Journal of foionality and Social Psychology

1984. Vol. 47. No 6. 1206-1217

Copyright 1984 by ife

Psychological A n o r m KM. \K

The Influence of Affect on Categorization

Alice M. Isen and Kimberly A. Daubman
University of Maryland
Three studies and a pilot experiment showed that positive affect, induced in any
of three ways, influenced categorization of either of two types of stimuliwords
or colors. As reflected by performance on two types of tasks (rating and sorting),
people in whom positive affect had been induced tended to create and use
categories more inclusively than did subjects in a control condition. On one task,
they tended to group more stimuli together, and on the other task they tended to
rate more low-prototypic exemplars of a category as members of the category.
These results are interpreted in terms of an influence of affect on cognitive
organization or on processes that might influence cognitive organization. It is
suggested that borderline effects of negative affect on categorization, obtained in
two of the studies, might result from normal people's attempts to cope with
negative affect.

Recent research has suggested that positive

affect may have a pervasive effect on cognitive
processes. For example, a mild positive affective state has been shown to be capable of
serving as a retrieval cue for positive material
in memory, regardless of the affective state
the subject was in when the list was learned,
influencing such measures as the reaction
time for recall of positive words and the
subset of words likely to be recalled from a
memorized list (e.g., Isen, Shalker, Clark, &
Karp, 1978; Laird, Wagener, Halal, & Szegda,
1982; Nasby & Yando, 1982; Teasdale &
Fogarty, 1979). Moreover, a mild positive
affective state of the kind likely to be experienced in everyday life has also been shown
to influence judgments of various kinds, decision-making strategy, and willingness to take
risks (e.g., Isen & Means, 1983; Isen, Means,
Patrick, & Nowicki, 1982; Isen & Patrick,
1983; Isen & Shalker, 1982; Isen et aL, 1978;
Johnson & Tversky, 1983).
These findings suggest that affect may influence not only memory but also cognitive
organization and consequences of this organization. One series of studies, for example,
investigated the influence of affective state on
memory, judgment, and evaluation. These
results indicated that when people had been
Requests for reprints should be sent to Alice M. Isen,
who is now at the Department of Psychology, University
of Michigan, Ann Arbor, Michigan 48106.

given a small free gift, they were more likely

to judge their consumer goods more favorably;
and these improved opinions were attributed
to affect-cued accessibility of positive material,
because results of a second study indicated
that positive feelings served to cue positive
material in memory (Isen et al., 1978). In
two other series of studies it was found that
mild positive affect influenced subjects' willingness to take risks (Isen & Patrick, 1983)
and their estimates of the amount of risk
present at a given timethat is, their estimates of the chances of certain events occurring (Johnson & Tversky, 1983). There is
growing evidence, then, that positive affect
can influence cognitive processes, possibly
including cognitive organization itself, through
the ideas that are brought to mind by the
affective state.
Moreover, additional work has indicated
that not only cognitive content, but the very
strategies that subjects use in solving problems, may be influenced by the presence of
positive feelings. In one series of studies,
subjects in whom good feelings had been
induced by placing them in especially comfortable surroundings, complete with refreshments, were found to be more likely to use
an intuitive solution, or a heuristic, in solving
two different types of problems, a physics
timer-tape problem and a relative frequency
judgment (Isen et al., 1982). In addition,
subjects in whom positive affect had been

1206

1207

AFFECT AND CATEGORIZATION

induced by telling them that they had performed above average on a task of perceptualmotor skills were significantly more quick to
reach a decision in a complex choice task
(involving choice of a car for purchase from
among 6 alternatives, differing along each of
9 dimensions). They took a mean of 11 min,
whereas the control group took, on average,
19 min; and they used significantly fewer
types of information and engaged in significantly less rechecking of information than
control subjects (Isen & Means, 1983). This
finding, obtained after a different type of
affect induction, is entirely compatible with
those of the studies that found increased use
of heuristics in problem solving as a function
of positive affect (Isen et al., 1982). Together
they suggest that people who are feeling good
tend to simplify decision-making or problemsolving situations, and this too may reflect
an impact of affect on cognitive organization.
Detailed analysis of the protocols of subjects in the car-choice study supports this
suggestion and provides additional information about the influence of positive affect on
decision making. It indicates that people in
the positive-affect condition tended to be
more efficient than control subjects, not just
more impulsive or careless: They were faster
and used fewer dimensions than control subjects in reaching a decision; but the two
dimensions eliminated were those described
as least important by all subjects, and the
ultimate car choice of the two groups did not
differ. In addition, those in the positive-affect
condition, to a significantly greater extent
than controls, tended to use a strategy similar
to the one identified by Tversky (1972) as
"elimination by aspects" (Isen & Means,
1983). This strategy is thought to be one of
the most efficient of the quasi-satisficing strategies (Janis & Mann, 1977; Tversky, 1972).
It involves the use of broad categories to
eliminate large amounts of data and to quickly
narrow the field of alternatives.
One interptetation that has been offered
for these tendencies of happy persons to be
more efficient and also more distractable is
that material may be multiply encoded and
multiply retrievable as a function of happiness. That is, people in positive affective
states may have multiple cues or encoding
and retrieval devices for a given item of

