PSYCHOLOGICAL SCIENCE

Research Article

416

Copyright © 2003 American Psychological Society VOL. 14, NO. 5, SEPTEMBER 2003

EMOTION AND INTUITION:
Effects of Positive and Negative Mood on Implicit Judgments
of Semantic Coherence

Annette Bolte,

1

Thomas Goschke,

2

and Julius Kuhl

3

1

Institute of Psychology, Braunschweig University of Technology, Braunschweig, Germany;

2

Department of Psychology, Dresden University of Technology, Dresden, Germany; and

3

Department of Psychology,

University of Osnabrück, Osnabrück, Germany

Abstract—

We investigated effects of emotional states on the ability to
make intuitive judgments about the semantic coherence of word triads.
Participants were presented word triads, consisting of three clue
words that either were weakly associated with a common fourth con-
cept (coherent triads) or had no common associate (incoherent tri-
ads). In Experiment 1, participants in a neutral mood discriminated
coherent and incoherent triads reliably better than chance level even if
they did not consciously retrieve the solution word. In Experiment 2,
the induction of a positive mood reliably improved intuitive coherence
judgments, whereas participants in a negative mood performed at
chance level. We conclude that positive mood potentiates spread of ac-
tivation to weak or remote associates in memory, thereby improving
intuitive coherence judgments. By contrast, negative mood appears to
restrict spread of activation to close associates and dominant word

meanings, thus impairing intuitive coherence judgments.

It is a common belief that the way people perceive the world, pro-
cess information, and solve problems is influenced by emotions and
moods. Probably all of us have experienced that when we are in a de-
pressed mood, our attention appears to be focused in a rigid and nar-
row way, whereas when we are in a happy mood, our minds seem to
bubble over with ideas and sometimes far-fetched associations. Such
everyday impressions are supported by a growing body of empirical
evidence showing that positive and negative moods are in fact accom-
panied by qualitatively different information processing modes (e.g.,
Gray, 2001; Isen, 1999; Kuhl, 1983, 2000). For instance, individuals
in a positive mood, compared with those in a negative mood, produce
more unusual associations (Isen, Johnson, Mertz, & Robinson, 1985),
show improved performance on tests of creative problem solving
(Isen, Daubman, & Nowicki, 1987), prefer heuristic over exhaustive
decision-making strategies (Isen & Means, 1983), and form more in-
clusive categories in a sorting task (Isen & Daubman, 1984; see Isen,
1999, for review).
In this study, we investigated a hypothesis derived from a more
general theoretical framework termed

personality systems interaction
theory

(Kuhl, 2000; Kuhl & Kazén, 1999). According to this theory,
an increase in positive affect supports a

holistic

processing mode,
which is characterized in memory by the activation of wide semantic
fields, which include weak or remote associates. In contrast, an in-
crease in negative affect supports an

analytic

processing mode, which
is characterized by a more restricted spread of activation to close asso-
ciates and dominant word meanings. We focus on a specific prediction
of this hypothesis—that a positive mood, because of increased activa-
tion of remote associates, improves the ability to make intuitive judg-
ments about the semantic coherence of verbal stimuli, whereas a negative
mood impairs intuitive coherence judgments.
We conceive of intuition as the ability to make above-chance judg-
ments about properties of a stimulus on the basis of information that is
not consciously retrieved (cf. Bastick, 1982; Westcott, 1968). An ex-
ample would be the ability to discriminate between pieces of music
from the classical and romantic periods, even if one is unable to de-
scribe the basis of this judgment verbally. Although the concept of in-
tuition has often been considered elusive, there is now accumulating
evidence that various types of judgments are influenced by informa-
tion that is not accessible to consciousness (for reviews, see Goschke,
1997; Greenwald, 1992; Kihlstrom, 1987; Lieberman, 2000; Perrig &
Wippich, 1995; Reber, 1993; Schacter, 1987). Directly related to our
investigation is a study of Bowers, Regehr, Balthazard, and Parker
(1990), who showed that participants were able to decide at an above-
chance level whether or not word triads (e.g., “playing, credit, report”)
were coherent (i.e., had a common weak associate—“card”) even if
they did not retrieve this solution word consciously. This finding sug-
gests that activation spreads from the memory representation of the
clue words in a coherent triad to the solution concept. Even if the solu-
tion word does not gain access to consciousness, its subthreshold acti-
vation presumably leads to an intuitive perception of coherence that
guides participants’ judgments (cf. Bowden & Beeman, 1998; Med-
nick, 1962; Metcalfe & Wiebe, 1987; Yaniv & Meyer, 1987).
The main purpose of our study was to investigate effects of posi-
tive and negative mood on intuitive coherence judgments. From per-
sonality systems interaction theory (Kuhl, 2000), we derived the
prediction that an increase in positive affect, by potentiating spread of
activation to weak associates, should improve the ability to make intu-
itive coherence judgments, whereas a negative mood, by restricting
spread of activation to close associates and dominant word meanings,
should impair intuitive coherence judgments. Note that this

affect-
modulation hypothesis

is not concerned with the effects of mood on
the activation of mood-congruent memory contents (cf. Bower, 1981;
Goschke, 1996). Rather, it concerns content-unspecific effects of af-
fective states on qualitatively different (analytic vs. holistic) process-
ing modes (Kuhl, 1983, 2000; cf. Forgas, 2000; Gray, 2001). To test
this hypothesis, we had participants perform a modified version of the
coherence judgment task with emotionally neutral material either in a
neutral mood (Experiment 1) or after the induction of a positive or
negative mood (Experiment 2).

