
and order.
Odorant
Group 1 Group 2
odor label order N percentage odor label order N percentage
I ± B acid parmesan cheese 33 83 vomit 33 83
vomit parmesan cheese
Pine oil spray disinfectant 25 63 Christmas tree 24 60
Christmas tree spray disinfectant
Menthol chest med icine 20 50 breath mint 27 68
breath mint chest medicine
Violet leaf fresh cucumber 33 83 mildew 35 88
mildew fresh cucumber
Patchouli musty basement 32 80 incense 33 83
incense musty basement
N the number of subjects who perceived the odorant as different as a function of label.
Maximum N 40.
Olfactory illusions 385
For subjects in group 2, violet leaf was the odor most often perceived as being a
different odor at each session (88%). I ^ B acid and patchouli tied for being the next
most l i kely to be perceived as different odors (83%). Percentage comparison tes ts
showed that there were no differences between these three odors ( p 4 0:05). P ine oil
was the least likely to be perceived as a different odor at each session (60%), followed
closely by menthol (68%). Percentage comparison tes ts showed that the number of
subjects who perceived pine oil to be a different odor at each session was significantly
less than for all other odors except menthol ( p 5 0:01). Notably, perc entage compar-
ison tests showed that significantly m ore subjects in group 2 perceived me nthol to be
`different' than subjects in group 1 (68% versus 50%; p 0:05).
3.2 Pleasantness evaluations
To assess how each odor was he donically evaluated as a function of the verbal context
it was presented in, analyses of variance (
ANOVA
s) were conducted on the pleasantness
rating-scale data. Four-way mixed design
ANOVA
s with Gender and Order (positive
label first, negative label first) as the between-subjects variables, Odor as the within-
subjects variables, and Lab el (positive, negative) as the repeated measure was performed.
Newman ^ Keuls, p 5 0:01, a poster iori comparisons were used to further insp ect the
data when significant findings were obtained.
A signifi cant Odor by Label interaction was found for odor pleasantness
(F
4 380
23:00, p 5 0:01)
ö
see figure 1. As can be seen, all odors were influenced by
the label manipulation; however, I ^ B acid was most strongly affected by verbal context,
and changed fro m a mean rating of 5.06 when called ``parmesan cheese'' to 1.68 when
called ``vomit''. Menthol was the least affected by verbal context, and changed by only
0.76 rating p oints between being l abeled ``breath mint'' or ``chest medici ne''.
A three-way interaction between Odor, Order, and Label indicated that pleasantness
perception was also variably altered as a functi on of the specific odor and which label
was given first (F
4 380
3:96, p 5 0:01). To investigate this finding further, two-way
mixed-design
ANOVA
s with Order as the between-subjects factors and Label as the
repeated measure, were performed on the pleasantness data obtained for each odor.
Analyses showed significant Order by Label interactions for pine oil (F
176
61:83,
p 5 0:01) and menthol (F
176
3:96, p 5 0:05)
ö
see table 3. A posteriori comparis ons
indicated that pine oil as ``Christmas tree'' was perce ived as sm elling more pleasant and
as ``spray disinfectant'' more unpleasant when the negative label (``spray disinfectant'')
was given first, compared with when ``Christmas tree'' was given as the first label.
For menthol, positive and negative labels elicited equivalent pleasantness ratings,
when the positive label (``breath mint'') was given first. However, if the negative label
(``chest medicine'') was given first, menthol was rated as significantly more pleasant when
called ``breath mint'' (at session 2) than when it was c al led ``chest me di cine'' (at session 1).
,
,
,
,
8
6
4
2
0
Mean pleasantness rating
I ^ B acid
pine oil
menthol
violet leaf
patchouli
Positive label Negative label
Figu re 1. Odor by label interactio n for ratings of pleasantness. Rating scale: 1 extremely
unpleasant, 9 extremely pleasant.