material (Isen, 1984). This may occur because

positive affect, in cueing positive material,
cues a wide variety of material (e.g., Boucher
& Osgood, 1969; Bousfield, 1944; Bousfield,
1950; Isen, Johnson, Mertz, & Robinson,
1984). Thus, the cognitive context present
when a person is feeling happy may be more
complex; such a complex context may lead
to more diverse (or multiple) interpretations
or encoding and retrieval devices. The contextualist position in cognitive psychology has
suggested that the cognitive context in which
material appears may influence the meaning
or interpretation given to that material (e.g.,
Bransford, 1979). Thus, positive affect, by
influencing both the content and the complexity of the cognitive context, may influence
the interpretation and organization of cognitive material.
Thus, affective state may influence cognitive
organizationmay influence the way in which
items or units are grouped together, or the
relatedness seen in cognitive material. For
example, if material is multiply encoded and
multiply retrievable, it may seem to belong
to multiple categories or overlapping categories; or the categories may be seen as more
inclusive. The present article begins to investigate this possibility by examining the effect
of induced feelings on categorization. It presents the results of a pilot study and three
experiments employing one negative and three
positive affect inductions (affect-inducing
films, a free gift, and refreshments), two types
of categorization tasks (rating and sorting),
and two types of stimuli (words and colors).
Pilot Study
The pilot study examined the impact of
positive affect, induced by means of refreshments presented at the experimental session,
on categorization, measured by ratings given
to. weak exemplars of categories (according
to the norms identified by Rosen, 1975).
Twenty male and female students who
were enrolled in an introductory psychology
course at the University of Maryland participated in this pilot study for course credit.
Participants were randomly assigned to either
the positive affect or control condition.
Subjects were tested in groups of three to

1208

ALICE M. ISEN AND KIMBERLY A. DAUBMAN

five. In the positive affect condition, subjects

were seated in comfortable furniture arranged
around a coffee table on which refreshments
(juice and cookies) had been placed. In the
neutral condition, no refreshments were
served and subjects were seated in folding
chairs arranged in rows.
Subjects were given 2 min to get settled
(in the positive affect condition, refreshments
were distributed during this time) and then
they performed a categorization task based
on one developed by Rosch (1975). For each
of four categories, the experimenter read
aloud a category name and nine potential
exemplars (three excellent, three moderately
good, and three weak, according to Rosch's
norms). Subjects were asked to rate items on
a 10-point scale, indicating the degree to
which they felt each item belonged or did
not belong to the given general category. A
rating of 1 was to be used to mean that the
item "definitely did not belong to that category," and a 10 was to mean that the item
"definitely did belong to the category" (and
thus a rating over 5 would affirm category
membership). Each subject was presented
four categories, tool, clothing, vehicle, and
weapon, in that order, the weak exemplars of
each were "crane," "scissors," and "rags";
"cane," "ring," and "purse"; "elevator," "camel," and "feet"; and "shoes," "screwdriver,"
and "fists," respectively.
For two of the categoriesvehicle and
clothingthe results indicated that persons
in the positive-affect condition were more
likely to rate the relatively unusual exemplars
as members of the categories (for example,
they tended to give a higher rating than did
control subjects to "elevator," "camel," and
"feet" as members of the category vehicle).
At the same time, for the other two categories, there was no significant difference
between the positive-affect and control conditions. This tends to speak, against certain
alternative interpretations of the effect of the
independent variable manipulation such as
response bias or a tendency to feel more free
or more careless in categorizing words. However, it suggests that the effect may depend
on particulars of the stimulus materials as
well as on the triggering of the cognitive
processes themselves.

On balance, in part because of the small

size of the sample, however, we thought it
inappropriate to overinterpret the result of
this pilot study; and we chose on this occasion
not to explore in detail the difference among
the dimensions. Rather, we retained our major
focus on the more basic issue of whether
positive affect can influence categorization.
The results of the pilot study were taken as
encouraging of the hypothesis, but preliminary; and the next two studies, run simultaneously, were conducted in order to replicate
and extend the finding that positive affect
tended to influence the pattern of categorization of stimuli.
Studies 1 and 2
Method
Subjects One hundred sixty-two students who were
enrolled in introductory psychology courses at the University of Maryland participated in these two experiments
for course credit. Participants were assigned randomly to
one of three experimental conditions or one of two
control conditions.
Procedure. Students participated in groups ranging
in size from one to six. In Study 1, they were assigned
to either a positive-affect condition (Free Gift, in which
participants received candy bars as an expression of
thanks for their participation) or a no-manipulation
control group; in Study 2, affect was manipulated by
means of films, and subjects were asked to view a 5-min
segment of "Gag Reel," a comedy film (positive-affect
treatment), "Area Under a Curve," a mathematics film
(control group), or "Night and Fog," a documentary of
Nazi concentration camps (negative-affect treatment).
Following the manipulation, as a check on affect
induction, a portion of the participants m each condition
of Studies 1, 2, and 3 (which follows) rated their feelings
on each of two 7-point scales, positive-negative and
amused-sober, with 1 indicative of the positive end of
the scale, and 7, the negative. Two other scales, alertunaware and refreshed-tired, were included as filler items
(and in the case of alert-unaware, to obtain some indication of more general, global arousal).
Next, participants performed the categorization task
described for the pilot study. Instructions for the task
explained to the participants that they were going to be
asked to rate items on a 10-point scale, indicating the
degree to which they felt each item belonged or did not
belong to a given general category. In addition to 1 and
10 (the endpoints of the scale), participants were instructed
about the break-point between 5 and 6. They were told
to use a 5 to mean that the item did not belong to the
category but was very similar to members of that category,
and a 6 to indicate that the item did belong to the
category but was not a very good example of i t As in
the pilot study, the experimenter then read a category
name and nine items (three excellent exemplars, three

1209

AFFECT AND CATEGORIZATION

Table 1

Mean Affect Rating and Variance on Each of Four Affect Scales in Each Condition
No

