Content uploaded by Tor D Wager
Author content
All content in this area was uploaded by Tor D Wager on Apr 16, 2018
Content may be subject to copyright.
Research Article
A Warm Heart and a Clear Head
The Contingent Effects of Weather on Mood and Cognition
Matthew C. Keller,
1
Barbara L. Fredrickson,
2
Oscar Ybarra,
2
Ste
´phane Co
ˆte
´,
3
Kareem Johnson,
2
Joe Mikels,
4
Anne Conway,
5
and Tor Wager
6
1
Department of Psychiatry, Virginia Institute for Psychiatric and Behavioral Genetics;
2
Department of Psychology,
University of Michigan at Ann Arbor;
3
Rotman School of Management, University of Toronto, Toronto, Ontario, Canada;
4
Department of Psychology, Stanford University;
5
Department of Psychology, Pennsylvania State University; and
6
Department of Psychology, Columbia University
ABSTRACT—Prior studies on the association between
weather and psychological changes have produced mixed
results. In part, this inconsistency may be because weath-
er’s psychological effects are moderated by two important
factors: the season and time spent outside. In two corre-
lational studies and an experiment manipulating partici-
pants’ time outdoors (total N5605), pleasant weather
(higher temperature or barometric pressure) was related
to higher mood, better memory, and ‘‘broadened’’ cogni-
tive style during the spring as time spent outside increased.
The same relationships between mood and weather were
not observed during other times of year, and indeed hotter
weather was associated with lower mood in the summer.
These results are consistent with findings on seasonal af-
fective disorder, and suggest that pleasant weather im-
proves mood and broadens cognition in the spring because
people have been deprived of such weather during the
winter.
Weather has long held a central place in human experience, and
if lay psychology is to be believed, weather continues to be an
important determinant of everyday mood and behavior in mod-
ern life (Persinger, 1980; Watson, 2000). Given the pervasive-
ness of this belief, the paucity of scientific knowledge on how
weather affects human psychology is surprising. Although the
effects of seasons on mood and depression are well documented
(e.g., Harmatz et al., 2000; Rosenthal et al., 1984), compara-
tively few studies have assessed the relationship between daily
variation in weather and human mood and cognition.
We found only two studies related to cognition and weather. In
a study manipulating temperature, Allen and Fischer (1978)
found that performance on a paired-association memory task
peaked at 72 1F (22 1C) and declined with warmer or cooler
temperature; Sinclair, Mark, and Clore (1994) found that days
that were both sunny and warm were associated with more
heuristic and less systematic processing than cloudy and cool
days. The number of studies on the relation between weather and
mood is somewhat larger. In some studies, low levels of humidity
(Sanders & Brizzolara, 1982), high levels of sunlight (Cun-
ningham, 1979; Parrott & Sabini, 1990; Schwarz & Clore, 1983),
high barometric pressure (Goldstein, 1972), and high temper-
ature (Cunningham, 1979; Howarth & Hoffman, 1984) have
been associated with high mood. However, high temperature has
also been associated with low mood (Goldstein, 1972) and low
potency (low potency is similar to low mood; Howarth & Hoff-
man, 1984), and two other studies found no relationships be-
tween mood and any weather variable (Clark & Watson, 1988;
Watson, 2000).
The largest test of the weather-mood hypothesis (Watson,
2000) collected daily mood reports from 478 undergraduate
students in Dallas, Texas, during the fall or the spring (a total of
20,818 observations). No significant correlations were found
between mood (measured by self-report using the Positive and
Negative Affect Scale, or PANAS) and any of the assessed
weather variables (sunshine, barometric pressure, temperature,
or precipitation
1
). These null findings were noteworthy because
they called into question the commonly held belief that weather
affects mood.
However, other lines of research focusing on population-wide
behaviors suggest that weather does have some effect on psy-
chological processes. High temperature is reliably associated
Address correspondence to Matthew C. Keller, Virginia Institute for
Psychiatric and Behavioral Genetics, 800 East Leigh St., Richmond,
VA 23219; e-mail: matthew.c.keller@gmail.com.
1
Watson (2000) conducted his primary analysis on sunshine and rain, but also
reported that neither temperature nor pressure was related to mood.
PSYCHOLOGICAL SCIENCE
724 Volume 16—Number 9Copyright r2005 American Psychological Society
with violent behavior (Anderson, 2001; Baron & Ransberger,
1978), though it is unclear whether this association is best ex-
plained by physiological effects of temperature on aggression
(Anderson, 2001) or by indirect effects due to the higher like-
lihood of interpersonal interactions in pleasant weather (Rotton
& Cohn, 2000). A second line of research documents that sun-
nier weather is related to slightly higher stock market returns
(Saunders, 1993). A possible interpretation for both findings is
that higher temperature and sunlight increase risk tolerance,
which in turn increases likelihood for aggression and buying
behavior, respectively.
