Rapid mood change and human odors

Physiology & Behavior 68 (1999) 241–250
Brief Communication
Rapid mood change and human odors
Denise Chen* and Jeannette Haviland-Jones
Rutgers University
Received 29 May 1998; received in revised form 22 March 1999; accepted 7 July 1999
Abstract
We demonstrate an immediate effect of airborne chemicals on human moods. We collected six groups of underarm odors, respectively, from five prepubertal girls, five prepubertal boys, five college women, five college men, five older women, and five older men. In
addition, we collected odors from homes of these donors, making them a seventh group of odor. Three hundred and eight odor observers
ranked the seven groups of odors of little girls, little boys, college women, college men, older women, older men, and homes by their perceived pleasantness, intensity, perceived masculinity, and age of the donors, among other qualities. On a separate task, the same odor observers assessed their depressive, hostile, and positive moods twice, once before and once a few minutes after they sniffed one of the
above seven groups of odors. Exposure to underarm odors for under 2 min led to significant, rapid, and small changes in the nonclinical
depressive mood of the odor observers. The mood changes were independent of the observers’ perceptions of odor qualities. Odors perceived as unpleasant and intense were as likely to relieve a depressive mood as were pleasant odors. © 1999 Elsevier Science Inc. All
rights reserved.
Keywords: Airborne chemical communication; Mood; Odor judgment; Humans
1. Introduction
Animals are known to produce airborne chemicals that
elicit long-term physiological and endocrinological effects
(e.g., estrous synchronization) [1,2] and immediate behavioral responses (e.g., approach, avoidance) [1] including
mood changes (fear) [3,4,]. Human airborne chemicals also
elicit long-term endocrinological effects (e.g., menstrual
synchronization) [5]. Rudimentary olfactory identification
(e.g., kin from nonkin, self from others, familiar from unfamiliar individuals) has been demonstrated in humans [6–9].
What has not been studied is whether mere exposure to such
odors has a short-term effect on the behavior of the odor recipients, and whether the effect is contingent upon, or independent of, the recipients’ perceptions of the odor qualities
and their inferences about the source of the odors.
Research shows that synthetic odors have an effect on mood
and memories [10–13] such that pleasant scents tend to have a
positive effect, whereas unpleasant chemicals tend to have a
negative effect. According to this hedonic congruency model, it
seems likely that natural body odors perceived to be pleasant
should affect people positively, and that body odors perceived
to be unpleasant should affect people negatively. On the other
* Corresponding author. Monell Chemical Senses Center, 3500 Market
Street, Philadelphia, PA 19104.
E-mail address: [email protected]
hand, given the evidence that odors can affect people without
their conscious awareness [5,14,15], it is also possible that the
emotional impact of natural body odors may be independent of
how the odors are consciously perceived.
Animal [1] and dermatological literature [6,9] suggests
that odors across different developmental stages vary in biological properties, making them possibly identifiable by age
and gender. Research on human olfactory identification of
gender, however, is inconclusive. The majority of the studies examined only the olfactory identification of the gender
of young adults, and relied on the direct-questioning method
of asking people to pick the right odor. Six out of the nine
studies found evidence supporting human identification of
the gender of young adults based on hand [16], breath [17],
and T-shirt [8,18] odors while the three studies based on underarm odors [19] did not. One interpretation of these results is that what is identified is odor intensity/pleasantness,
and not gender per se [6,19]. It remains unclear whether discrimination is possible in young adults and in other age
groups such as prepubertal children and older adults. A recently published study showed that neither adults nor children themselves identified gender of 9-year-old children
[20]. However, hygiene and soap usage in donor children
was not controlled, which, as an adult study has suggested
[21], could have hampered gender identification. Further,
the direct questioning methods may have tapped only the
more conscious level of discrimination.
0031-9384/99/$–see front matter © 1999 Elsevier Science Inc. All rights reserved.
PII: S0031-9384(99)00 1 4 7 - X
242
D. Chen and J. Haviland-Jones / Physiology & Behavior 68 (1999) 241–250
The purpose of this study is to investigate: (1) whether
underarm odors from different age and gender groups have
a differential impact on the mood of young adult men and
women observers who smell the odors; and (2) whether the
emotional responses to odors are contingent upon the observers’ perceptions of odor qualities (e.g., pleasantness, intensity) and inferences about odor source (e.g., donor age
and gender).
2. Materials and methods
2.1. Participants and design
2.1.1. Odor donors
Underarm odors were collected from 30 donors of six
age-by-gender groups: (1) five little girls (mean 6 SD 5
5 6 1.48 years, specific age 5 3, 4, 5, 5, 7); (2) five little
boys (mean 6 SD 5 6 6 1.22 years, specific age 5 5, 5, 6,
6, 8); (3) five college women (mean 6 SD 5 20 6 1.14
years, specific age 5 18, 19, 20, 20, 21); (4) five college
men (mean 6 SD 5 23 6 3.28 years, specific age 5 19,
21, 21, 25, 27); (5) five older women (mean 6 SD 5 71 6
5.28 years, specific age 5 64, 68, 73, 76, 76); and (6) five
older men (mean 6 SD 5 73 6 8.44 years, specific age 5
58, 74, 76, 78, 78). All were Caucasian, healthy, not on hormone treatment, and nonsmoking. The college women were
not pregnant and reported regular menstrual cycles (i.e., between 27 to 33 days per cycle). All donors were recruited
through ads from the New Brunswick, NJ area. The children
were recruited from day care centers, young adults from a
state university, and older adults from faculty/staff from the
same university, and from a senior activity center. All received $10 and three T-shirts for their participation.