Positive
film
n = 57

Control
film
n=36

Negative
film
n = 20

Free
gift
n = 41

manipulation
n = 17

2.74
1.21

3.26
1.39

4.70
2.64

2.62
1.57

3.65
2.02

2.68
1.24

3.94
1.79

5.15
1.40

3.61
1.73

3.32
1.78

3.65
2.88

4.45
1.18

4.49
3.48

3.67
2.53

3.59
1.55

2.64
1.24

3.03
1.70

2.91
2.05

2.61
1.28

3.32
1.81

Scale*
Positive-negative
o2
Amused-sober
o2
Refreshed-tired
Alert-unaware

M
<r
1

7-point scales, with 1 as the positive pole.

moderately good exemplars, and three weak exemplars,

according to Rosen's, 1975, norms). The first item to be
rated was always an excellent exemplar of the category;
the remaining items were presented in random order.
This was done with four categories, presented in the
following order, furniture, vehicle, vegetable, and clothing.
The weak exemplars in these categories were stove, fan,
and telephone, camel, feet, and elevator, pickles, seaweed,
and rice, and purse, ring, and cane. Participants indicated
their ratings individually on a sheet of notebook paper.

Results

Manipulation check. Table 1 presents participants' mean ratings of their feelings following the manipulation of affect. Ratings on
the positive-negative and amused-sober scales
constituted the affect manipulation check
(both of these scales, for the film-induced
negative and positive affect states; only positive-negative for the free-gift induced affect
state). To test whether the intended states had
been induced by the film manipulations, a
two-way, mixed-design analysis of variance
(ANOVA) was performed, with three levels of
affect (positive, neutral, and negative) and
two levels of affect-scale (positive-negative
and amused-sober); this revealed a significant
main effect of affect, ^ 2 , 110) = 40.42, p =
.0001. Planned comparisons showed that participants in the "Gag Reel" condition rated
their feelings as significantly more positive,
t(9l) = 2.17, p < .025, and more amused,
?(91) = 4.85, p < .001, than participants who
had seen the control film. Conversely, partic-

ipants in the "Night and Fog" as compared

to the "Area Under a Curve" condition saw
themselves as significantly more negative,
1(54) = 3.51, p < .001, and more sober,
f(54) = 3.46, p < .001. For the free-gift affect
manipulation, the positive-negative scale was
the intended affect manipulation check; thus
a one-way ANOVA comparing these two conditions on this scale was performed. A significant difference was obtained, JFX 1, 56) =
7.42, p = .009, withfree-giftparticipants rating their feelings as more positive than did
the control group. The differences in these
ratings suggest that the appropriate affect was
successfully induced in each case.
Categorization. Our hypothesis was that
positive affect would influence categorization
such that items not normally considered part
of a category would either (a) be seen as
category members, or (b) be seen as more
similar to category members. (No hypothesis
was offered for a possible effect of negative
affect, but negative affect was included in
order to extend perspective on the effects
observed). Thus, our analysis centers on the
ratings of those items not usually considered
category members.
To test the first part of this hypothesis, we
compared the mean number of poor exemplars rated 6 or above (i.e., rated as category
members in each condition). (See Table 2.)
Second, it seemed possible and quite consistent with our reasoning that positive-affect

1210

ALICE M. ISEN AND KIMBERLY A. DAUBMAN

Table 2
Mean Number of Poor Exemplars Judged as Category Members, Variance, and Number of
Participants in Each Condition
Study 2

Study 1
Measures

Free gift

No
manipulation

"Area under
a curve"

"Gag reel"

"Night
and Fog"

4.10
8.33
42

2.86
3.76
28

3.09
3.54
34

4.61
10.24
36

4.05
5.65
21

a2

subjects might not judge poor exemplars

quite as category members (a rating of 6 or
higher) but might be more likely to judge
these items as more similar to category members than would subjects in the control condition. Thus, we also examined subjects'
mean ratings of the 12 poor exemplars (see
Table 3).
A one-way ANOVA on the first dependent measure revealed significant differences
among affect conditions, F(4, 156) = 2.67,
p = .03. Planned comparisons indicated the
nature of these differences more specifically:
Subjects in the free-gift condition rated more
poor exemplars as category members than
did no-manipulation controls, f(68) = 1.98,
p < .05; and subjects in the comedy-film condition rated more poor exemplars as category
members than did subjects viewing the control
film, f(68) = 2.48, p < .01. There was a tendency for subjects viewing the negative film,
in comparison with neutral-film subjects, also
to identify more poor exemplars as category

members; but this trend did not reach customarily acceptable levels of significance,
/(53) = 1.34, .1 <p<2.
Effects on the second dependent variable
followed the same pattern. The overall analysis
indicated a significant difference among conditions, ^ 4 , 156) = 2.45, p < .05; subjects
in the free-gift condition rated the poor exemplars higher (that is, as tending more to
be members of the categories) than did nomanipulation controls, /(68) = 2.08, p < .025;
comedy-film subjects rated them higher than
did control-film subjects f(68) = 1.84, p<
.05; and there was a nonsignificant tendency
for negative-film subjects to rate the poor
exemplars higher than did control-film subjects, /(53) = 1.46, .1 < p < .2.
Although we were not primarily concerned
with the impact of affect on the ratings of
moderate and good exemplars (we had included them as filler items and had offered
no hypothesis concerning these data), we felt
our readers might be interested in this infor-

Table 3
Mean Ratings of Poor, Moderate, and Good Exemplars, Variance, and Number
of Participants in Each Condition
Study 1

