Integration of Impulsivity and Positive Mood to Predict Risky Behavio, mgradam
[ Pobierz całość w formacie PDF ]
2007, Vol. 19, No. 1, 107–118
Copyright 2007 by the American Psychological Association
Integration of Impulsivity and Positive Mood to Predict Risky Behavior:
Development and Validation of a Measure of Positive Urgency
Melissa A. Cyders, Gregory T. Smith, and
Nichea S. Spillane
University of Kentucky
Sarah Fischer
University of Chicago Medical Center
Agnes M. Annus
University of Kentucky
Claire Peterson
University of Dayton
In 3 studies, the authors developed and began to validate a measure of the propensity to act rashly in
response to positive affective states (
positive urgency
). In Study 1, they developed a content-valid
14-item scale, showed that the measure was unidimensional, and showed that positive urgency was
distinct from impulsivity-like constructs identified in 2 models of impulsive behavior. In Study 2, they
showed that positive urgency explained variance in risky behavior not explained by measures of other
impulsivity-like constructs, differentially explained positive mood-based risky behavior, differentiated
individuals at risk for problem gambling from those not at risk, and interacted with drinking motives and
expectancies as predicted to explain problem drinking behavior. In Study 3, they confirmed the
hypothesis that positive urgency differentiated alcoholics from both eating-disordered and control
individuals.
Keywords:
impulsivity, positive mood, risky behavior, urgency, validation
The aim of this series of studies was to test the possibility that
there is an impulsivity-like construct that involves the tendency to
act rashly or maladaptively in response to positive mood states
(
positive urgency
) and that individual differences on this trait help
explain risky behavior. This possibility stems from the following
concerns. First, there is evidence that the term
impulsivity
actually
subsumes several moderately related constructs that play different
roles in accounting for risky behavior. However, none of the
existing constructs specifically reflects the capacity for risk taking
in response to positive moods. Second, there is suggestive evi-
dence that risky and maladaptive behaviors can follow very pos-
itive mood.
fined the construct of impulsivity in many different ways. Concepts as
diverse as acting without thinking, sensation seeking, risk taking,
boredom susceptibility, adventuresomeness, and anxious impulsivity
have all been included in discussions of impulsive behavior (Depue &
Collins, 1999; Reed & Derryberry, 2005). Conceptually, these con-
cepts appear quite diverse. There are likely to be important distinc-
tions among them. For example, there are likely many individuals
(e.g., pilots) who seek thrilling sensations but who plan carefully
before acting. Two prominent models have identified distinct sub-
scales that measure different aspects of impulsivity: the behavioral
activation system framework (Carver & White, 1994; Gray, 1987;
Patterson & Newman, 1993) and Whiteside and Lynam’s (2001)
description of the four types of impulsivity within the five-factor
model of personality.
Gray (1987) held that there are two neurological systems that
regulate certain aspects of behavior: the behavioral inhibition
system and the behavioral activation system. The behavioral inhi-
bition system is thought to be activated primarily by signals of
punishment and nonreward. Its activation results in interruption of
ongoing behavior, enhanced analysis of stimuli in the environ-
ment, and cautious responding. The behavioral activation system is
thought to be activated by signals of reward and nonpunishment.
Its activation results, primarily, in approach or reward-seeking
behavior. In mixed incentive conditions (i.e., in which both reward
and punishment cues are present), individuals high in behavioral
activation continue to act, apparently in the pursuit of reward, even
though they are being punished. That propensity earns them the
label
disinhibited
or
impulsive
(Patterson & Newman, 1993).
Whiteside and Lynam (2001) factor analyzed self-report mea-
sures of impulsivity and identified four distinct but related con-
The Many Impulsivity-Like Constructs
Impulsivity is an important construct for understanding many
forms of dysfunctional behavior. In addition to being included as a
diagnostic criterion for many disorders in the
Diagnostic and Statis-
tical Manual of Mental Disorders
(4th ed.;
DSM–IV
; American Psy-
chiatric Association, 1994), impulsivity has been implicated in risk
models for a number of disorders, including alcoholism, eating dis-
orders, and pathological gambling. However, researchers have de-
Melissa A. Cyders, Gregory T. Smith, Nichea S. Spillane, and Agnes A.