EFFECTS OF SEASON ON PSYCHOLOGICAL
PROCESSES
In contrast to the relatively sparse literature on the psycholog-
ical changes associated with the weather, hundreds of articles in
the past 20 years have considered seasonal effects on psycho-
logical functioning, and in particular on seasonal affective
disorder (SAD). SAD is seasonally recurrent depression with
typical onset during the fall or winter and remission in the
spring. It is characterized by typical depressive symptoms as
well as atypical symptoms, such as longer sleep duration and
carbohydrate craving (Rosenthal et al., 1984). Cognitive im-
pairments in memory, learning, and visual-spatial ability have
also been documented (Michalon, Eskes, & Mate-Kole, 1997;
O’Brien, Sahakian, & Checkley, 1993). Given that mood tends to
reach a low point in the general population during the winter
(Harmatz et al., 2000), and that about half of nondepressed
people manifest some degree of SAD symptoms during northern
winters (Dam, Jakobsen, & Mellerup, 1998; Kasper, Wehr,
Bartko, Gaist, & Rosenthal, 1989), SAD can be seen as one
extreme along a continuum of normal wintertime behavioral
changes.
Several findings about seasonal effects suggest that exposure
to sunlight immediately affects mood and cognition. Placebo-
controlled studies document that artificial sunlight (produced by
a very bright lamp) improves mood and diminishes SAD
symptoms for a majority of SAD and non-SAD depressed pa-
tients (Kripke, 1998; Stain-Malmgren, Kjellman, & A
˚berg-
Wistedt, 1998), and, most tellingly, improves mood and vitality
among nondepressed subjects (Leppamaki, Partonen, & Lonn-
quist, 2002; Leppamaki, Partonen, Piiroinen, Haukka, &
Lonnquist, 2003). Effects are often observed after the first
bright-light treatment (Kripke, 1998). Moreover, Lambert, Reid,
Kaye, Jennings, and Esler (2002) found that brain serotonin
production in 101 healthy, non-SAD males rose or dipped as
naturally occurring daily sunlight increased or decreased, re-
spectively.
These findings appear inconsistent with the weak and variable
weather findings we reviewed earlier. Exposure to at least one
weather phenomenon, sunlight, appears to immediately affect
mood and serotonin levels among both depressed and nonde-
pressed people. This suggests that weather does indeed affect
mood, and possibly cognition.
OVERVIEW OF THE CURRENT RESEARCH
For the same reasons that the hedonic value of any emotion-
inducing stimulus decreases with continued exposure (Cabanac,
1971), it would be maladaptive for pleasant weather to have the
same hedonic effect irrespective of prior exposure. Thus, we
predicted that warm and sunny days in the spring (when people
have been deprived of such weather) boost mood and alter
cognition more than warm and sunny days later in the year, when
pleasant weather is less of a novelty.
In addition, it is reasonable to assume that one must be ex-
posed to the weather for it to affect one’s psychological proc-
esses. However, people in industrialized countries spend an
average of 93% of their time inside (Woodcock & Custovic,
1998) and thus are largely disconnected from the weather out-
side. This suggests that surveys correlating mood with weather
might fail to uncover any connection simply because many
people have little exposure to the weather.
We conducted three studies to test the hypothesis that the
effects of weather on mood and cognition are moderated by
season and by degree of direct exposure to the weather. Although
sunlight has received the lion’s share of attention in the SAD
literature, the link between temperature and violence suggests
that temperature is also a likely candidate for affecting psy-
chological processes. Other researchers have reported memory
impairment associated with SAD (Allen & Fischer, 1978; Mic-
halon et al., 1997; O’Brien et al., 1993) and manipulated tem-
perature (Allen & Fischer, 1978). Thus, the focus of the current
studies was on how atmospheric pressure (an assay of sunlight)
and temperature are related to cognition and memory.
STUDY 1: RELATIONSHIPS AMONG WEATHER, MOOD,
AND COGNITION IN THE SPRING
Method
We collected data from 97 participants (54 female and 43 male
students ages 18–29) in Ann Arbor, Michigan (421north lati-
tude) between April 5 and June 15, 2001. Participants re-
sponded to a newspaper advertisement and were paid for their
time. They completed all measures once, during a single session,
and were run individually. Upon arrival, participants filled out
questionnaires to report their current mood, how much time they
spent outside the day that they came to the lab, their activity
level that day (on a verbally anchored scale from 6, very active,
to 1, very inactive), and demographic information. They then
completed two cognitive tasks.
Weather was not mentioned until debriefing. We obtained data
on temperature and barometric pressure from the National
Climatic Data Center (NCDC). Measurements of sunlight were
unavailable, but barometric pressure served as a good substitute
Volume 16—Number 9 725
M.C. Keller et al.
for sunlight. High pressure is typically associated with clear,
sunny weather, whereas low pressure is associated with clouds,
precipitation, and storm fronts (Ahrens, 2000). We did not
combine temperature and pressure to form an underlying ‘‘good
weather’’ variable because the two variables were unrelated (r5
.06, n.s.).