Study 2

Free gift
n = 42

No
manipulation
n = 28

4.28
2.52

Goodness of
exemplar

"Area under
a curve"
n = 34

"Gag reel"
n = 36

"Night
and Fog"
n = 21

3.56
1.33

3.90
1.32

4.52
2.65

4.47
1.72

7.71
1.29

7.51

7.34

.93

.97

7.95
1.10

7.74
.76

9.74
.14

9.68

9.46

.27

.33

9.80
.11

9.77
.17

Poor
a2
Moderate
a2
Good

M
a3

1211

AFFECT AND CATEGORIZATION

mation, and therefore it is presented in Table

3. We did not analyze these data in one
overall ANOVA, along with the data on the
weak exemplars, for two reasons: First, responses on the moderate and strong goodnessof-exemplar items were not related conceptually to our hypothesis regarding the impact
of affect on cognitive organization or categorization. Second, as can be seen from Table
3, the variances of these data were so different
from those of the weak-exemplar results, that
if all of the data were treated as one single
set for analysis, that set would have had
markedly unequal cell variances and would
not have met the homoscedasticity requirement of the ANOVA. (Moreover, one excellent
exemplar was always presented first in the
list of items to be rated, so that the order of
presentation might have influenced results on
these items to some degree.) Thus, we analyzed the data of the different strength-ofexemplar levels separately.
To test for effects of induced feeling states
on moderate and good exemplars, two oneway ANOVAS were performed; the analysis of
the mean rating of the moderate exemplars
indicated no significant differences among
the 5 groups, F\4, 156) = 1.77, p = .14. The
overall analysis of the rating of excellent
exemplars revealed a significant main effect,
F\4, 156) = 3.03, p = .02. Post hoc Scheffe
tests designed to locate the source of the
effect indicated that only the difference between the comedy and control films of Study
2 was significant, p < .05.
The findings of Studies 1 and 2 as well as
of the pilot study suggest that a person in a
positive-affect state (induced in any of three
ways) tends to view (rate) fringe exemplars
of categories more as members of the categories or as similar to category members than
do persons in comparable control groups. We
interpret these results as indicative of a change
in cognitive organization or perceived relations among stimuli under conditions of positive affect.
We did not expect the moderate and strong
exemplars to show the same effect as the
weak, because these items presumably would
already tend to be rated as members of their
categories, regardless of the affective state of
the observer (i.e., even by control subjects).
Moreover, because such items are more pro-

totypic of their respective categories, rating

them as members of the category would not
require seeing them in new ways or seeing
additional aspects of them, or seeing potential
relations between them and other items. (This
is the effect that we expected of positive
affect.) As noted, these items were included
only as filler material, to provide a normal
context for the task. However, results of our
analyses revealed a significant influence of
affect condition on the rating of good exemplars (although not of the moderate exemplars) in Study 2 (which used films to induce
affect), though not in simultaneously conducted Study 1 (which used a free gift to
induce affect). Because we were not able to
include all goodness-of-exemplar levels in
one analysis, it was not possible to test the
interaction between affect and strength of
exemplar. Thus, it remains possible that the
results for the strong and moderate exemplars
are not significantly different from those of
the weak exemplars.
One possible interpretation of this difference between affect conditions on the good
exemplars (and of the possibility that affect
influenced all types of exemplars equally)
that might be proposed is that affect produced
a response bias or tendency to prefer the
upper end of the rating scale. This observation
may invite some to interpret all of our findings
in terms of a response bias rather than of a
more meaningful influence of affect. We do
not believe this to be the correct interpretation
of our results, in part because postive-affect
subjects used the lower end of the rating scale
to indicate their more positive affect on the
manipulation check. Moreover, the apparent
influence of affect on the filler items may not
be reliable, because it was obtained only in
one study and only with good (but not moderate) exemplars. However, we address the
response-bias issue more directly in Study 3,
by employing a different taska sorting task
rather than the rating task used in the first
three studies (including the pilot study). Study
3 also introduces different stimuli, colors, in
order to broaden the investigation further.
Study 3
Method
Subjects Seventy-four male and female students enrolled in an introductory psychology course at the Uni-

1212

ALICE M. ISEN AND KIMBERLY A. DAUBMAN

versity of Maryland participated in this experiment

Participants were randomly assigned to one of three
experimental conditions or a control condition, "Gag
Reel," "Free Gift," "Night and Fog," or "Area Under a
Curve," described in Studies 1 and 2. This procedure
resulted in two groups of 18 subjects, one group of 22,
and one group of 16.
Apparatus Sets of 14 two-inch-square color chips of
different colors, taken from the Munsell Color System,
were used as the stimuli to be categorized for this
experiment. The colors selected for inclusion in the
stimulus sets (with values of 10R, 5YR, 10YR, 5Y, 10Y,
5GY, 10GY, 5G, 10G, 5BG, 10BG, 5B, 10B, and 5PB)
form a psychophysically equal-interval continuum spanning 7/10 of the color wheel, from red to purple-blue, at
Saturation 6.
Procedure. Subjects reported for participation in
groups ranging from one to five. After the affect-inducing
or control manipulation, participants were seated at
isolated tables so that they were unable to see each other.
A set of the 14 color chips was placed in front of each
subject in a prearranged order so that the continuum of
color was not immediately apparent. The participants
were told to place the colors into categories making as
few or as many categories as they wanted and putting as
few or as many colors in each category as they wanted.
We expected that persons who were feeling good would
create fewer categories, grouping more colors together.
We were still not prepared to make a prediction for the
negative-affect condition.