2.1.2. Odor observers
The odor observers were 316 undergraduate students. All
received course credit for their participation. They were
screened for gross olfactory deficit at the end of the experiment. Eight out of the original 316 observers (2%, five
women, three men) failed to meet the criteria and were excluded from subsequent analysis. The tests and criteria will
be described later under Olfactory Screening Tests. The final odor observers consisted of 308 undergraduate students
(154 men, 154 women) ages 17 to 29 (mean 6 SD 5 19 6
1.64 years). Fifty-two percent described themselves as Caucasian, 24% Asian, 9% Hispanic, 8% African-American,
and 7% other. The majority of them (97%) were nonsmokers. None of the women reported that they were pregnant.
2.1.3. Design
The experiment was a 7 (odor sources: little girls, little
boys, college women, college men, older women, older
men, home odors) 3 2 (observer gender) mixed-model design, with 22 observers in each cell. The percentage of
White, African-American, Hispanic, Asian, and other odor
observers were equally distributed across odor conditions
(p 5 0.691).
2.2. Odor stimuli collection
Adult donors were instructed not to use any fragrance/
antiperspirant nor to shave their underarms starting from 4
days prior to odor collection. Some used the fragrance-free
deodorant provided by the experimenter up to the day prior
to odor collection. Both child and adult donors were instructed not to eat any odorous food such as garlic and onion both on the day before and on the days of odor collection. A diet diary was kept during odor collections to check
if dietary instructions were followed. Any strenuous activities by the donors on the days of odor collection were reported.
Both child and adult donors took a bath/shower (or at
least washed their underarm regions) on the evening before
each odor collection with the fragrance-free soap and shampoo provided by the experimenter, taped a clean 4 3 40
gauze pad around each armpit the next morning, and kept
them on for 8 to 10 h. This was repeated for 3 days. Following each day of collection, donors stored the pads in double
plastic bags in their freezer. Each day of odor collection, donors wore a new T-shirt provided by the experimenter to
avoid odor contamination from laundry detergent. Donors
were instructed to leave open blank pads at home as controls for environmental influences. On average, pads were
collected by the experimenter the same or the second day
after the collection was over, separated by donor age, gender, and odor type, cut into 2 3 20 squares, placed in enclosed glass jars, and frozen at 280 8C until used in testing.
Each donor contributed underarm sweat collected on a total
of 24 pads (two arms 3 3 days 3 four squares per pad).
Subsequent analyses revealed that donor diet, activity
level, and the length of time the pads had been worn did not
vary by odor conditions. The seven donors who reported to
have consumed garlic- or onion-flavored food on the days
of odor collection were evenly distributed in each age by
gender group (i.e., one little girl, one college man, one older
man, two college women, and two older women). Overall,
the majority of children and older adult donors reported to
be physically active such as engaging in play (children), or
gardening or exercising in the gym (older adults). More
young adults reported to engage in school-related work such
as attending classes and reading. A repeated-measures
ANOVA with the number of hours worn on Day 1, Day 2,
and Day 3 of odor collection as a within-subjects factor, and
with the seven odor conditions as a between-subject factor,
found no significant difference in the number of hours the
pads were worn on each day, F(2, 23) 5 0.794, p . 0.05,
nor in the number of hours the pads were worn for each
odor condition, F(10, 48) 5 1.448, p . 0.05.
2.3. Odor stimuli preparation
Odor stimuli were defrosted to room temperature at least
40 min before testing. The seven groups of odor stimuli (six
groups of underarm odors from little girls, little boys, college women, college men, older women, and older men, re-
D. Chen and J. Haviland-Jones / Physiology & Behavior 68 (1999) 241–250
spectively, and one group of odors combined from the
homes of each donor) were each covered up in a 150 3 20
mm glass Petri dish. Glass Petri dishes were chosen over
plastic bottles because the latter gave off a plastic odor.
They were chosen over glass jars because Petri dishes’
short, flat, and wide surfaces allowed the pads to be evenly
spread out and allowed easier sniffing. Each dish of underarm odors contained five 2 3 20 pads representing each individual within that donor group. The home odor dish contained six pads, one randomly drawn from each donor
group. The extra pad was hidden underneath the other five
pads, so that the home odor dish did not appear different
from the underarm odor dishes.
Each set of five groups of pads was reused for an average
of 7 days (mean 6 SD 5 6.62 6 2.50), and was tested on
an average of 13 observers (mean 6 SD 5 13.12 6 3.94).
The old pads were then discarded and new pads were used.
The pads were changed a total of eight times (or cycles).
Three sets of seven groups of odors were used at a cycle
during the first seven cycles, and two sets of seven groups
of odors were used during the last cycle. After being transferred into enclosed glass jars, the pads were refrozen in between the test days, as well as between sessions on a single
test day if the two sessions were separated by more than 3 h.
2.4. Testing condition
The experiment was described as a study of people’s reactions to various odors, and did not specify the source/type
of any odors. The odor observers were instructed not to
wear any fragrances or strongly scented deodorants to the
experiment. The experiment was administered by a single
experimenter who also did not use any of the above fragrances. On average, observers were either tested individually or two at a time in a room of approximately 6 meters
long and 3 meters wide. The room had seven windows; all
were opened in between the test sessions. When more than
one subject was tested at a time, they were seated apart from
one another, to ensure that they were not able to smell the
target odor assigned to each other.
The experiment consisted of three assessments, in their
chronological order: (1) olfactory impact on mood, (2) olfactory impact on memory recollections (not reported here),
and (3) odor judgment.
Observers smelled only one group of odors on the olfactory-impact-on-mood task but smelled all seven groups of
odors on the odor-judgment task. For the purpose of clarity,
the odor-judgment task will be presented before the olfactory-impact-on-mood task.