We collected information on the following three dependent
variables:
Mood valence: Participants reported their mood using the
PANAS mood scale (Watson, Clark, & Tellegen, 1988). We
subtracted Negative Activation from Positive Activation to
create a measure of mood valence, with higher scores de-
noting better mood (see Barrett & Russell, 1998, for justifi-
cation of this rotation). In this article, we focus on mood
valence rather than Positive Activation (from the PANAS)
because (a) our findings in Study 1 indicated that mood va-
lence was more strongly related to weather than positive ac-
tivation, and (b) mood valence can be assessed quickly with a
single-item measure, which was important for Studies 2 and 3.
Digit span: Digit span is an excellent index of working
memory capacity (Wechsler, 1997). It was defined as the
maximum number of digits participants were able to repeat
immediately after hearing a digit string.
Openness to new information: We were also interested in how
weather affects cognitive broadening. Cognitive broadening
describes a style of thinking in which people become more
creative and is hypothesized to be an adaptive shift in cog-
nition that leads to behavioral flexibility and exploration
(Fredrickson, 2001; Isen, 2000). Individuals who are in a
broad mind-set should modify previously formed attitudes
when new information contradicts those attitudes. To measure
cognitive broadening, we randomly assigned participants
either to read favorable and then unfavorable information
about a fictitious employee or to read the unfavorable infor-
mation first (see Kruglanski & Freund, 1983). Participants
then rated the employee’s intelligence and performance.
Openness to new information was defined as the participant’s
overall rating of the employee if the unfavorable information
was presented first and the reversed rating if the favorable
information was presented first. Higher scores indicate a
willingness to update initial impressions, reflecting a broad
mind-set.
Results and Discussion
We used multiple regression, controlling for activity level and
the time participants came to the lab, to test our prediction that
the effects of weather would be moderated by the amount of time
spent outside.
2
As in some of the previous research (Clark &
Watson, 1988; Watson, 2000), neither temperature nor baro-
metric pressure was directly related to mood valence. However,
the interactions of time spent outside with temperature and with
barometric pressure were both significantly related to mood
valence in the expected direction: As time spent outside in-
creased, the temperature-mood and pressure-mood relation-
ships became more positive (Table 1). These relationships are
illustrated in Figures 1a and 1b using median-splits on time
spent outside: Among participants who spent more than 30 min
outside, higher temperature and pressure were associated with
higher moods, but among those who spent 30 min or less outside,
this relationship was reversed.
A similar pattern occurred for the cognitive measures (Table
1). Pressure (but not temperature) became more positively re-
lated to digit span (Fig. 1c) and to openness to new information
(Fig. 1d) as time spent outside increased; that is, among people
who spent more than 30 min outdoors, clearer days were asso-
ciated with higher digit spans and more flexible thinking styles.
The relation between digit span and barometric pressure is
noteworthy because digit span is a common component of IQ
scales (e.g., Wechsler, 1997) and is often considered a stable,
trait variable. A supplementary analysis revealed that mood did
not mediate these cognitive effects.
TABLE 1
Simultaneous Regression Model Relating Weather and Time
Spent Outside to Dependent Measures in Study 1
Dependent variable and predictor B SE(B) r
2
Mood valence
Temperature .002 .101 .000
Pressure .109 .127 .010
Time outside .009 .118 .000
Time Outside Temperature .216
n
.097 .046
Time Outside Pressure .249
w
.132 .063
Digit span
Temperature .046 .105 .002
Pressure .009 .104 .000
Time outside .202
w
.118 .032
Time Outside Temperature .042 .100 .002
Time Outside Pressure .214
n
.101 .048
Openness to new information
Temperature .054 .104 .003
Pressure .181
w
.104 .028
Time outside .097 .102 .009
Time Outside Temperature .022 .099 .000
Time Outside Pressure .378
nn
.110 .113
Note. Sample sizes vary because of equipment failure and other random errors
in data collection. All variables are standardized. Interaction terms are the
product of the two standardized predictors in question and are interpreted as
the change in the regression slope between the standardized weather and de-
pendent variables when time spent outside increases by one standard deviation
(Jaccard, Turrisi, & Wan, 1990). The analysis controlled for activity level and
time of day participants came in. The omnibus tests for mood valence (n582)
and openness to new information (n596) were significant, F(7, 74) 53.38, p5
.004, and F(7, 88) 54.30, p5.007, respectively. The omnibus test for digit
span (n597) was not significant, F(7, 89) 51.60, p5.112.
w
p<.10.
n
p<.05.
nn
p<.01.
2
Assumptions regarding normality of the sampling distributions and equality
of variances were satisfied unless otherwise noted.