386 R S Herz, J von Clef
3.3 Subsidiary ratings
A significant interaction between Odor and Label was obtained for familiarity ratings
(F
4 380
2:34, p 5 0:05)
ö
see table 4. I ^ B acid was rated as significantly le s s familiar
when given a negative label than when given a positive label. For the other o dors th e re
were no differences in familiarity ratings as a function of label. That is, the difference
in perc eived familiarity a s function of verbal context changed more for I ^ B acid than
any other odor.
A significant Odor by Order interaction was also observed (F
4 380
4:48, p 5 0:01).
For I ^ B acid, pine oil, and menthol whether a positive or negative label came first did not
influence familiarity ratings. However, for violet leaf and patchouli the o rder of the labels
given did matter. For violet leaf, subjects who were given the positive name ``fresh
cucumber'' first and the negative name ``mi ldew'' s econd, rated violet leaf as m ore
familiar overall than subjects who received the names in the reverse order. However, for
patchouli, subje cts who received the positive name ``incense'' first followed by ``musty
basement'' rated patchouli as lower in familiarity than di d subjects who received the
names in the reverse order. Thus, there appears to be something about the prototypicality
of the n ame given to the odor in questio n which influence s odor p e rception (eg familiar-
ity) more so than whether the hedonic valence of the odor name is positive or negative.
,
,
Tab le 3. Pleasantne ss ratings as a func tion of label order and label hedo nics.
Label order Label hedonics Mean SEM
Pine oil
Christmas tree positive: Christmas tree 6:70 0:24
Spray disinfectant negative: spray disinfectant 5:70 0:30
Spray disinfectant positive: Christmas tree 7:38 0:25
Christmas tree negative: spray disinfectant 5:00 0:30
Menthol
Breath mint positive: breath mint 6:63 0:20
Chest medicine negative: chest medicine 6:20 0:23
Chest medicine positive: breath mint 7:18 0:23
Breath mint negative: chest medicine 6:08 0:32
Tab le 4. Familiarity and intensity ratings as a function of odor and label (me an SEM).
Odorant
Familiarity ratings Intensity ratings
positive negative positive negative
label label label label
I ± B acid parmesan cheese vomit parmesan cheese vomit
7:66 0:17 6:75 0:22** 7 : 31 0:15 7:46 0:19
Pine oil Christmas tree spray disinfectant Christmas tree spray disinfectant
7:14 0:21 6:81 0:18 7:22 0:13 7:04 0:14
Menthol breath mint chest medicine breath mint chest medicine
7:91 0:10 8:06 0:11 7:09 0:14 7:24 0:16
Violet leaf fresh cucumber mildew fresh cucumber mildew
5:51 0:24 4:94 0:27 6:00 0:19 6:75 0:17**
Patchouli incense musty basement incense musty basement
5:63 0:25 5:30 0:22 6:29 0:16 6:45 0:18
** denotes p 5 0:01.
Olfactory illusions 387
A significant main effe ct was found for intensity ratings by Order (F
1 380
4:94,
p 5 0:05). Subjects who rece ived a negative label first rated the odo rs as smelling
stronger across both sessions (mean 7:00) than subjects who rec eived a positive label
first (mean 6:76). A significant Odor by Label interaction was also obtained
(F
4 380
3:70, p 5 0:01)
ö
see table 4. A poster iori comparisons indicated that for violet
leaf the difference in ratings between when it was called ``mildew'' and when it was
called ``fresh cucumb er'' was greater than with any other odor.
Finally, a significant Gender by Label interaction was found (F
1380
3:63, p 0:05).
A p osterior i compar isons revealed that when me n were given a positive label they
rated the odors as smelling weaker (m ean 6:60) than subjects in any other group.
Means for females given positive and negative l abels and men given negative labels
were 6.97, 7.02, and 6.96, respectively. These me ans did not differ.
4 Discussion
The present exper iment demonstrated that subjects' perceptual responses to an odor
could be inverted as a function of the verbal label given to it. The claim was made at
the outset of this article that if the influence of verbal context on the perception of certain
odors could induce olfactory misperceptions in accord with our operational definition
of an olfactory illusion (the chemical stimulus re mains invariant but perception of the
stimulus is altered as a function of the verbal context in which it was presented) then
an olfactory illusion would be demonstrated. We propose that evidence of olfactory
illusions was shown for violet leaf, patchouli, and, most significantly, I ^ B acid.