2.5. Olfactory screening tests
Two screening tests were given. One was a three-item
scratch-and-sniff test (Sensonics, Inc.) in which subjects
were asked to identify the correct name of a scent out of
four available choices for a total of three scents that are familiar to most people who are raised in this country (i.e.,
243
grass, root beer, coconut). The other was an androstenone
identification test. Androstenone is an odorous steroid believed to contribute to underarm odors [22,23]. Subjects
were presented with three bottles, and asked to identify over
two trials the only one bottle that contained an odor (5 mg
of 99% pure 5a-androst-16-en-3-one purchased from
Sigma). The order of the tests was random across subjects.
Two different tests were used because a previous study [24]
suggested that andronstenone sensitivity might not be directly related to overall olfactory sensitivity. Only subjects
who made no more than one mistake on the scratch-andsniff test regardless of their performance on the androstenone test, or subjects who correctly identified androstenone
on both trials regardless of their performance on the scratchand-sniff test, were included in the final analysis. Five
women (one each from the little-girl, little-boy, collegewomen, college-men, and older-women odor condition) and
three men (one each from the little-girl, college-women, and
older-men odor condition) did not meet such criterion and were
excluded. The criterion was arbitrary, and aimed at excluding individuals who were sensitive to neither common/familiar odors nor one of the odorous ingredients of sweat.
Out of the final 308 observers who passed the screening
tests, one-third of women (34% or 52 women) and over onefourth of men (27% or 42 men) performed correctly on both
the scratch-and-sniff and the androstenone test. Sixty percent of women (92 women) and close to 50% of men (48%
or 75 men) detected androstenone, compared with 56% of
each gender (86 females, 87 males) who correctly identified
the names of all three scents on the scratch-and-sniff test.
2.6. Experimental procedure
2.6.1. Odor judgment
Observers had all seven odors presented to them. Each
odor was ranked independently on perceived pleasantness,
intensity, and familiarity, and on perceived erotic quality/
allure, dependency/neediness for care and protection, and
perceived age and gender of the donor. The odors were resorted before each ranking. There was a minimum of a 1min interval in between each ranking to minimize the effects of odor adaptation. The order of the evaluations was
random except for questions on age and gender, which always appeared last.
2.6.2. Olfactory impact on mood
The Differential Emotion Scale IV (DES) [25] consists
of 36 questions that assess how frequently observers experienced 12 emotions. Each question is rated on a scale from 1
(rarely or never) to 5 (very often). Observers’ mood was
twice assessed on two different forms of the DES, once at
the beginning of the experiment, and once about 2 min later,
after they completed a brief demographic information form
and sniffed one target odor. All observers sniffed only one
type of odor. Neither observers nor the experimenter knew
which odor was presented. The experimenter randomly selected one dish of odors from the seven dishes of odors. Be-
244
D. Chen and J. Haviland-Jones / Physiology & Behavior 68 (1999) 241–250
cause the identification number was written on the bottom
of the dish, the experimenter did not know which odor it
was. The identification was only uncovered and recorded
before the odor-judgment task. At the end of the day, the experimenter recorded the number of subjects in each odor
condition. If one condition was filled, subjects were assigned odors from the remaining conditions. With the exception of the last few subjects, the experimenter did not
know which condition the subjects were assigned to.
The perceived intensity of the target odor was defined as
the ranked intensity of that odor relative to the six other
odors. Other perceived qualities of the target odor were similarly defined, based on how the target odor was ranked relative to the other odors on the odor judgment task.
When presented with the target odor, observers were instructed to first sniff all five pads in the dish one by one to
get an idea of what each smelled like. They were then instructed to take three quick sniffs followed by three long
sniffs of all the pads at once. This was done to ensure that
all subjects would smell the stimuli in a similar fashion, and
that each individual odor within the donor group was
smelled.
2.7. Statistical analyses
2.7.1. Odor judgment
To determine observers’ perceptions of odor qualities,
repeated-measures ANOVAs were conducted with rank
within each judgment category (for perceived pleasantness,
intensity, perceived masculinity and age of the odor donors,
respectively) as a within-subjects factor, and with observer
gender as a between-subjects factor.
To investigate whether observers on the odor-judgment
task might rank the target odor that they had been exposed
to earlier on the mood task to be less intense than observers
who had not been exposed to it, seven ANOVAs were performed, with the rank of the perceived intensity for each target odor group relative to the other odors as a dependent
variable, and with odor condition (observers in that target
odor condition versus those who were not) as an independent variable. Results indicate that observers in a particular
target odor condition did not rate the intensity of the target
odor any differently than those who were not in the condition except that observers who had been exposed to the
odors of college men on the mood task rated those odors to
be less intense than did observers who had not been exposed
to such odors (mean 5 5.77 versus 6.40, F(1, 304) 5 11.26,
p , 0.005). Consequently, ranked intensity for odors of college men by observers in that odor condition was adjusted
such that any rank less than 4 was increased by 4 (e.g., on a
scale where the maximum intensity ranking is 7, an intensity ranking of 1 of college men’s odors by a subject in the
college men’s odor condition was adjusted to 5). This
brought the mean intensity rank for odors of college men by
observers in that condition equivalent to the same as those
not in that condition (mean 5 6.30 versus 6.40).