726 Volume 16—Number 9
Weather, Mood, and Cognition
STUDY 2: MANIPULATION OF TIME SPENT OUTSIDE
Although the results of Study 1 were consistent with the hy-
pothesis that the effect of weather on mood and cognition de-
pends on spending time outdoors, the results could also be
accounted for if better moods or broader mind-sets are associ-
ated with greater willingness or ability to go outside in pleasant
weather. To address this self-selection issue, in Study 2 we
manipulated the time that participants spent outside before and
after we assessed their mood and memory.
Method
We collected data from 121 participants (85 females and 36
males ages 18–32) in Ann Arbor, Michigan, between April 16
and July 27, 2003. Participants responded to an advertisement
in the local newspaper that sought people who relieved stress by
‘‘walking outside, dancing, or meditating’’ and were paid for
their time.
Study 2 was part of an unrelated experiment (to be reported
elsewhere) on how specific stress-relieving activities affect
coping with stressful life events. We matched weather conditions
to the extent possible by yoking each participant’s session with
the session of another participant. Yoked sessions were sched-
uled in immediate succession on the same day, with one par-
ticipant randomly assigned to be in the inside condition and the
other to be in the outside condition (see the next paragraph).
The first part of each session took place in a windowless room.
The participant filled out a baseline questionnaire packet that
asked for demographic information and included a measure of
mood, as well as several stress measures unrelated to the present
study. A research assistant then assessed the participant’s digit
span. If the participant danced to relieve stress (n542), he or
she was then randomly assigned to either dance indoors (inside
condition) or walk around an outdoor track (outside condition). If
the participant walked outdoors to relieve stress (n551), he or
she was randomly assigned to either walk in a nearby arboretum
(outside condition) or walk indoors on a treadmill (inside con-
dition). Finally, if the participant meditated to relieve stress (n5
28), he or she was randomly assigned to either meditate (inside
or outside) or proofread a passage (inside or outside).
3
The
participant engaged in the assigned activity for 30 min and then
returned to complete postactivity measures either outdoors
(outside condition) or indoors (inside condition). These mea-
sures included the same mood and digit span measures that were
completed earlier. Data on temperature and barometric pressure
were again obtained from the NCDC. Weather was not men-
tioned until debriefing.
Openness to new information could not be collected because
of time constraints. The following two variables were relevant to
the current study:
Mood valence change: Mood valence was measured using an
affect grid (J.A. Russell, Weiss, & Mendelsohn, 1989). We
subtracted the baseline score from the postactivity score to
create an index of mood change over the course of the study.
Digit span change: Digit span was measured using the same
procedure employed in Study 1. Digit span change was the
postactivity score minus the baseline score.
Results and Discussion
The interactions of outside/inside condition and weather were in
the same direction and of similar magnitude for all three groups
of participants (i.e., dancers, walkers, and meditators), so analy-
ses are collapsed across these three groups. Table 2 shows the
values of the parameters for the regression equations.
4
As in
Study 1, neither temperature nor pressure was directly related to
mood, but moods improved for participants who were randomly
assigned to be outside on warm, high-pressure (clear) days,
whereas moods declined for those randomly assigned to be in-
side on such days (Figs. 2a and 2b). This interaction was sig-
nificant for temperature and marginally significant for pressure.
Temperature (but not pressure) was positively related to digit
span change among participants assigned to the outside con-
dition (Fig. 2c). It should be noted that pressure, not tempera-
Fig. 1. Study 1 results: mood valence (a, b), digit span (c), and openness
to new information (d) as a function of temperature or barometric pressure
and amount of time spent outdoors in the spring.
3
It should be noted that neither the inside condition nor the outside condition
was equivalent for the dancers and nature walkers (in the outside condition,
dancers walked on a track, but nature walkers walked in a park; in the inside
condition, dancers danced to music, but nature walkers walked on a treadmill).
The empty cells of this design made it impossible to assess the interaction be-
tween stress-relief activity and outside/inside condition, which was of little
theoretical interest, but were not problematic to testing our hypothesis.
4
We analyzed difference scores for each dependent variable, which is equiva-
lent to conducting repeated measures analyses (Maxwell & Delaney, 1990).
Volume 16—Number 9 727
M.C. Keller et al.
ture, was related to digit span in Study 1. We discuss this dis-
crepancy in the General Discussion.
STUDY 3: RELATIONSHIPS AMONG WEATHER AND
MOOD ACROSS LOCATIONS AND SEASONS
Study 2 substantively replicated the results of Study 1 and
suggested that being inside or outside causally changes the
weather-mood and weather-memory associations. Both studies,
however, were performed in the spring and early summer in a
northern climate, when warm and sunny weather is still some-
what novel. In Study 3, we collected mood data from people in
varied geographical locations across 1 year to assess whether the
weather-mood association differed across seasons and locations.