Annus, Department of Psychology, University of Kentucky; Sarah Fischer,
Department of Psychiatry, University of Chicago Medical Center; Claire
Peterson, Department of Psychology, University of Dayton.
Correspondence concerning this article should be addressed to Melissa
A. Cyders, Department of Psychology, University of Kentucky, Lexington,
KY 40506-0044. E-mail: melissa.cyders@uky.edu
107
Psychological Assessment
1040-3590/07/$12.00 DOI: 10.1037/1040-3590.19.1.107
108
CYDERS ET AL.
structs, each of which corresponds to a facet of one of the five
factors of personality, as measured by the NEO Personality Inven-
tory—Revised (NEO-PI–R; Costa & McCrae, 1995). The four
were
sensation seeking
(tendency to seek out novel and thrilling
experiences: the excitement
-
seeking facet of extraversion),
lack of
deliberation
(tendency to act without thinking: the deliberation
facet of conscientiousness),
lack of persistence
(inability to remain
focused on a task: the self-discipline facet of conscientiousness),
and
urgency
(tendency to act rashly in response to distress: the
impulsivity facet of neuroticism). There is evidence that these four
factors represent distinct constructs. In addition to representing
facets on different factors in the NEO-PI–R measure of the five-
factor model, they have different correlates and explain different
aspects of risky behavior (Fischer, Smith, & Anderson, 2003;
Miller, Flory, Lynam, & Leukefeld, 2003; Smith, Fischer, Cyders,
Annus, Spillane, & McCarthy, in press; Whiteside, Lynam, Miller,
& Reynolds, 2005).
Neither of these concepts of impulsivity includes rash action in
response to a positive mood. If there are individual differences in
the tendency to respond rashly or impulsively to extremely posi-
tive mood states, they do not appear to be captured by existing
scales.
Study 1: Item Development, Item Refinement, and Factor
Analytic Investigation
In Study 1, after developing an item pool and subjecting it to
content validity analyses, we conducted factor analyses to (a)
examine the factor structure of positive urgency and (b) determine
whether positive urgency was distinct from each type of impul-
sivity described in the behavioral activation system and five-factor
model work reported previously. The original positive urgency
items either were developed on an a priori, theoretical basis or
were adapted from items from a scale that measured negative
urgency (the tendency to act rashly when in a negative mood; this
scale is described below). This process resulted in 17 items.
Method
Participants
Participants for the content validity portion of the study were
three trained raters. The raters were doctoral students trained
extensively in risky behavior and impulsivity through course work,
clinical experiences, and research in these areas.
Participants for the factor analytic portion of the study com-
prised two samples. The first sample included 1,322 undergraduate
students (mean age
18–40 years). Sixty-
four percent of the sample was female, and 36% was male. Ninety
percent of the sample was Caucasian, 5% African American, and
5% other. The second sample consisted of 300 college student
participants (175 women); the mean age of this second sample was
19.16 years (age range
19.34 years, range
Positive Mood–Based Rash Action: Positive Urgency
There is empirical evidence of tendencies toward rash action in
response to very positive mood. In general, induced positive mood
produces increased risk taking (Yuen & Lee, 2003). Undergradu-
ate college students are more likely to drink on days of celebration
than during the week (Del Boca, Darkes, Greenbaum, & Goldman,
2004; Kornefel, 2002), and that drinking tends to be heavy and
associated with increased physical violence, alcohol-related inju-
ries and deaths, driving while under the influence, and unwanted
sexual intercourse (Del Boca et al., 2004). There is also evidence
that some individuals engage in risky drinking to enhance an
existing positive mood. M. L. Cooper, Agocha, and Sheldon
(2000) found that drinking for mood enhancement leads to in-
creased drinking, drinking-related problems, and involvement in
risky behaviors. In addition, positive mood has been identified as
a temptation to resume gambling among recently quit pathological
gamblers (Holub, Hodgins, & Peden, 2005).
In light of this evidence, it is important to determine whether
there are individual differences in the tendency to respond to
positive mood with risky behavior. We hypothesized that rash
action in response to positive mood states is related to rash action
in response to negative mood states, and both represent an under-
lying dysregulation in response to extreme mood states.