2.7.2. Olfactory impact on mood
Mood scores before and after the odor exposure were
separately analyzed by Principal Components Analysis
(PCA) with varimax rotation. Principal Components Analysis, similar to factor analysis, is a data reduction method
widely used in the social sciences. PCA successively extracts a small number of components that account for a large
amount of variance while varimax rotation makes the components easier to interpret [26]. Three mood factors were revealed based on the scree plot for both preodor mood and
postodor mood scores: depressive mood (like other measures of depression, includes sadness, inward hostility, disgust, and anxiety associated with fear, shame, shy, and
guilt), hostile mood (anger and contempt), and positive
mood (interest, joy, and surprise). Factor loadings are presented in Table 1. Questions 4 and 31 had low loadings on
both mood tests. Question 31 also loaded on different factors on the second mood test. Consequently, both were excluded from subsequent analyses. The three-factor mood
structure in the present study was consistent with the threefactor structure found by Stapley and Haviland [27]. No ethnicity difference was found in mood changes or in odor
judgment; therefore, subsequent analyses combined observers across different ethic backgrounds.
To investigate the relative predictive value of odors on
mood above and beyond observers’ preodor mood state, six
hierarchical multiple regression analyses were conducted,
two for each mood factor. The following variables were entered as predictors in step 1: preodor mood, perceived intensity of the target odor, odor conditions, and observer gender.
Odor conditions either appeared as individual odors
(dummy coded into six variables), or were grouped by donor age (dummy coded into two variables) and donor gender. Two-way interactions between observer gender and
odor conditions were entered as predictors in step 2. Any
three-way interactions were entered in step 3. As one of the
purposes of the study was to investigate whether mood
scores changed as a function of the target odor, the dependent variable was changes in mood (defined as a preodor
mood factor subtracted from its postodor mood factor).
The seven perceived odor qualities (perceived intensity,
erotic quality, pleasantness, familiarity, donor dependency,
donor age, and gender) were not entered simultaneously in
the same regression because they were significantly correlated with one another. Instead, separate regressions were
conducted as above, each controlling for one of the six remaining odor qualities. The same hierarchical multiple regression analyses were also conducted with the adjusted
perceived intensity as one of the predictors.
The ability to detect androsterone was not entered as a predictor in the regression because it was not significantly correlated with observed mood changes (point-biserial correlation
5 20.029, 20.069, 0.011 with depressive, hostile, and positive mood changes, respectively, p . 0.05), nor was it correlated with observers’ accuracy at identifying the age and gender of the odors (point-biserial correlation 5 0.067, p . 0.05).
D. Chen and J. Haviland-Jones / Physiology & Behavior 68 (1999) 241–250
245
Table 1
Item content and factor loadings (Varimax rotation) of the pre-odor and post-odor mood scores
Item content on
Izard’s original scale
Preodor Components
23. Feel bashful, embarrassed
28. Feel sick about yourself
35. Feel afraid
14. Feel mad at yourself
21. Feel sad and gloomy, almost like crying
5. Feel you can’t stand yourself
22. Feel like you did something wrong
34. Feel discouraged, like you can’t make it, nothing’s going right
7. Feel unhappy, blue, downhearted
10. Feel shy, like you want to hide
19. Feel fearful, like you are in danger, very tense
26. Feel like people laugh at you
6. Feel embarrassed when anybody sees you make a mistake
2. Feel sheepish, like you don’t want to be seen
1. Feel regret, sorry about something you did
12. Feel scared, uneasy, like something might harm you
27. Feel like things are so rotten they could make you sick
30. Feel like you ought to be blamed for something
36. Feel like people always look at you when anything goes wrong
24. Feel disgusted, like something is sickening
31. Feel the way you do when something unexpected happens
9. Feel like somebody is a low-life, not worth the time of day
16. Feel like somebody is “good for nothing”
33. Feel angry, irritated, annoyed with somebody
13. Feel mad at somebody
29. Feel like you are better than somebody
20. Feel like screaming at somebody or banging on something
4. Feel like somthing stinks, puts a bad taste in your mouth
15. Feel happy
3. Feel glad about something
25. Feel joyful, like everything is going your way, everything is rosy
18. Feel amazed, like you can’t believe what’s happened, it was so unusual
17. Feel so interested in what you’re doing that you’re caught up in it
32. Feel alert, curious, kind of excited about something
11. Feel like what you’re doing or watching is interesting
8. Feel surprised, like when something suddenly happens you had no idea would happen
0.680
0.667
0.648
0.645
0.645
0.637
0.624
0.612
0.603
0.596
0.591
0.585
0.584
0.583
0.572
0.562
0.559
0.520
0.508
0.503
0.397
Depressive Hostile
0.421
Postodor Components
Positive Depressive Hostile
0.490
0.332
0.778
0.774
0.668
0.641
0.544
0.505
0.480
0.348
0.793
0.677
0.718
0.712
0.678
0.740
0.749
0.725
0.626
0.757
0.636
0.681
0.679
0.748
0.622
0.637
0.611
0.635
0.608
0.637
0.427
0.434
0.705
0.667
0.646
0.575
0.568
0.555
0.542
0.516
Positive
0.488
0.830
0.838
0.727
0.724
0.591
0.644
0.440
0.623
0.658
0.613
0.691
0.647
0.667
0.673
0.650
Principal Components Analysis is used as the extraction method. Loadings on both components are displayed if the same item loads on two components
and their loading difference is less than 0.19.
Corrections for inclusion of multiple tests were taken to
protect the alpha levels. For example, to ensure that the significance level for the variable of odor condition was 0.05,
the significance level of each of the six dummy variables to
represent the seven categories of odor condition had to be
less than 0.0083 to be considered significant. To ensure that
the significance level for the variable of donor age was 0.05,
the significance level of each of its two dummy variable had
to be less than 0.025 to be considered significant.