Method
From January to December 2002, we collected information from
387 participants who volunteered to participate on a Web site
dedicated to on-line psychological studies: 281 were females
and 106 were males; 201 lived in the northern United States and
Canada (381N), 174 lived in the southern United States (<
381N), 12 lived in Europe (>381N); and ages ranged from 18
through 56 (M525.9, SD 58.7).
Potential participants clicked on a link titled ‘‘Short Dispo-
sition Survey,’’ which led to our consent form. Those who agreed
to participate completed a demographic page, an implicit mood
task, and a single-question mood survey. They then reported how
much time they had spent outside that day (average of 64 min)
and how active they had been that day (using the same 6-point
verbally anchored scale used in Study 1). Weather was not
mentioned until debriefing.
As in Studies 1 and 2, we obtained data on temperature and
barometric pressure from the NCDC. We analyzed sea-level
pressure rather than station-level pressure to account for pres-
sure differences due to elevation. To account for geographical
differences in mean temperature, we subtracted the state’s,
province’s, or (if outside North America) country’s average
temperature across the year from the observed temperature on
the day the participant completed the survey (referred to as
‘‘temperature’’ unless otherwise noted).
We collected the following two dependent measures:
Implicit mood valence: First, we administered an implicit
measure of mood to expand our previous findings regarding
weather-mood associations. Participants were asked to fill in
the blank letters of eight words that had one or two letters
removed from them (e.g., two of these words were ‘‘G L O _ _Y’’
and ‘‘J O _’’). In each case, a neutral word (‘‘G L O S S Y’’ or ‘‘J
O B’’) or mood-descriptor words (‘‘G L O O M Y’’ or ‘‘J O Y’’)
could be created. Four of these mood descriptors had a pos-
TABLE 2
Simultaneous Regression Model Relating Weather and Time
Spent Outside to Dependent Measures in Study 2
Dependent variable and predictor B SE(B) r
2
Mood valence change (affect grid)
Temperature .021 .090 .000
Pressure .027 .090 .001
Outside (1)/inside (1) .017 .090 .000
Time Outside Temperature .299
nn
.090 .082
Time Outside Pressure .220
w
.090 .040
Digit span change
Temperature .115 .093 .018
Pressure .133 .091 .022
Outside (1)/inside (1) .062 .092 .004
Time Outside Temperature .194
n
.093 .038
Time Outside Pressure .083 .091 .008
Note. Sample sizes vary because of equipment failure and other random errors
in data collection. All variables are standardized. Interaction terms are the
product of the two standardized predictors in question and are interpreted as
the change in the regression slope between the standardized weather and de-
pendent variables when time spent outside increases by one standard deviation
(Jaccard, Turrisi, & Wan, 1990). The type of activity the participant used to
relieve stress (dancing, meditation, or nature walking) was controlled by en-
tering two variables (created using effect codes) into each regression equation.
Activity level and time of day were not controlled because of random assign-
ment. The omnibus test for mood valence change (n5120) was significant, F(7,
112) 52.22, p5.045. The omnibus test for digit span change (n5119) was not
significant, F(7, 111) 51.65, p5.141.
w
p<.10.
n
p<.05.
nn
p<.01.
Fig. 2. Study 2 results: change in mood valence (a, b) and digit span (c) as a function of temperature or barometric pressure and random
assignment to the inside or outside condition in the spring.
728 Volume 16—Number 9
Weather, Mood, and Cognition
itive valence, and four had a negative valence. Implicit mood
valence was defined as the number of completed words that
were mood descriptors with a positive valence minus the
number of completed words that were mood descriptors with a
negative valence. Such scales have been found to correlate
with momentary mood (Rusting & Larsen, 1998), ostensibly
because people tend to perceive stimuli as mood congruent in
ambiguous situations.
Explicit mood valence: Because prior experience indicated
that an affect grid is potentially confusing without in-person
instruction, we did not use an affect grid in this study. Instead,
participants indicated their current mood valence on a 9-
point Likert scale that was anchored by intensity descriptors
of mood valence (1 5very low,95very high).
Results and Discussion
The primary analyses were conducted separately for each sea-
son. We controlled for activity level, the time of day the ques-
tionnaire was completed, and (given the greater range of ages in
this study than in the others) age. No main effects or interaction
terms were statistically significant in the winter (January–
March) or fall (October–December) subsamples. The most ro-
bust results were in the spring (April–June; see Table 3). Results
were consistent with the results of the previous two studies in
that the main effects for temperature were nonsignificant for
each of the two dependent variables, whereas the interactions of
temperature and time spent outside were significant. As par-
ticipants spent more time outside in the spring, temperature
became significantly more related to explicit mood valence and
implicit mood valence (see Figs. 3a and 3b). The effects of
pressure during the spring were weaker than the effects of
temperature. The interaction of time spent outside and pressure
was marginally significant for implicit mood valence ( p<.10;
see Fig. 3c), but was unrelated to explicit mood valence.