The first step in examining this hypothesis was to develop a
measure of positive urgency. We developed a series of items to
fully tap the construct of positive urgency, subjected them to
content validity analysis by trained raters, examined their psycho-
metric properties, and determined their factor structure in a large,
developmental sample. We then tested positive urgency’s relations
with risky behaviors, its incremental validity in explaining risky
behavior over other forms of impulsivity, and its ability to differ-
entiate among disordered groups.
18–52 years). Mean scores on scales
used in Study 1 can be found in Table 1.
Measures
Positive Urgency Measure (PUM).
The psychometric evalua-
tion of the PUM was the object of this investigation. The nature of
the scale and the results of its evaluation are described below.
Items are assessed on a 4-point scale ranging from 1 (
agree
strongly
)to4(
disagree strongly
).
UPPS Impulsive Behavior Scale—Revised (UPPS–R).
The
UPPS–R (Whiteside & Lynam, 2001) is a 4-point Likert-type scale
used to assess four different types of impulsivity (internal consis-
tencies in parentheses): urgency (.87), deliberation (.91), persis-
tence (.82), and sensation seeking (.90). Items are assessed on a
scale ranging from 1 (
agree strongly
)to4(
disagree strongly
).
Drinking Motives Questionnaire (DMQ).
The four-factor
DMQ (M. L. Cooper, 1994) is a 20-item scale that reflects four
main motivations for drinking, including coping motives (drinking
to cope or deal with negative affect), enhancement motives (drink-
ing to enhance positive motives), social motives (drinking to
increase socialization), and conformity motives (drinking to fit in
with a group). Items on this questionnaire are rated ona1(
almost
never/never
)to5(
almost always/always
) Likert-type scale. Each
factor has an internal consistency of .84–.85, and each item loads
uniquely on one of the four factors (M. L. Cooper, 1994).
Behavioral Activation Scale (BAS).
The BAS (Carver &
White, 1994) measures individual differences in behavioral acti-
vation. It is made up of 13 items divided into three subscales:
Reward Responsiveness (5 items; e.g., “When I get something I
want, I feel excited and energized”), Drive (4 items; e.g., “When
POSITIVE URGENCY
109
Table 1
Mean Scores and Standard Deviations for Study Samples 1 and 2
Study 1
Study 2
Sample 1
(
n
1,322)
Sample 2
(
n
300)
Sample 1
(
n
326)
Sample 2
(
n
216)
Measure
MSD MSD MSD MSD
Positive urgency
1.65
.56
1.86
.57
1.58
.59
1.75
.66
Urgency
2.28
.50
2.21
.58
2.28
.61
Perseverance
1.94
.45
1.83
.44
1.74
.43
Deliberation
2.08
.44
2.05
.44
1.91
.47
Sensation seeking
2.89
.58
2.93
.56
2.83
.58
Behavioral activation
Reward responsiveness
1.42
.42
Drive
2.21
.60
Fun seeking
2.05
.62
Note.
All scales were coded so that higher scores indicate more impulsive action. Scales ranged from 1 (
low
impulsivity
)to4(
high impulsivity
).
I want something, I usually go all out to get it”), and Fun Seeking
(4 items; e.g., “I crave excitement and new sensations”). In the
developmental sample and the current sample, the BAS had an
overall internal consistency of .87. Items were assessed on a
4-point Likert-type scale ranging from 1 (
agree strongly
)to4
(
disagree strongly
).
any other scale as a positive urgency item. The full set of items
resulting from this study is listed in Table 2.
Exploratory Factor Analysis of the PUM
We first conducted an exploratory factor analysis on the PUM
items, using a random sample of
n
666 from the total sample of
1,322 participants. Because items were scored on a 4-point Likert-
type scale, we used polychoric correlations. Although we antici-
pated a one-factor solution following the content validity analysis,
we considered solutions from one to four factors. A principal
factor analysis with oblique oblimin rotation produced one factor
with a scree plot that gave clear indication of a one-factor solution.
Using parallel analysis, we found that the eigenvalue for the
second factor was smaller than the average eigenvalue produced
from 50 factor analyses of random data and smaller than the
eigenvalue at the 95th percentile of eigenvalues produced from
random data. The Factor 1 loadings were quite high, ranging from
.76 to .99. In addition, each of the highest loading items on Factor
2 loaded much more highly on Factor 1. We, therefore, concluded
that a one-factor solution best fit the data for the exploratory
subsample. The factor loadings for the PUM are presented in Table 2.