Previous research has shown that co ntext can embellish olfactory perception
(Moskowitz 1979; Zellner and Kautz 1990). Indeed, the perception of an odor can
even b e fabricated by verbal and expe ctation effects. In a classic early demonstration,
Slosson (1899) opened a vial of `odor ' in a cla ssroom and asked students to raise the ir
hand when they detected it. T he vial was completely empty b ut most students raised
their hand. Even more striking, O'Mahoney (1978) showed that informing a television
or radio audience that a certai n sound frequency could produce the perception of
odors was able to generate reports of odor detection and in some cases allergic reac-
tions! More recently, Kn asko et al (1990) found that the m ere suggestion of an ambient
pleasant or unpleasant odor i n an unscented room could produce changes in mood
that were consistent with odor expectation. These findings show how susc eptible olfac-
tory perception is to verbal suggestion. What we have shown in the present exper im ent
is that the perception of an invar i ant physical stimulus (an odorant) can be inverted
as a function of the verbal label that it is presented with. That is, opposite percep tions
to the sam e physical odor stimulus can be produced merely by the verbal context it i s
presented in. We believe that this demonstrates a case for olfactory illusions.
Notably, not all of the odors tested were equally susce ptible to our verbal-c ontext
manipulations. We did not anticipate that all the odors would be equally affected but
we did not know to what extent they would vary and which odors would be most or
least susceptible. An analysis of the specific odors used also helps us to understand
what the mechanisms involved in verbally induced olfactory illusions may be. From
the perceptual interpretation and rating-scale data it is cl ear that I ^ B acid was the
most affected by verbal context (most illusory) and that menthol was the most weakly
influenced, though illusory effects were still obtained in over half of the sample. We
interpret the finding that different odor s were more or less susceptible to the verbal
labeli ng manipulation to be a function of experiential familiarity with the specific odor
and the degree of anchoring the odor had with the verbal labels provided. It seems
that the more equally anchored to the two possible labels the odor was, the mo re likely
it was believed to be either of the two odorants. Thus, I ^ B acid could be experientially
perceived as ``vomit'' or ``parmesan cheese'' equally well, whereas menthol was much
,
,
,
388 R S Herz, J von Clef
more firmly rooted in the ``chest medicine'' than in the ``breath mint'' connotation.
This idea is supported by the comments made by one subje c t, who was a German
exchange student. She remarked that she only knew of the smell of menthol as breath
mints and was unfamiliar with the chest medicine designation. For m ost North A m e r i-
cans, however, experience with menthol in medicine is more common than it is in
candy. A similar cross-cultural variation exists between the US and U K regarding
the p erception of wintergreen (methyl s alicylate). For peop le who have grown up in the
USA, wintergree n is exclusively a pleasant mint candy smell, whereas in the UK this
same odor is associated w ith medicine and tends to elicit highly negative odor ratings
(Cain and Johnson 1978; Mon cr ieff 1966).
The order of labels given to particular odors was also found to be linked to the
degree of verbal-context susceptibility in some c ases. In particular, menthol change d in
terms of h ow much it was perceived to be the same or a different odor across the
two sessions as a function of what name it had been given first. When m enthol was
first pre sented with the label ``chest medicine'', 50% of th e subjects p erceived the odor
to b e different at each se ssion; however, when it was first labeled ``breath mint'', 68%
considered it to be different at each session. Additionally, positive and negative labels
elicited equivalent pleasantness ratings, when the first name was ``breath mint'', but if
the first label applied was ``chest medicine'' the ``breath mint'' label elicited high er
pleasantness ratings. For violet leaf, subjects who were given the positive name ``fresh
cucumber'' first rated violet leaf as more familiar than subjects who received the nega-
tive name ``mildew'' first. However, for patchouli, subjects who rece ived the positive
name ``in cense'' first rated patchouli as being less familiar than subjects who were told
``musty basement'' first.