Of note, warmer temperature in the summer was associated
with decreased explicit mood as time spent outside increased
(Time Outside Temperature B5.27, p5.02, r
2
5.08). This
effect was driven by participants living in southern climates
(Time Outside Temperature B5.36, p5.03, r
2
5.11; for
participants in northern climates, Time Outside Temperature
B5.05, p5.88, r
2
5.00). This result is similar to the findings
on the relation between temperature and violence (Rotton &
Cohn, 2000) and suggests a curvilinear relationship between
mood and temperature. We tested this possibility by regressing
raw (geographically uncorrected) temperature and squared raw
temperature across the whole year against explicit mood. As
expected, there was an inverted-U temperature-mood relation-
ship among participants who had spent more than 45 min out-
side (temperature squared B5.11, p5.03, r
2
5.01), with the
predicted maximum mood occurring at 67.4 1F (19.7 1C). This
effect was again much more prominent in southern climates (tem-
TABLE 3
Simultaneous Regression Model Relating Weather and Time
Spent Outside to Dependent Measures in Study 3 (Spring Only)
Dependent variable and predictor B SE(B) r
2
Explicit mood valence
Temperature .199 .118 .033
Pressure .262
n
.114 .060
Time outside .024 .124 .001
Time Outside Temperature .228
n
.111 .048
Time Outside Pressure .051 .118 .002
Implicit mood valence
Temperature .202
w
.120 .033
Pressure .074 .118 .006
Time outside .060 .125 .004
Time Outside Temperature .235
n
.111 .050
Time Outside Pressure .228
w
.119 .042
Note. All variables are standardized. Interaction terms are the product of the
two standardized predictors in question and are interpreted as the change in
the regression slope between the standardized weather and dependent vari-
ables when time spent outside increases by one standard deviation (Jaccard,
Turrisi, & Wan, 1990). The model controlled for age, activity level, and time of
day the questionnaire was completed. The omnibus tests for explicit and im-
plicit mood valence (n593) were not significant, F(8, 84) 51.34, p5.235, and
F(8, 84) 51.43, p5.198, respectively.
w
p<.10.
n
p<.05.
Fig. 3. Study 3 results: explicit mood valence (a) and implicit mood valence (b, c) as a function of temperature or
barometric pressure and amount of time spent outdoors in the spring.
Volume 16—Number 9 729
M.C. Keller et al.
perature squared B5.22, p5.02, r
2
5.04) than in northern
climates (temperature squared B5.06, p5.29, r
2
5.00).
Squared temperature was unrelated to mood among participants
who spent less than 45 min outside (temperature squared B5
.01, p5.87, r
2
5.00), again demonstrating the moderating
effect of being outdoors.
It should be noted that the curvilinear effect of temperature
does not explain the pattern of findings across the seasons. If it
did, a stronger positive association between mood and temper-
ature would be expected in the fall than in the spring; if mood is
optimal at 67 1F (19 1C), temperature has more room to increase
mood during the fall, which has an average temperature of 62 1F
(17 1C), than during the spring, which has an average temper-
ature of 70 1F (21 1C). It should also be noted that the effect of
temperature change is asymmetrical: Temperature changes to-
ward cooler weather in the fall did not predict higher mood.
Rather, there appears to be something uniquely uplifting about
warm days in the spring.
GENERAL DISCUSSION
Three studies examined how temperature and pressure relate to
mood and cognition. Study 1, conducted during a northern
spring, indicated that spending time outdoors increases the
relationships of temperature and barometric pressure with
mood, digit span, and openness to new information. Study 2, in
which participants were randomly assigned to be indoors or
outdoors, suggests that being outdoors is a causal factor that
changes weather-mood and weather-memory relationships.
Study 3 indicated that, in addition to time spent outside, season
is a critical moderator of weather’s effects on mood. Exposure to
higher temperatures predicted increased mood during the spring
but had the opposite effect on mood during the summer, espe-
cially among participants living in southern climates, where
high temperatures are increasingly unpleasant.
Contrary to our initial expectations, the effects of the weather
on people who spent almost all of their time indoors (i.e., less
than 30–45 min outside) was nearly as strong (in the opposite
direction) as the effects on those who spent their time outdoors.
This result was obtained in all three studies. One possible ex-
planation for this result is that people consciously resent being
cooped up indoors when the weather is pleasant in the spring.
Another possibility is that brief exposure to pleasant weather
places people in mood and mind states that make normal day-to-
day indoor activities feel boring or irritating. The current find-
ings do not address the question of whether the effects of weather
observed in these studies are due to conscious mediation, to
direct physiological effects of the weather, or to some other
process.
The overall 95% confidence interval for the springtime Time
Outside Weather Bs across all 14 tests was .18 .07, meaning
that spending about 30 to 45 min more outside increased the
slope of the relation between standardized temperature or
pressure and standardized mood or cognition by .18 units. This
probably underestimates the true effect given the error certain to
exist in the psychological and behavioral measures (D.W. Rus-
sell, Kahn, Spoth, & Altmaier, 1998). Nevertheless, even if er-
ror-free measures were available, it is doubtful these effects
would be very large simply because weather is likely to be but
one among many factors that influence interpersonal differences
in mood and cognitive style.