The effect size d is reported. It is calculated at t times the
square root of the sum of one over the first sample size and one
over the second sample size. According to Cohen, d 5 0.20 indicates a small effect size, d $ 0.50 indicates a medium effect
size, and d $ 0.80 indicates a large effect size [28].
measures ANOVAs were performed for perceived intensity
and perceived pleasantness respectively with the rank of the
seven groups of odors as a within-subjects factor and with
the cycle of the pads as a between-subjects factor. Although
there was a significant rank by cycle interaction, F(39,
1689) 5 2.254 and F(38, 1649) 5 2.438, p , 0.0001, for
pleasantness and intensity, respectively. Scheffe post hoc
analyses showed that none of the pairwise comparison
reached significance after the significance level for each
pair was adjusted. There were a total of 56 (seven odors 3
eight cycles) pairs of comparisons. The significance level
for each pair needed to be less than 0.0009 in order to
achieve an overall significance level of 0.05. Thus, we conclude that the perceived odor pleasantness and intensity respectively did not differ by cycle.
3. Results
3.1. Odor judgment
To examine whether pads from differ cycles varied in
their perceived pleasantness and intensity, two repeated-
Odor observers ranked the seven odors on qualities such
as perceived pleasantness, intensity, and perceived gender
246
D. Chen and J. Haviland-Jones / Physiology & Behavior 68 (1999) 241–250
and age of the donors (Figs. 1 and 2). Home odors and odors
of children were perceived to be most pleasant, least intense, most feminine, and youngest in age among the seven
target odors. The odors of college men followed closely by
odors of older women were perceived as most unpleasant,
intense, most masculine, and oldest in age. Odors of college
women and older men were in the middle. The finding that
odors of older women were judged more similar to odors of
college men than to odors of college women and that odors
of older men were judged more similar to odors of college
women than to odors of college men could be due to the
change in the hormonal profile in older women and men.
The estrogen/testosterone ratio has been found to decrease
in postmenopausal women [29,30] and increase in older
men [30–32]. Although apocrine gland activities are believed to stop in old age [9], the strong underarm odors found
in older women in this study seem to indicate otherwise.
There have been claims that people can discriminate gender from the hand, breath, and T-shirt odors of young adult
men and women [8,16–18], but such results may have been
biased by age restrictions in the samples. When we added an
older adult sample, we found neither gender nor age identification was consistently accurate. Our results support Doty
and colleagues’ findings [6,19] that such discrimination
might have been based on observers’ perceptions of odor intensity and pleasantness. The more intense and unpleasant
an odor, the older and more masculine the donors were
judged to be. This stereotypic rule led observers to misidentify little boys as little girls, and to misidentify both college
men and older women as old men. Older men, whose odor
was judged moderate on most scales, might be classified as
any age or gender. Odors of college women were identified
as female, but their low intensity caused them to be easily
confused with children’s odors.
Fig. 1. Average rank of perceived odor pleasantness, intensity, and perceived masculinity and age of the donors by odor conditions. Vertical bars
represent standard errors about the means. This is based on repeated measures ANOVAs, F(6, 301) 5 125.14, 266.24, 77.17, and 46.17, respectively, p , 0.0001.
Fig. 2. Average rank of perceived erotic quality, intensity, familiarity, and
perceived dependency of the donors by odor conditions. Vertical bars represent standard errors about the means. This is based on repeated measures
ANOVAs, F(6, 301) 5 88.82, 266.24, 14.84, and 9.70, respectively, p ,
0.0001.
3.2. Olfactory impact on mood
The regression results for depressive mood are presented
in Tables 2 and 3, and mean changes in depression by odor
conditions are presented in Fig. 3. Based on research on the
effect of synthetic odors on mood and behavior [10–13], we
had predicted that mild and pleasant odors, such as the underarm odors of children, would have an uplifting effect on
the mood of the observers, whereas intense and unpleasant
odors, such as the underarm odors from men, would increase the hostile and depressive moods of the observers.
However, we actually found that even after preodor depressive mood and perceived odor intensity had been controlled,
observers still reported lower ratings of depression when
presented with odors of older adults in contrast to younger
adults, and odors of women in contrast to men (b 5 20.166
and 20.144, p , 0.02, d 5 0.36 and 0.29, respectively,
small effect size). Identical results were obtained after the
perceived intensity had been adjusted for habituation (b 5
20.167 and 20.145, p , 0.02, d 5 0.36 and 0.29).
In particular, observers reported significantly lower ratings of depression when presented with odors of older
women than when presented with odors of college men (b 5
20.287, p , 0.0001, d 5 0.82, large effect size). Again,
identical results were obtained after the perceived intensity
had been adjusted for habituation (b 5 20.290, p ,
0.0001, d 5 0.81).
ANCOVAs for each mood factor with the preodor mood
and perceived odor intensity as covariates, with odor conditions and observer gender as independent variables, and
with the mood change as the dependent variable, yielded the
same result as the regression analyses.
Regression analyses respectively controlling for perceived odor pleasantness, erotic quality, familiarity, and donor dependency, age, and gender yielded results that are
D. Chen and J. Haviland-Jones / Physiology & Behavior 68 (1999) 241–250
247
Table 2
Hierarchial multiple regressiona of preodor depressive mood, perceived odor intensity,b observer gender, and odor conditions (for individual odors) on
changes in depressive moodc
Variable
Step 1
Preodor depressive mood
Perceived odor intensity
Observer gender (w 5 1, m 5 0)
Odor 1 (girlse 5 1, rest 5 0, cm 5 reference)f
Odor 2 (boys 5 1, rest 5 0, cm 5 reference)
Odor 3 (cw 5 1, rest 5 0, cm 5 reference)
Odor 4 (ow 5 1, rest 5 0, cm 5 reference)
Odor 5 (om 5 1, rest 5 0, cm 5 reference)
Odor 6 (home 5 1, rest 5 0, cm 5 reference)
Step 2
Observer gender 3 odor 1
Observer gender 3 odor 2
Observer gender 3 odor 3
Observer gender 3 odor 4
Observer gender 3 odor 5
Observer gender 3 odor 6
R2
DR2
0.086
0.086
3.034**
0.112
0.011
0.139*
20.136
20.136
20.136
20.287***
20.108
20.202*
0.095
0.009
0.475
0.026
20.069
0.051
20.016
0.028
20.090
Incremental F
Betad
R 5 0.308, R2 5 0.095, F(15, 285) 5 1.991, p , 0.02, for the total equation.