Results
Table 4 presents the mean number of
categories created in each of the four conditions of this study. A one-way ANOVA on
these data, with four levels of the affect
variable (two positive, one neutral, one negative) indicated significant differences among
the groups, i=l(3, 70) = 3.01, p = .04. As
expected, the two positive-affect conditions
did not differ from each other, thus, they
were combined and compared to the neutral
condition in testing our hypothesis. This procedure revealed a significant difference in the
expected direction, r(54) = 2.72, p < .005. A
comparison between the negative and control
conditions indicated a tendency toward a
difference in the same direction,fl(32)= 2.01,
p < .1, two-tailed.
Discussion
Positive affect. The findings of these four
studies, using two different tasks and three
positive-affect inductions, provide evidence
for the proposal that positive affect can influence the way in which cognitive material is
categorized or grouped together. More specif-

Table 4
Study 3: Mean Number of Categories Created,
Variance, and Number of Participants
in Each Condition
Mea- Free gift "Gag reel"
sures n = 18
n = 22
M

a2

3.72
.80

3.55
2.74

"Area under
a curve"
n = 16

"Night and
Fog"
n = 18

4.88
2.78

3.83
1.79

ically, they show that weak exemplars are

more likely to be rated as category members,
and that colors' are sorted into more inclusive
categories, by persons who are feeling happy.
Thus, these findings also lend support to the
more general hypothesis that under conditions
of positive affect, people tend to see relatedness and interconnections among cognitions,
and perhaps process material in a more integrated fashion (Isen, 1984; Isen, Daubman,
& Gorgoglione, in press).2
The process by which positive affect influences categorization, or integration and perception of relatedness, is still unclear. Three
possibilities have been suggested, and these
may not be mutually exclusive (Isen et al., in
press). One is that positive affect may prime
an affective dimension of material that is
normally not seen as affective, and in this
way it may serve to change and broaden the
relations perceived. For example, under conditions of positive affect, neutral concepts
such as "tower" or "spark" may be seen in
added, figurative ways; and thus their meanings and their ranges of associates may be
broadened. Support for this possibility comes
from the fact that positive affect has been
shown to serve as a retrieval cue for other
1
Recently, similar findings have also been obtained
using animal names and trait words.
2
Another possibility is to view this effect as one of
"overinclusiveness" rather than "integration," thus putting
a somewhat negative (undesirable) cast on it rather than
a positive one. Results of other studies suggest that there
are at least some desirable outcomes of such processes,
such as improved efficiency or unproved creative problemsolving ability (Isen et al., in press; Isen & Means, 1983;
Isen et al., 1982). However, there can be undesirable
effects as well (see, for example, Isen et al., 1982), perhaps
under specific circumstances; and such possibilities need
to be explored.

AFFECT AND CATEGORIZATION

positive material in memory (e.g., Isen et al.,

1978). It would be difficult to account for the
findings of all of the present studies by reference to this concept alone, however, because
it is difficult to understand how the grouping
of color chips might come to be affected by
acquired or primed affective components.
However, a related idea emerges from the
suggestion that positive aspects of stimuli
may be cued by the positive affective state of
the perceiven The affective tone or impact of
the stimuli may become more positive (e.g.,
Isen & Shalker, 1982; Isen et al., 1978; Schiffenbauer, 1974), and this more positive tone
may serve to link objects together because it
is shared by them. That is, stimuli may be
grouped or seen as similar because of the
similar affective impact that they have on us.
Tversky and Gati (1978) have pointed out
that we view things that behave the same way
as similar (p. 98). Likewise, if things affect
us in the same way, we may see them as
similar in some sense.
A second possible mechanism by which
positive affect may result in the organization
of material into larger units, which we have
observed, may revolve around the fact that
positive affect has been shown to promote
the use of heuristics or reliance on information that comes to mind most easily (Isen et
al., 1982). In the task involving rating of
weak exemplars, it may be that the presence
of good and moderately good exemplars of a
category, or even just the task of dealing with
the category, results in a bias toward use of
that category; and persons in a positive state
may be expected to be more subject than
others to the effects of such biases. There is
some difficulty in using this interpretation to
account for the pattern of color sorting among
those in the positive-affect condition. It can
be done if one assumes that a grouping is
facilitated when it is begun and therefore
becomes more likely to continue to be used
by those in a positive affective state, but this
is somewhat awkward.
Thus, the tendency to use heuristics is not
an entirely satisfying interpretation of the
present results. In addition, it is not sufficient
to account for all of the relevant data, because
it implies that persons who are feeling good
might be expected to be more repetitive,
typical, unoriginal, or unimaginative in their

1213

responding; and this suggestion is inconsistent

with the results of other studies, which indicate that increased creativity and greater integration of disparate elements are also associated with positive affect (Isen et al., in
press; Isen et al. 1984). However, the tendency
to use heuristics still may be a contributor to
the overall process. Moreover, if the tendency
to use heuristics does play a role in the
observed effect on categorization this process
does not preclude organizational changes and,
in fact, may even mediate them.
A third possibility stems from the fact that
under conditions of positive affect, more material and more diverse material comes to
mind (Isen et al., in press). Context theory
in cognitive psychology has suggested that
cognitive context (as distinct from stimulus
context) can play a role in the meaning or
interpretation given a stimulus (e.g., Bransford, 1979); and recently Medin and his
associates have discussed how such processes
might influence the way in which cognitive
material is categorized or combined with
other cognitive cues or components (e.g.,
Medin, 1983). Thus, if positive affect serves
to cue a large variety and amount of material,
and this material creates a complex, extended
context for cognitive activity, then positive
affect may be setting a broad context that
influences categorization and organization of
material. As suggested earlier, material may
be multiply encoded and/or retrievable in
such a broad context (Isen, 1984). Thus, the
contiguity of diverse material, or the sheer
volume of material present under conditions
of positive affect, may result in the use of
larger or more integrated categories for processing information (Isen et al., in press); or
categories may be seen as overlapping. Perhaps
material is chunked differently under conditions of positive affect because there is a lot
of it to be dealt with; or perhaps the presence
of diverse cues impacts the meaning of the
material, simply or conngurally, resulting in
many more groupings and associations than
at another time. (Support for this suggestion
has recently been obtained in a series of
studies, Isen et al., 1984, reporting more
diverse word associations as a function of
positive affect.) Perhaps, also, this possibility
is related to, or even in part responsible for,
the increased use of heuristics that has been