Although the pattern of results forms a coherent picture across
the three studies, two apparent discrepancies deserve fuller
consideration. First, the interaction between time spent outside
and barometric pressure did not approach significance ( p<.10)
for explicit mood in Study 3 during the spring, although this
interaction was significant or marginally significant for all other
tests involving mood. Second, in Study 1, as time outside in-
creased, pressure significantly predicted increases in digit span,
whereas in Study 2, it was temperature that had this effect. What
should be made of these seeming inconsistencies? Not very
much, we argue. Across varied locations and different meth-
odologies, 12 of 14 springtime Time Outside Weather Bterms
were in the predicted direction (B>0), a highly improbable
pattern of results given the null hypothesis of no effect (exact
binomial test p<.001). Moreover, 10 of these 14 Bterms were
significant or marginally significant, whereas fewer than 2
should have been if there really were no effects (exact binomial
test p<10
8
). It is vanishingly unlikely that this pattern of
results was due to random error. The ‘‘inconsistencies’’ between
studies (as judged by the p5.05 threshold) are exactly what
should be expected given the sample sizes employed and the
likely size of the effect.
Our findings support the hypothesis that both the amount of
time people spend outdoors and the season moderate weather’s
effects on mood and cognition. We hypothesize that pleasant
springtime weather is a zeitgeber for changing mood and cog-
nition from their wintertime settings back to their baseline
settings. If future work continues to support the hypotheses of
this article, the behavioral prescription is straightforward: If you
wish to reap the psychological benefits of good springtime
weather, go outside.
Acknowledgments—We thank O. Schultheiss, N. Schwarz, J.
Priester, P. Samson, and W. Kuhn for suggestions. We also are
grateful for the hard work of several research assistants: Gloria
Jen, Melissa McGivern, Christine Crosby, and Danelle Filips.
This work was supported by a fellowship from the National
Science Foundation (M.C.K.), Grant MH59615 from the Na-
tional Institute of Mental Health (B.L.F), and funds from the
John Templeton Foundation (B.L.F.).
REFERENCES
Ahrens, D.C. (2000). Meteorology today. Pacific Grove, CA: Brooks/
Cole.
730 Volume 16—Number 9
Weather, Mood, and Cognition
Allen, M.A., & Fischer, G.J. (1978). Ambient temperature effects on
paired associate learning. Ergonomics,21, 95–101.
Anderson, C.A. (2001). Heat and violence. Current Directions in Psy-
chological Science,10, 33–38.
Baron, R.A., & Ransberger, V.M. (1978). Ambient temperature and the
occurrence of collective violence: The long, hot summer revisited.
Journal of Personality and Social Psychology,36, 351–360.
Barrett, L., & Russell, J. (1998). Independence and bipolarity in the
structure of current affect. Journal of Personality and Social
Psychology,74, 967–984.
Cabanac, M. (1971). Physiological role of pleasure. Science,173,
1103–1107.
Clark, L.A., & Watson, D. (1988). Mood and the mundane: Relations
between daily life events and self-reported mood. Journal of
Personality and Social Psychology,54, 296–308.
Cunningham, M.R. (1979). Weather, mood, and helping behavior:
Quasi-experiments with the sunshine samaritan. Journal of Per-
sonality and Social Psychology,37, 1947–1956.
Dam, H., Jakobsen, K., & Mellerup, E. (1998). Prevalence of winter
depression in Denmark. Acta Psychiatrica Scandinavica,97, 1–4.
Fredrickson, B.L. (2001). The role of positive emotions in positive
psychology: The broaden-and-build theory of positive emotions.
American Psychologist,57, 218–226.
Goldstein, K.M. (1972). Weather, mood, and internal-external control.
Perceptual Motor Skills,35, 786.
Harmatz, M.G., Well, A.D., Overtree, C.E., Kawamura, K.Y., Rosal, M.,
& Ockene, I.S. (2000). Seasonal variation of depression and other
moods: A longitudinal approach. Journal of Biological Rhythms,
15, 344–350.
Howarth, E., & Hoffman, M.S. (1984). A multidimensional approach
to the relationship between mood and weather. British Journal of
Psychology,75, 15–23.
Isen, A.M. (2000). Positive affect and decision making. In M. Lewis
(Ed.), Handbook of emotion (pp. 417–434). New York: Guilford
Press.
Jaccard, J., Turrisi, R., & Wan, C. (1990). Interaction effects in multiple
regression. Newbury Park, CA: Sage Publications.
Kasper, S., Wehr, T.A., Bartko, J.J., Gaist, P.A., & Rosenthal, N.E.