Perceived odoar intensity 5 Ranked intensity of the target odor relative to the six other odors. The ranks were obtained from the odor judgment task.
c
Changes in depressive mood 5 postodor depressive mood 2 preodor depressive mood.
d
Beta weights are for the simultaneous entry of variables. Beta is the standardized multiple regression coefficient that allows comparisons of the relative
effects of the predictor variables. It indicates the predicted change (in standard deviation units) in the dependent variable of a mood difference for a standard
deviation change in a particular predictor, when all other predictor variables are held constant [39].
e
Girls 5 little girls, boys 5 little boys, cw 5 college women, cm 5 college men, ow 5 older women, om 5 older men, home 5 home odors, w 5 women,
m 5 men.
f
Each level of odor condition and each level of gender by odor condition is counted as significant only if p , 0.0083.
*p , 0.05, **p , 0.005, ***p , 0.0001.
a
b
Table 3
Hierarchial multiple regressiona of preodor depressive mood, perceived odor intensityb, observer gender, and odor conditions (by donor age and donor gender)
on changes in depressive moodc
Variable
R2
DR2
Incremental F
Step 1
Preodor depressive mood
Perceived odor intensity
Observer gender (w 5 1, m 5 0)
Donor gender (w 5 1, m 5 0)
Donor age 1 (ce 5 1, rest 5 0, ya 5 reference)f
Donor age 2 (oa 5 1, rest 5 0, ya 5 reference)
Step 2
Observer gender 3 donor gender
Observer gender 3 donor age 1
Observer gender 3 donor age 2
Donor gender 3 donor age 1
Donor gender 3 donor age 2
Step 3
Observer gender 3 donor gender 3 donor age 1
Observer gender 3 donor gender 3 donor age 2
0.081
0.081
3.701***
Betad
0.131*
20.001
0.160***
20.144**
20.090
20.166**
0.094
0.013
0.711
0.046
20.074
20.038
0.133
20.051
0.097
0.003
0.362
0.073
20.066
R 5 0.311, R2 5 0.097, F(13, 245) 5 2.018, p , 0.03, for the total equation.
Perceived odor intensity 5 ranked intensity of the target odor relative to the six other odors. The ranks were obtained from the odor judgment task.
c
Changes in depressive mood 5 postodor depressive mood 2 preodor depressive mood.
d
Beta weights are for the simultaneous entry of variables. Beta is the standardized multiple regression coefficient that allows comparisons of the relative
effects of the predictor variables. It indicates the predicted change (in standard deviation units) in the dependent variable of a mood difference for a standard
deviation change in a particular predictor, when all other predictor variables are held constant [39].
e
C 5 children, ya 5 young adults, oa 5 older adults, w 5 women, m 5 men.
f
Donor age 1 and donor age 2 are counted as significant only if p , 0.025 for each.
*p , 0.05, ** p , 0.02, ***p , 0.01.
a
b
248
D. Chen and J. Haviland-Jones / Physiology & Behavior 68 (1999) 241–250
Fig. 3. Changes in depressive mood by odor conditions. Vertical lines represent standard errors about the mean. The part below zero indicates the
amount of reduction in depression after the observers smelled the target
odor. This is based on an ANCOVA with changes in depressive mood as
the dependent variable, observer gender, and odor condition as independent variables, and with preodor depressive mood and perceived intensity
of the target odor as covariates, F(6, 285) 5 2.771, p , 0.02.
identical to the regression analyses reported above that controlled for perceived odor intensity. Namely, even after preodor depressive mood and perceived odor qualities (e.g.,
pleasantness, erotic quality, familiarity, dependency, donor
age, donor gender) had been controlled, observers still reported significantly lower ratings of depression when presented with odors of older adults in contrast to younger
adults (b 5 20.166 to 20.17, p , 0.016 to 0.018), and
odors of women in contrast to men (b 5 20.144 to 20.154,
p , 0.013 to 0.019.
Rapid exposure (i.e., around 2 min) to odors also led to a
nonsignificant change in the positive mood consistent with
that found in the depressive mood. That is, when depression
was significantly lightened, positive moods were nonsignificantly elevated. There was no change in the hostile moods.
Mean changes in depressive, hostile, and positive mood, are
presented in Table 4. Although women reported higher ratings of depression than did men (b 5 0.139, p , 0.02, d 5
0.28, small effect size and identical results were obtained
after perceived intensity had been adjusted for habituation),
there was no interaction.
4. Discussion
To summarize, our data demonstrate that airborne chemicals produced by humans can modulate the moods of other
humans independent of perceived odor intensity and pleasantness, and independent of the attributions of the donors’
age and gender. In particular, exposure to underarm odors
of older women, women, and older adults, led to a greater
reduction in depressive mood than exposure to underarm
odors of young men, men, and young adults. This could indicate a subtle effect of airborne chemicals on human mood.
Even though the effect of airborne chemicals on mood is
very small, though significant, we believe that the rapid
mood change is impressive, given that observers smelled
the target odor for under 2 min, and given that the two mood
tests were only separated by 2 or 3 min. Longer exposure or
repeated exposure may show significant cumulative effects
on individual and group moods.