EXPERIMENT 1

In Experiment 1, we investigated whether participants in a neutral
mood are able to judge the semantic coherence of word triads above

Address correspondence to Annette Bolte, Braunschweig University of
Technology, Institute for Psychology, Spielmannstr. 19, D-38106 Braunschweig,
Germany; e-mail: annette.bolte@tu-bs.de.

PSYCHOLOGICAL SCIENCE

Annette Bolte, Thomas Goschke, and Julius Kuhl

VOL. 14, NO. 5, SEPTEMBER 2003

417

chance in the absence of conscious knowledge of the solution word.
We changed several features of the coherence task used in the original
study by Bowers et al. (1990), in order to avoid some methodological
shortcomings. Moreover, we used a more conservative criterion for
classifying coherence judgments as implicit.

Method

Participants

Participants were 24 undergraduates from the University of Osna-
brück, Germany (9 men and 15 women; mean age

!

25, range:
22–34).

Materials and apparatus

Stimuli were German translations of 72 word triads taken from the
study by Bowers et al. (1990); they had been adapted from the Remote
Associates Test (Mednick & Mednick, 1967). Half of the triads were
coherent; that is, each of the three clue words was a low associate of a
common fourth word (the solution word). The other half of the triads
were incoherent; that is, the three words were not associated with a
common solution word. Coherent and incoherent triads were matched
with respect to the semantic relatedness among the three clue words
(cf. Bowers et al., 1990, p. 76). Each participant was presented 12 co-
herent and 12 incoherent triads.

Procedure

Participants were instructed that they would have to judge whether
or not word triads were semantically coherent. It was explained that
the clue words of coherent triads were all weakly associated with a
common target word. On each trial, the three clue words were pre-
sented one above the other at the center of the computer monitor for
4 s. After the offset of the triad, participants indicated their coherence
judgment by pressing the “-” or “y” key of the computer keyboard.
After each response, participants indicated their confidence in the co-
herence judgment on a 4-point rating scale (1

!



low confidence

; 4

!

high confidence

).

1

Finally, participants were prompted to enter a solution
word via the computer keyboard. When participants were insecure or
thought that there was no solution word, they were forced to guess and
type in any word that came to mind.

Results and Discussion

Solved triads

We classified a coherent triad as solved if the participant produced
the preordained solution word or an exact synonym of the solution
word. In addition, we considered a word to be a solution if it was
judged to be semantically related to the three clue words by at least
one of two independent raters.

2

We followed this procedure to detect
trials on which participants came up with plausible, but unanticipated
solution words (e.g., when “hike,” instead of the preordained solution,
“mountain,” was given as a solution for the triad “goat, pass, green”).
Across participants, a total of 101 out of 288 coherent triads (35%)
were classified as solved; that is, participants solved on average 4.2 of
the 12 coherent triads. Of the 101 produced solution words, 37 were
preordained solutions, whereas 64 were synonyms of the preordained
solutions or words that were semantically related to the clue words. As
expected, none of the produced words for incoherent triads was classi-
fied as semantically related to the clue words.

Coherence judgments

Solved coherent triads were almost always correctly classified by
participants as coherent (see Table 1). For the unsolved triads, we
computed

hit rates

(i.e., the proportion of unsolved coherent triads
that were correctly classified as coherent) and

false alarm rates

(i.e.,
the proportion of unsolved incoherent triads that were falsely classi-
fied as coherent). Participants correctly classified 61% of the unsolved
coherent triads as coherent; this hit rate was significantly above the
chance value of 50%,

t

(23)

!

2.14,

p



"

.05. Participants falsely clas-
sified 48% of the incoherent triads as coherent; this false alarm rate
did not significantly depart from chance level,

t

(23)

!



#

0.50. As an
index for the ability to discriminate between unsolved coherent and in-
coherent triads, we computed the difference between the hit and false
alarm rates. This

intuition index

was 13% and departed significantly
from 0,

t

(23)

!

2.21,

p



"

. 04.

Response times

For unsolved triads, correct coherence judgments were made faster
than incorrect judgments, both for coherent triads (3,343 vs. 3,597 ms)
and for incoherent triads (2,598 vs. 3,268 ms). However, a 2

$

2 anal-
ysis of variance (ANOVA) with triad type (coherent vs. incoherent)
and judgment (coherent vs. incoherent) as independent variables
yielded no reliable effects.

Conclusions

Participants’ performance in judging the semantic coherence of
word triads was reliably better than chance level even when they did
not consciously retrieve the solution words. We obtained this finding
even though we took several precautions to be as conservative as possi-
ble in classifying triads as unsolved. First, in addition to preordained

1. In both experiments, average confidence for unsolved triads was rela-
tively high (2.8 and 3.0 on a 4-point scale, respectively) and differed only min-
imally between coherent and incoherent triads and between mood conditions.
2. The two raters were the first author and a second rater who was blind to
whether or not triads were coherent and to whether or not participants had
judged a triad as coherent.

Table 1.

Mean number of solved and unsolved coherent and
incoherent triads that were classified as coherent or incoherent,
Experiment 1

Judgment
Coherent triads Incoherent triads
Solved Unsolved Solved Unsolved
Coherent 3.63 (1.97) 4.42 (1.91) 0.00 5.38 (2.22)
Incoherent 0.58 (0.83) 3.00 (2.00) 0.00 5.88 (2.35)

Note

. Standard deviations are given in parentheses. Because of
occasional missing values, the sum of the individual frequencies is
slightly smaller than the expected value of 24. Incoherent triads by
definition had no solutions.