(1989). Epidemiological findings of seasonal changes in mood and
behavior. A telephone survey of Montgomery County, Maryland.
Archives of General Psychiatry,46, 832–833.
Kripke, D.F. (1998). Light treatment for nonseasonal depression: Speed,
efficacy, and combined treatment. Journal of Affective Disorders,
49, 109–117.
Kruglanski, A.W., & Freund, T. (1983). The freezing and unfreezing of
lay-inferences: Effects on impressional primacy, ethnic stereo-
typing, and numerical anchoring. Journal of Experimental Social
Psychology,19, 448–468.
Lambert, G.W., Reid, C., Kaye, D.M., Jennings, G.L., & Esler, M.D.
(2002). Effects of sunlight and season on serotonin turnover in the
brain. Lancet,360, 1840–1842.
Leppamaki, S., Partonen, T., & Lonnquist, J. (2002). Bright-light ex-
posure combined with physical exercise elevates mood. Journal of
Affective Disorders,72, 139–144.
Leppamaki, S., Partonen, T., Piiroinen, P., Haukka, J., & Lonnquist, J.
(2003). Timed bright-light exposure and complaints related to
shift work among women. Scandinavian Journal of Work, Envi-
ronment, & Health,29, 22–26.
Maxwell, S.E., & Delaney, H.D. (1990). Designing experiments and
analyzing data. Belmont, CA: Wadsworth.
Michalon, M., Eskes, G.A., & Mate-Kole, C.C. (1997). Effects of light
therapy on neuropsychological function and mood in seasonal af-
fective disorder. Journal of Psychiatry & Neuroscience,22, 19–28.
O’Brien, J.T., Sahakian, B.J., & Checkley, S.A. (1993). Cognitive im-
pairments in patients with seasonal affective disorder. British
Journal of Psychiatry,163, 338–343.
Parrott, W.G., & Sabini, J. (1990). Mood and memory under natural
conditions: Evidence for mood incongruent recall. Journal of
Personality and Social Psychology,59, 321–336.
Persinger, M.A. (1980). The weather matrix and human behavior. New
York: Praeger Press.
Rosenthal, N.E., Sack, D.A., Gillin, J.C., Lewy, A.J., Goodwin, J.C., Dav-
enport, P.S., Newsome, D.A., & Wehr, T.A. (1984). Seasonal affective
disorder: A description of the syndrome and preliminary findings
with light therapy. Archives of General Psychiatry,41, 72–80.
Rotton, J., & Cohn, E.G. (2000). Violence is a curvilinear function of
temperature in Dallas: A replication. Journal of Personality and
Social Psychology,78, 1074–1081.
Russell, D.W., Kahn, J.H., Spoth, R., & Altmaier, E.M. (1998). Analyz-
ing data from experimental studies: A latent variable structural
equation modeling approach. Journal of Consulting Psychology,5,
18–29.
Russell, J.A., Weiss, A., & Mendelsohn, G.A. (1989). The affect grid: A
single-item scale of pleasure and arousal. Journal of Personality
and Social Psychology,57, 493–502.
Rusting, C.L., & Larsen, R.J. (1998). Personality and cognitive
processing of affective information. Personality and Social Psy-
chology Bulletin,24, 200–213.
Sanders, J.L., & Brizzolara, M.S. (1982). Relationships between weath-
er and mood. Journal of General Psychology,107, 155–156.
Saunders, E.M. (1993). Stock prices and Wall Street weather. American
Economic Review,83, 1337–1345.
Schwarz, N., & Clore, G.L. (1983). Mood, misattribution, and judge-
ment of well-being: Informative and directive functions of affec-
tive states. Journal of Personality and Social Psychology,45,
513–523.
Sinclair, R.C., Mark, M.M., & Clore, G.L. (1994). Mood-related per-
suasion depends on (mis)attributions. Social Cognition,12, 309–
326.
Stain-Malmgren, R., Kjellman, B.F., & A
˚berg-Wistedt, A. (1998).
Platelet serotonergic functions and light therapy in seasonal af-
fective disorder. Psychiatric Research,78, 163–172.
Watson, D. (2000). Mood and temperament. New York: Guilford Press.
Watson, D., Clark, L.A., & Tellegen, A. (1988). Development and val-
idation of brief measures of Positive and Negative Affect: The
PANAS Scales. Journal of Personality and Social Psychology,54,
1063–1070.
Wechsler, D. (1997). Manual for the Wechsler Adult Intelligence Scale
(3rd ed.). San Antonio, TX: Psychological Corp.
Woodcock, A., & Custovic, A. (1998). ABC of allergies: Avoiding ex-
posure to indoor allergens. British Medical Journal,316, 1075–
1078.
(RECEIVED 8/12/04; REVISION ACCEPTED 9/2/04;
FINAL MATERIALS RECEIVED 9/17/04)
Volume 16—Number 9 731
M.C. Keller et al.