1214

ALICE M. ISEN AND KIMBERLY A. DAUBMAN

discussed. These and other possible mechanisms underlying the influence of positive
affect on cognitive organization need to be
explored in more detail.
Note, however, that these studies have not
addressed cognitive organization directly. It
may be, as proposed here, that positive affect
does result in changes in underlying organization or structure; however, it may be that
affect does not influence underlying structure
or relation between cognitive elements, but
only affects certain processes that may be
responsible for the effects that we have observed. Some obvious possibilities, such as
an affect-induced response bias in use of the
rating scales, seem unlikely, in view of the
full pattern of data obtained and the fact that
multiple methods of measurement have been
employed. However, there are other influences
on responding that might be considered. We
have already mentioned as one possibility the
tendency to rely on heuristics, for example.
Thus, whereas this work on categorization is
compatible with an interpretation in terms
of changes in cognitive organization and suggests that affect may influence organization,
other interpretations remain possible. Because
of this, studies using a priming paradigm
(which is a widely accepted means of studying
cognitive organizational structure) and investigating priming patterns among those in
various affective states might be useful at this
point to address more directly the question
of cognitive organization. If positive affect
influences cognitive organization in the way
hypothesized, then material that is not normally primed by a given stimulus may indeed
be facilitated by that stimulus during positive
affect. Moreover, the details of the ways in
which these patterns change may provide
more specific information about the nature
of the change in cognitive organization that
occurs.
Negative affect. An unexpected finding,
of borderline significance, in two of the four
studies reported in this article was that negative affect appeared to influence the categorization process in the same way that positive
affect did: People in the negative-affect condition tended, though not significantly so, to
give higher ratings to poor exemplars of a
category than did those in the control group;
and they also tended to group more colors
together. Although these findings did not

reach customarily accepted levels of significance, and although negative affect was induced in only one way, some discussion of
this result may be warranted, because it was
observed in two sequential studies using diverse methods of measurement.
These results cannot easily be intrepreted
in terms of the hypothesis regarding the more
basic impact of positive affect on cognitive
organizationthat is, increased breadth of
focus, integration, or chunking of cognitive
material. Even if negative material is cued
by negative affect (a finding that has sometimes been obtained but more frequently
not), such material is not as extensive as
positive material and would not be expected
to require or promote increased chunking
and broader organization. In fact, evidence
regarding at least some forms of negative
affect suggests that the range of attention
becomes more restricted rather than more
inclusive (e.g., Bruner, Matter, & Papanek,
1955; Easterbrook, 1959). Moreover, results
of other studies investigating the aspects of
the current formulation that bear on creative
integration and cue use fail to find evidence
that negative affect facilitates those processes
(Isen et al., in press). Thus, perhaps the
borderline effects of negative affect observed
in the present studies, to the extent that they
are reliable, are reflective of a different process
from that underlying the effect of positive
affect.
It has been proposed that negative affect
may initiate a process of affect repair wherein,
it is hypothesized, normal people who are
feeling bad may engage in processes designed
to make themselves feel better (e.g., Cialdini,
Darby, & Vincent, 1973; Isen, 1984; Isen et
al., 1978). Such a process, which is consistent
with the self-regulatory models of social
learning theory (Bandura, 1973, 1977; Mischel, 1973), has been put forth to explain
instances of asymmetry in the cognitive and
behavioral effects of positive and negative
affect, in dependent measures ranging from
memory to altruism. (See Isen, 1984, for a
more complete discussion of the literature
relating to this point.)
The process of affect-repair following negative-affect induction might contribute to the
effect that we observed in either of two ways:
First, this process is presumed to be primarily
a controlled or effortful process; it therefore

AFFECT AND CATEGORIZATION

would be expected to require processing capacity and might also thus result in the use
of simplifying strategies such as heuristics.
To date, there has not been investigation of
the influence of negative affect on use of
heuristics as there has been of positive; and
the possibility remains that negative affect,
like positive, though possibly for different
reasons, increases the tendency to use heuristics. If so, then it may be for this reason
that processing capacity demands are increased because of efforts at affect repair, and
this increased demand results in a tendency
to use heuristicsthat negative-affect subjects
may tend to categorize more broadly. As
noted above, however, by itself this interpretation in terms of use of heuristics seems
insufficient to account for all of the relevant
data.
The second way in which the process of
affect-repair might result in the findings that
we obtained is that people might focus on
positive material in order to try to make
themselves feel better. To quote the words to
a popular song, in illustration, "When the
dog bites, when the bee stings, when I'm
feeling sad, I simply remember my favorite
things, and then I don't feel so bad" (Rodgers
& Hammerstein, 1959, p. 27). Thus, after
negative-affect induction, but after the passage
of a little bit of time and the expenditure of
at least a little bit of effort, people may end
up thinking about positive things; and, at
least to some extent and in some situations
(e.g., where the tasks were not cognitively
demanding), this might produce cognitive
effects similar to those occurring in responses
to positive-affect inducers.
This suggestion may lead one to propose
that positive and negative affect will frequently
or generally have the same effects, or will do
so eventually, given time and energy for strategy deployment.3 However, we are not suggesting such a global interpretation. We do
not yet know enough about how or how
successfully such mood-repair strategies work.
For example, they may produce only reduction in unhappiness rather than increased
happiness; and they may not work equally
well in all situations. There is already reason
to believe that these qualifications on the
efficacy of affect repair exist. Nonetheless, in
the current instance, it is possible that the
similar, rather than symmetrical, effects of