In addition, results of this study suggest that people may
differentially react to differences between odors of different
ages and gender, but may not be able to articulate them on a
simple discrimination task. For example, although underarm odors of older women and of young men had different
effects on observers’ depressive moods, both were perceived to be intense, unpleasant, masculine, and aging on
odor judgment tasks. This could be because the mood assessment task provided a relatively meaningful and self-relevant context, whereas the simple discrimination task did
not. Context may be important to process olfactory information. This is consistent with observations that meaningful
scents trigger emotional memories [33,34]. Van Toller and
Kendal-Reed [35] made the distinction between an olfactory experience that lends itself to linguistic description
(e.g., naming or labeling an odor) versus an olfactory experience that is intuitive and nonlinguistic (e.g., emotional relationship with one’s grandmother evoked by the smell of
lavender water). Perhaps the mood task used in this work
provided such a social linguistic environment for people to
describe something that may be social but may not be intrinsically linguistic.
The hedonic congruency effect was found in previous
studies but not in the present study. This could be due to a
number of reasons. One of them could be that natural body
odors carry biologically significant information that impacts
on people differently from the way artificial fragrances or
chemical irritants do. Another reason could be due to differences in analyses. The present study controlled for differences due to subjects’ perceived odor intensity/pleasantness,
Table 4
Mean changes in depressive, hostile, and positive mood by types of odor
Depression
Hostility
Positive
Types of odor
Mean
SE
Mean
SE
Mean
SE
Little girls
Little boys
College women
College men
Older women
Older men
Home odor
22.062
22.078
22.020
20.0543
24.213
21.688
23.093
0.768
0.788
0.764
0.852
0.792
0.758
0.799
20.317
20.408
20.876
20.418
20.774
20.908
20.295
0.307
0.311
0.308
0.343
0.317
0.303
0.317
20.892
21.147
20.842
21.955
20.924
20.859
21.624
0.378
0.383
0.374
0.419
0.390
0.381
0.389
Children
Young adults
Older adults
22.072
21.002
22.934
0.597
0.568
0.561
20.400
20.629
20.820
0.229
0.221
0.217
21.008
21.385
20.876
0.295
0.281
0.281
Women’s odors
Men’s odors
22.766
21.239
0.455
0.454
20.658
20.576
0.175
0.175
20.874
21.305
0.224
0.226
D. Chen and J. Haviland-Jones / Physiology & Behavior 68 (1999) 241–250
whereas previous studies did not. In addition, the intensity/
pleasantness of the odors in the present study was measured
relative to one another, and not in the absolute sense.
We found in this study that underarm odors differentially
impacted on the depressive mood but not on the positive
mood. One possible explanation is that almost all existing
mood tests have more items on negative than on positive
moods, a potential bias that might have made the tests more
sensitive to detecting negative than positive changes. Another plausible explanation is that the observed olfactory
impact on negative moods reflected a primordial impression
of men and young adults. In many species, including humans [36], males and young adults tend to be more aggressive then females and older adults.
There are several important limitations of this study. For
one thing, target odors were not judged individually but were
judged as a group. It is possible that individual differences
exist within each donor group. However, the experimenter
observed that at least three of the five college men and two of
the five older women had strong odors. Thus, it was not simply the odor of one individual within the group that biased the
judgment of the entire group. Nevertheless, future studies
may want to evaluate the odors both on an individual as well
as on a group basis. Secondly, a number of studies have indicated that the olfactory sensitivity in women varies by the different phases of their menstrual cycles [37,38]. However,
menstrual cycles of female observers were not recorded in
this study, largely because the olfactory screening tests would
already discriminate between the olfactorily more sensitive
observers from those who were not.
In conclusion, the findings of the present work replicate
and extent past research on olfactory identification of gender. This work examines olfactory identification of gender
in conjunction with age from a developmental perspective,
and examines the effects of body odors from different ages
and genders on peoples’ mood. It demonstrates for the first
time that body odors carrying social-biological information
differentially bias odor recipients’ moods. This finding may
have an important implication for existing research on human perception and interaction.
Acknowledgments
We are grateful to Dr. Erich Labouvie for helpful discussions. We are also grateful to Drs. Julie Mennella, Mort
Bart, and Paul Vincenti for discussions, to Dr. Terry McGuire
for comments on the first draft, Dr. Pamela Dalton for comments and suggestions on a later draft, and to the anonymous
reviewers for comments and suggestions. We thank all our
participants. This research was supported in part by Tova
Dissertation Fellowship from the Olfactory Research Fund.
References
[1] Brown RE, Macdonald DW. Social Odors in Mammals, vols 1 & 2.
Oxford: Clarendon Press, 1985.
249
[2] McClintock MK. Pheromonal regulation of the ovarian cycle: enhancement, suppression, and synchrony. In: Vandenbergh JG, editor. Pheromones and Reproduction in Mammals. New York: Academic Press, 1983. pp. 113–49.
[3] Eisenberg JF, Kleiman DG. Olfactory communication in mammals.
Annu Rev Ecol Sys 1972;3:1–32.
[4] Rottman SJ, Snowndown CT. Demonstration and analysis of an alarm
pheromone in mice. J Comp Physiol Psychol 1972;81:483–90.
[5] Stern K, McClintock MK. Regulation of ovulation by human pheromones. Nature 1998;392:177–9.
[6] Doty R. Olfactory communication in humans. Chem Senses 1981;6:
351–76.
[7] Kaitz M, Eidelman AI. Smell-recognition of newborns by women
who are not mothers. Chem Senses 1992;17:225–9.
[8] Russell MJ. Human olfactory communication. Nature 1976;260:520–2.
[9] Schaal B, Porter RH. “Microsmatic humans” revisited: the generation and perception of chemical signals. In: Slater PJB, Rosenblatt
JS, Beer C, Milinski M, editors. Advances in the Study of Behavior,
vol 20. San Diego, CA: Academic Press, Inc., 1991. pp 135–199.