This illustrates a first-label effect, where the connotation of the f irst label given
influences subsequent respo nding to the same odor. The first-label effect shows that
there is verbal pr iming in odor perception. Notably, perceptual priming by the first
label appears to be linked to the degree of experiential anchoring and thus how
susceptible the o dor was to illusory perception. Perceptio ns of odors with weaker
experiential anchoring (eg menthol) were primed more by the first label than odors
with strong experiential anchoring (eg I ^ B acid). Therefore, both what one is primed
to think of an odor as being as well as one's standard for experience play an important
role in the manipulation of perception for ce rtain o dors.
A subsidiary issue in the present study was to examine the influence of gender on
the responses obtained. Our data showed gender difference s in perceived odor intensity
only, with the finding that men rated odors given a positive lab el as smelling weaker
than subjects in any othe r group. Typically when sex differences are found, they demon-
strate greater sensitivity by females. Howeve r, our findings underscore the potency of
the labeling effe ct, as the results suggest that gender-based expectations rather than
gender-based olfactory sensitivity produced this outcome.
What is m issing in the present study is a condition where the odorants were as sessed
without any verbal c ontext. Subjects could have been asked to try to name and evaluate
the odor s purely on th e basis of the ir p ersonal interpretation. This condition would
be useful to deter mine what the base-rates were for various alternate perceptio ns of
the same odors. However, we did not consider it necessary to in clude this condition
for the following reasons. First, the two alternate verbal labels for each of the odors
were already chosen on the basis of pretesting subjects for what they thought the odor
stimuli could be. Second, it is well established that odors are very difficult to identify
verbally even when they are very familiar (eg Cain 1979; Desor and Beauchamp 1974);
thus in many cases subjects would likely not have be en able to provide themselves with
much useful verbal information regarding the odors in question. As such, a label-free
condition would have tapped into lexical-access issues, the ``tip-of-the-nose phen omenon''
Olfactory illusions 389
(Lawless and Engen 1977), rather than being an indication of the absence of a verbal
context effect. Finally, the aim of this study was to evaluate the effect of verbal context
on odor perception and thus we believed that it was ne cessary to provide subjects
with a verbal context prior to their perception of the actual odorants. Had subjects
smelled the odors without first being provided with verbal labels, they woul d have tried
to generate some kind of label after smelling them and this would have confounded
our goals.
The purpose of this study was to assess whether verbal context could produce
illusory changes in the perception of a set of `ambiguous' odors. Our data showed that
the perception of certain odors could be inverte d at both hedon ic and perc eptual levels
as a function of the verbal context in which the odor was presented. Reber (1985)
defines an illusion as any stimulus situation where that which is perceived cannot be
predicted, prima facie, by a simple analysis of the physical stimulus. This account fits
well with our olfactory case, as one would not be able to predict how I ^ B acid
(for example) would be perceived in the absence of a verbal context. On the basis of
our operational d efi nition of illusion and that of Reber it would seem that we have
provided the first empirical evidence for the existence of olfactory illusions. Notably,
susceptibility to verbal context appears to be influenced by the degree to which pr io r
expe riential factors (prototypicality) bias perc eption. Am ong the odor s tested, I ^ B
acid showed the least bias and thus the greatest illusory effects, whereas menthol
showed the mo st bias and, conversely, the least illusory effects. Future studies should
assess a broad range of different odor s under various verbal, visual, and environmental
conditions to determine more comprehensively what the governi ng principles that
define this phenomenon are.
Acknowledgements . We are indebted to detailed discussions with Trygg Engen and his insightful
commentaries on earlier drafts of this manuscript. We thank George Preti and Henry Lawley at
the Monell Chemic al Senses Center for assis tance with the preparation of the odors used in
this research. We also appreciate the helpful remarks of Gary Beauchamp, William Heindel, and
two anonymous reviewers. This research was supported by a grant from Haarmann and Reimer.
RSH is a Morley R Kare Fellow.