1215

positive and negative affect are attributable

to processes of affect improvement or efforts
at affect improvement, on the part of those
who are feeling bad.
Finally, responses to positive and negative
affect may be subtly but importantly distinct,
even when the reactions appear similar on
the surface. For example, in the area of word
association, it has been found that positive
affect and positive words as stimuli can each
lead to an increased number of unusual first
associates (Isen et al., 1984); it has also been
reported that negative words as stimuli sometimes produce unusual associates (Cramer,
1968). However, it is assumed that the reasons
for this response may be different in the two
cases, because other aspects of word-association responses to positive and negative affect
differ. The pool of additional responses to
negative words is small, whereas positive affect
results in a greater number of associates
overall; similarly, reaction time to negative
words is slower, whereas associates to positive
words are supplied more quickly (Cramer,
1968). Thus, in the present studies as well,
although responses to negative and positive
affect may appear similar in some respects,
there may be meaningful differences between
them.
In view of the fact that negative affect was
induced in only one way, and the fact that
the impact of this affect induction was of
only borderline significance, however, it may
be premature to speculate further about how
negative affect influences responding on tasks
such as these. It is possible that the effect
observed with negative affect is not a reliable
one; and it is also possible that any such
effect might be attributable to aspects of the
manipulation other than its negative-affect
inducing quality. On reflection, we realize
that "Night and Fog" differs from our other
manipulations in several ways besides its va-

3
Various homeostatic theories of affect have also predicted similarity of response to positive and negative
affect, but for a somewhat different reason. The opponentprocess theory described by Solomon (1980), an interesting
variant that has been applied to various phenomena
involving acquired motivation, postulates that both positive and negative affective processes initiate their own
neutralization. It is not clear, however, whether positive
and negative affective processes would have exactly the
same parameters.

1216

ALICE M. ISEN AND KIMBERLY A. DAUBMAN

lence: For example, (a) the affect that it

generates may be more intense, and (b) it
focuses on a specific theme, about which
subjects may have their own thoughts and
associations (it may thereby direct a subject's
train of thought or attention in a way that
the other films do not). Thus, we suggest that
it may not be appropriate to base conclusions
about the impact of negative affect generally
on results obtained with only this one means
of affect induction; and we will resist the
temptation to speculate further on how to
account for the borderline effects of the negative film observed in Studies 2 and 3.
Arousal. Finally, in part because the effects
of positive and negative affect tended to be
similar rather than opposite, we will address
the questions of (a) whether the categorization
effects that we observed can be attributed to
a more general stimulation or arousal, rather
than to the affective states that we hypothesize,
and (b) whether the findings may represent
an effect on the control group rather than on
the affect groups. In Study 2 and again in
Study 3, we compared the behavior of subjects
in the two affect-induction conditions ("Gag
Reel" and "Night and Fog") to that of subjects
exposed to a treatment that was designed to
constitute a control condition ("Area Under
a Curve"). However, the crucial factor might
have been, not that affect was induced, but
that both of the affect-inducing treatments
were relatively stimulating or interesting,
whereas the control condition might have
been boring. Thus, one might argue that the
lowered cognitive stimulation produced by
the control film might have narrowed categorization, not the other way around. It is
for this reason that we included the conditions
of Study 1a no-manipulation control and
positive-affect induction by means other than
filmand analyzed allfiveconditions together
(but did not combine the two positive-affect
groups nor the two control groups for analysis). There was no difference between these
two control conditions in either ratings of the
weak exemplars or ratings of subjects' own
affective states following experimental treatment. The comparable effects in this pair of
conditions suggest that the obtained effect on
categorization is not attributable to any unintentional boredom or negative affect produced by the film that was designed to serve
as a control for affect induction.
The issue of arousal is partly answered by

the material in the preceding paragraph, but

it still requires some additional clarification.
One might argue that, independent of the
matter of the neutrality of the control groups,
the affect-inducing manipulations may have
induced arousal and may have had their
effects through arousal rather than affect It
is hard to see how this effect might occur
without affect, but perhaps something like
drive level might be postulated as the crucial
factor in producing our results. Thus, we will
present some evidence that may bear on this
issue.
We mentioned when we discussed the manipulation check that it included four scales,
positive-negative, amused-sober, refreshedtired, and alert-unaware. For purposes of the
manipulation check, we presented only the
results on the relevant scales, positive-negative
and amused-sober. However, these ratings
also provide at least suggestive information
bearing on the question of whether differences
in arousal level might account for our observed categorization results (see Table 1).
It was the dimension of alert-unaware that
was originally included in the list in order to
tap a possible effect of the manipulation on
arousal (see also Isen & Gorgoglione, 1983);
and on this scale there were no differences
between conditions. On the dimension refreshed-tired, which some might consider an
indicator of arousal, the two control conditions did differ, with those in the film condition ("Area") reporting more relative tiredness. Subjects in the comedy film condition
reported feeling more refreshed than those in
the film-control group (although not in comparison with subjects in the other control
group), but subjects in the free-gift condition
did not differ from the no-manipulation control group. Likewise, responses of subjects in
the negative-affect condition did not differ
from those of persons in the film-control
condition. Thus, if responses on either of
these two scales were taken to reflect general
arousal, or physical energy level, the results
of our studies could not be attributed to
general arousal, because the pattern of responding on the scales simply does not fit the
pattern of categorization data.
Finally, in this same context of arousal
conceptualized as drive, it might be noted
that a Hullian formulation, which suggests
that behavior is a function of Habit Strength X
Drive, would not account for the pattern of