[10] Ehrlichman H, Bastone L. The use of odor in the study of emotion.
In: Van Toller S, Dodd GH, editors. Fragrance: The Psychology and
Biology of Perfume. New York: Elsevier Applied Science, 1992.
pp. 143–59.
[11] Ehrlichman H, Bastone L. Olfaction and emotion. In: Serby MJ,
Chobor KL, editors. Science of Olfaction. New York: Springer Verlag, 1992. pp. 410–38.
[12] Schiffman SS, Sattely-Miller EA, Suggs MS, Graham BG. The effect of pleasant odors and hormone status on mood of women at
midlife. Brain Res Bull 1995;36:19–29.
[13] Schiffman SS, Suggs MS, Sattely-Miller EA. Effect of pleasant
odors on mood of males at midlife: comparison of African-American and European-American men. Brain Res Bull 1995;36:31–7.
[14] Badia P, Wesensten N, Lammers W, Culpepper J, Harsh J. Responsiveness to olfactory stimuli presented in sleep. Physiol Behav
1990;48:87–90.
[15] Schwartz GE. Levels of awareness and “awareness without awareness”: From data to theory. In: Hameroff SR, Kaszniak AW, Scott
AC, editors. Towards a Science of Consciousness: The First Tuscon
Discussions and Debates: Cambridge, MA: The MIT Press, 1996.
pp 279–93.
[16] Wallace P. Individual discrimination of humans by odor. Physiol
Behav 1977;19:577–9.
[17] Doty RL, Green PA, Ram C, Yankell SL. Communication of gender
from human breath odors: relationship to perceived intensity and
pleasantness. Horm Behav 1982;16:13–22.
[18] Schleidt M, Hold B, Attili G. A cross-cultural study on the attitude
towards personal odors. J Chem Ecol 1981;7:19–31.
[19] Doty RL, Orndorff MM, Leyden J, Kligman A. Communication of
gender from human axillary odors: relationship to perceived intensity and hedonicity. Behav Biol 1978;23:373–80.
[20] Mallet P, Schaal B. Rating and recognition of peers’ personal odors
by 9-year-old children: an exploratory study. J Gen Psychol 1998;
125:47–64.
[21] Schleidt M. Personal odor and nonverbal communication. Ethol Sociobiol 1980;1:225–31.
[22] Brooksbank BW, Brown R, Gustafsson JA. The detection of 5aandrost-16-en-3a-o1 in human male axillary sweat. Experientia
1974;30:864–5.
[23] Gower DB, Bird S, Sharma P, House FR. Axillary 5a-androst-16en-3-one in men and women: relationships with olfactory acuity to
odorous 16-androstenes. Experientia 1985;41:1134–6.
[24] Dorries KM, Schmidt HJ, Beauchamp GK, Wysocki CJ. Changes in
sensitivity to the odor of androstenone during adolescence. Dev
Psychobiol 1989;22:423–35.
[25] Izard CE, Libero DZ, Putnam P, Haynes OM. Stability of emotion
experiences and their relations to traits of personality. J Pers Soc
Psychol 1993;64:847–60.
250
D. Chen and J. Haviland-Jones / Physiology & Behavior 68 (1999) 241–250
[26] Stevens J. Applied Multivariate Statistics for the Social Sciences,
2nd. ed.. Hillsdale, NJ: Lawrence Erlbaum Associates, 1992.
[27] Stapley JC, Haviland JM. Beyond depression: gender differences in
normal adolescents’ emotional experiences. Sex Roles 1989;20:
295–308.
[28] Cohen J. Statistical Power Analysis for the Behavioral Sciences.
New York: Academic Press, 1977.
[29] Judd HL, Korenman SG. Effects of aging on reproductive function
in women. In: Korenman SG, editor. Endocrine Aspects of Aging.
New York: Elsevier Biomedical Press, 1982. pp. 163–97.
[30] Perskey H. Psychoendocrinology of Human Sexual Behavior. New
York: Praeger Publishers, 1987.
[31] Kent JR, Acone AB. Plasma testosterone and aging in males. In:
Vermuelen A, editor. Androgens in Normal and Pathological Conditions. Amsterdam: Excerpta Medica, 1966. pp. 31–5.
[32] Perskey H, Smith KD, Basu GK. Relation of psychological measures of aggression and hostility to testosterone production in man.
Psychosom Med 1971;33:265–77.
[33] Proust M. Remembering of things past (translated by GKS Moncrieff & T Kilmartin), vol. 1. New York: Random House, 1981.
[34] Synnott A. Roses, coffee and lovers: the meanings of smell. In: Gilert
A, editor. Compendium of Olfactory Research, 1982–1994. Dubuque,
IA: Kendall/Hunt Publishing Company, 1995. pp. 117–28.
[35] Van Toller S, Kendal-Reed M. A possible protocognitive role for
odor in human infant development. Brain Cognit 1995;29:275–93.
[36] Wilson M, Daly M. competitiveness, risk taking, and violence: the
young male syndrome. Ethol Sociobiol 1985;6:59–73.
[37] Doty RL, Snyder PJ, Huggins GR, Lowry LD. Endocrine, cardiovascular, and psychological correlates of olfactory sensitivity
changes during the human menstrual cycle. J Comp Physiol Psychol
1981;95:45–60.
[38] Pause BM, Bernfried S, Krauel K, Fehm-Wolfsdorf G, Ferstl R. Olfactory information processing during the course of the menstrual
cycle. Biol Psychol 1996;44:31–54.
[39] Lewis-Beck MS. Applied Regression: An Introduction. Beverly
Hills, CA: Sage Publications, 1980.