References
Cain W S, 1979 ``To know with the nose: Keys to odor identification'' Science 203 467 ^ 470
Cain W S, 1982 ``Odor identification by m ales and females: predictions vs performance'' Che m-
ical Senses 7 129 ^ 142
Cain W S, Johnson F Jr, 1978 ``Lability of odor pleasantnes s: influence of mere exposure'' Pe rcep-
tion 7 459 ^ 465
Cain W S, Potts B C, 1996 ``Switch and bait: Probing the discrimi native basis of odor identification
via recognition me mory'' Chemical Senses 21 35 ^ 44
Clemens S L, 1882 The Invalid's Story, republished in The Complete Short Stories of Mark Twain
Ed. C Neider (Garden City, NY: International Collectors Library, 1957)
Delk J L, Fillenbaum S, 1965 ``Differences in perc eived color as a function of characteristic
color'' American Journal of Psychology 78 290 ^ 293
Desor J A, Beauchamp G K, 1974 ``The human capacity to transmit olfactory information''
Perception & Psychophysics 16 551 ^ 556
Doty R L, Snyder P, Huggins G, Lowry L D, 1981 ``Endocrine, cardiovascular and psychological
correlates of olfactory sensitivity changes during the human menstrual cycle'' Journal of Com-
parative and Physiological Psychology 95 45 ^ 60
Duncker K, 1939 ``The influence of past experience upon perceptual properties'' American Jour-
nal of Psychology 52 255 ^ 265
Engen T, 1987 ``Remembering odors and their names'' American Scientist 75 497 ^ 503
Gregory R L, 1997 Eye and Brain: The Psychology of Seeing 5th edition (New York: Oxford
University Press)
Herz R S, 20 00 ``Verbal coding in olfactory versus non-olfactory cognition'' Memory & Cognition
28 957 ^ 964
390 R S Herz, J von Clef
Knasko S C, Gilbert A N, Sabini J, 1990 ``Emotional state, physical well-being and performance
in the presence of feigned ambient odor'' Journal of Applied Soc ial Psychology 20
1345 ^ 1357
Ku
«
nnapas T M, 1955 ``Influence of frame size on apparent length of a line'' Journal of Experimental
Psychology 50 168 ^ 1 70
Lawless H, Engen T, 1977 ``Associations to odors: Interference, mnemonics, and verbal labelling''
Jou rnal of Experimental Psychology 3 52 ^ 59
Luo C R, Wang S, 1997 ``Effects of figure context on the apparent length of a lin e'' Perceptual
and Motor Skills 85 551 ^ 558
Moncrieff R W, 1966 Odour Preferences (New York: John Wiley)
Moskowitz H R, 1979 ``Mind, body and pleasure: An analysis of factors which influence sensory
hedonics''', in Preference Behaviour and Chemoreception Eds J H A Kroeze (London: Informa-
tion Retrieval Ltd) pp 131 ^ 144
O'Mahoney M, 1978 ``Smell illusio ns and suggestions: reports of smells contingent on tones
played on television and ratio'' Chemical Sen ses and Flavor 3 183 ^ 189
Reber A S, 1985 Dictionary of Psychology (Harmondsworth, Middx: Penguin Books)
Rechtschaffen A, Mednich S A, 1955 ``The autokinetic word technique'' Journal of Abnormal
and Social Psychology 51 346
Rozin P, Fallon A E, 1987 ``A perspective on disgust'' Psychological Bulletin 94 23^41
Schi ffman H R, Thomp son J G, 1978 ``The role of apparent depth and context in the perception
of the Ponzo illusion'' Perception 7 47 ^ 50
Sherif M, 1935 ``A study of some social factors in perception'' Archives of Psychology 187
Sherif M, 1937 ``An experimental approach to the study of attitudes'' Sociometry 1 90^98
Slosson E E, 1899 ``A lecture experiment in hallucinations'' Psychological Review 6 407^408
Spivey-Knowlton M, Bridgeman B, 1993 ``Spatial context affec ts the Poggendorff illusion''
Perception & Psychophysics 53 467 ^ 474
Zellner D A, Kautz M A, 1990 ``Color affects perceived odor intensity'' Journal of Experimental
Psychology: Human Perception and Performance 16 391 ^ 397
Olfactory illusions 391
ß 2001 a Pion publication printed in Great Britain