AFFECT AND CATEGORIZATION

results that we obtained. Rather, that formulation implies that increased arousal
should make the dominant response more
likely; and thus it would predict that affect,
if it constitutes arousal, should lead to more
usual responding or more typical categorization responses.
In conclusion, the findings of these studies
suggest that categorization is influenced by
positive affect and may be influenced by
negative affect as well. In combination with
other data showing effects of positive affect
on word association, memory, and creative
problem solving, these data support our hypothesis that positive affect is associated with
cognitive reorganization such that more relations among concepts or ideas are seen than
is the case under neutral-affect conditions.
This phenomenon seems worthy of further
investigation.
References
Bandura, A. (1973). Aggression. A social learning analysis
Englewood Cliffs, NJ: Prentice-Hall.
Bandura, A. (1977). Self-efficacy: Toward a unifying
theory of behavioral change. Psychological Review, 84.
191-215.
Boucher, J., & Osgood, C. E. (1969). The pollyanna
hypothesis. Journal of Verbal Learning and Verbal
Behavior, 8, 1-8.
Bousfield, W. A. (1944). An empirical study of the
production of affectively toned hems. Journal ofGeneral
Psychology, 30. 205-215.
Bousfield, W. A. (1950). The relationship between mood
and the production of affectively toned associates. The
Journal of General Psychology, 42, 67-85.
Bransford, J. D. (1979). Human cognition Belmont, CA:
Wadsworth.
Bruner, J. S., Matter, J., & Papanek, M. L. (1955).
Breadth of learning as a function of drive-level and
mechanization. Psychological Review, 62, 1-10.
Cialdini, R. B., Darby, B., & Vincent, J. (1973).
Transgression and altruism: A case of hedonism. Journal of Experimental Social Psychology, 9, 502-516.
Cramer, P. (1968). Word association. New York: Academic
Press.
Easterbrook, J. A. (1959). The effect of emotion on cue
utilization and the organization of behavior. Psychological Review, 66, 183-201.
Isen, A. M. (1984). Toward understanding the role of
affect in cognition. In R. Wyer & T. Srull (Eds.),
Handbook of social cognition (pp. 174-236). Hillsdale,
NJ: Erlbaum.
Isen, A. M., Daubman, K. A., & Gorgoglione, J. M. (in
press). The influence of positive affect on cognitive
organization. In R. Snow & M. Fair (Eds.), Aptitude,
learning and instruction III: Conative and affective
factors. Hillsdale, NJ: Erlbaum.
Isen, A. M.r & Gorgoglione, J. M. (1983). Some specific
effects of four affect-induction procedures. Personality
and Social Psychology Bulletin, 9, 136-143.
Isen, A. M., Johnson, M. M. S., Mertz, E., & Robinson,

1217

G. F. (1984). The influence of positive affect on the

unusualness of word associations Unpublished manuscript. University of Maryland.
Isen, A. M., & Means, B. (1983). The influence of positive
affect on decision making strategy. Social Cognition.
2, 18-31.
Isen, A. M., Means, B., Patrick, R., & Nowicki, G.
(1982). Some factors influencing decision-making
strategy and nsk-taking. In M. S. Clark & S. T. Fiske
(Eds.), Affect and Cognition- The 17th Annual Carnegie
Symposium on Cognition (pp. 243-261). Hillsdale,
NJ: Erlbaum.
Isen, A. M., & Patrick, R. (1983). The effect of positive
feelings on risk-taking: When the chips are down.
Organizational Behavior and Human Performance, 31,
194-202.
Isen, A. M., & Shalker, T. E. (1982). Do you "Accentuate
the positive, eliminate the negative" when you are in
a good mood? Social Psychology Quarterly, 45, 58-63.
Isen, A. M., Shalker, T, Clark, M., & Karp. L. (1978).
Affect, accessibility of material in memory and behavior
A cognitive loop? Journal of Personality and Social
Psychology. 36, 1-12.
Janis, I. L., & Mann, L. (1977). Decision making. New
York: The Free Press.
Johnson, E., & Tversky, A. (1983). Affect, generalization,
and the perception of risk. Journal of Personality and
Social Psychology. 45. 20-31.
Laird, J. D., Wagener, J. J., Halal, M., & Szegda, M.
(1982). Remembering what you feel: The effects of
emotion on memory. Journal of Personality and Social
Psychology. 42. 646-657.
Medin, D. L. (1983). Structural principles tn categorization. In T. Tighe, B. Shepp, & H. Pick (Eds.),
Interactions Perception, cognition, and development.
Hillsdale, NJ: Erlbaum.
Mischel, W. (1973). Toward a cognitive social learning
reconceptualization of personality. Psychological Review.
80. 252-283.
Nasby, W., & Yando, R. (1982). Selective encoding and
retrieval of affectively valent information. Journal of
Personality and Social Psychology 43, 1244-1253.
Rodgers, R., & Hammerstein, Q, II. (1959). The sound
of music. New York: Williamson Music.
Rosch, E. (1975). Cognitive representations of semantic
categories. Journal of Experimental Psychology General, 1040), 192-233.
Schiffenbauer, A. (1974). Effects of observer's emotional
state on judgments of the emotional state of others.
Journal of Personality and Social Psychology, 30(1),
31-35.
Solomon, R. L. (1980). The opponent-process theory of
acquired motivation. American Psychologist, 35, 691712.
Teasdale, J. D., & Fogarty, S. J. (1979). Differential
effects of induced mood on retrieval of pleasant and
unpleasant events from episodic memory. Journal of
Abnormal Psychology, 88, 248-257.
Tversky, A. (1972). Elimination by aspects. Psychological
Review. 79. 281-299.
Tversky, A., & Gati, I. (1978). Studies of similarity. In
E. Rosch & E. Lloyd (Eds.), Cognition and categorization (pp. 73-98). New York: Academic Press.

Received September 6, 1983

Revision received July 24, 1984