Procedure
Raters were trained for this study through explanation of the
new proposed trait of positive urgency. They were then asked to
differentiate items that represent positive urgency from items rep-
resenting the three BAS subscales, Whiteside and Lynam’s (2001)
four impulsivity-like constructs, and motives to drink alcohol. If
raters could consistently differentiate positive urgency items from
BAS and UPPS–R items, we would conclude that the positive
urgency items were specific in two senses: They would not be
indices of broad, general impulsivity that overlapped with other
scales, nor would they be inadvertent measures of other particular
types of impulsivity. If raters could differentiate the items from
motives to enhance positive mood by drinking, they would be
specific in the sense that they did not refer to that specific mood-
enhancement strategy. Raters were also asked to judge whether the
items clearly reflected the defined construct of positive urgency
and to identify items that did not.
Participants in the first factor analysis sample completed the
PUM in a group-administration format. Participants in the second
sample completed the PUM, the BAS, and the UPPS–R in a
group-administration format. All participants completed informed
consent procedures.
Confirmatory Factor Analysis of the PUM
We next subjected the one-factor solution to confirmatory factor
analysis on the second random sample (
n
656), again using
polychoric correlations. We used the weighted least squares esti-
mation method. The model resulted in a comparative fit index
(Bentler, 1990) of .99 and a Tucker–Lewis fit index of .99 (Tucker
& Lewis, 1973). These values indicate an acceptable fit: Conven-
tion holds acceptable fit at .90 (Kline, 2005) or even more strin-
gently at .95 (Hu & Bentler, 1999). The model had a root-mean-
square error of approximation (Browne & Cudeck, 1993) of .06
(90% confidence interval from .05 to .07), which is considered a
fair fit. The loadings were again quite high, ranging from .59 to
.85. See Table 3 for the correlation matrix of items included in the
confirmatory factor analysis. The PUM scale, consisting of the 14
Results
Content Validity Analyses
Positive urgency items that were misclassified by any of the
three experts were deleted. That procedure resulted in the deletion
of 3 items. There was 100% agreement that the remaining 14 items
reflected positive urgency, and no rater mislabeled an item from
110
CYDERS ET AL.
Table 2
Exploratory Factor Loadings for the Positive Urgency Measure Items
Item
Loading
1. When I am very happy, I can’t seem to stop myself from doing things that can have
bad consequences.
.92
2. When I am in great mood, I tend to get into situations that could cause me problems.
.83
3. When I am very happy, I tend to do things that may cause problems in my life.
.93
4. I tend to lose control when I am in a great mood.
.92
5. When I am really ecstatic, I tend to get out of control.
.97
6. Others would say I make bad choices when I am extremely happy about something.
.90
7. Others are shocked or worried about the things I do when I am feeling very excited.
.97
8. When I get really happy about something, I tend to do things that can have bad
consequences.
.96
9. When overjoyed, I feel like I can’t stop myself from going overboard.
.76
10. When I am really excited, I tend not to think of the consequences of of my actions.
.98
11. I tend to act without thinking when I am really excited.
.92
12. When I am really happy, I often find myself in situations that I normally wouldn’t be
comfortable with.
.97
13. When I am very happy, I feel like it is OK to give in to cravings or overindulge.
.99
14. I am surprised at the things I do while in a great mood.
.97
Note.
Data are from a principal factor analysis with oblique rotation.
n
666.
items, showed an internal consistency of
.94 in the combined
sample (
n
1,322), with a median interitem correlation of
r
.79
(ranging from
r
.37 to
r
.85).
the fourth factor consisted of the 4 items on the Fun Seeking
subscale of the BAS.
There was a clear simple structure to the results of this factor
analysis. All PUM items loaded most highly on the PUM factor;
their factor loadings ranged from .82 to .91, with an average factor
loading of .86. The loading of each PUM item on the PUM factor
was at least .67 higher than its loading on any other factor. For
reward responsiveness, item loadings ranged from .78 to .97; for
drive, item loadings ranged from .75 to .94; for fun seeking, item
loadings ranged from .81 to .91. No BAS items had secondary
loadings higher than .15 on the PUM factor. The positive urgency
factor’s correlations with the other three factors were reward
responsiveness,
r
.03; drive,
r
.04; and fun seeking,
r
.36.
PUM and BAS Factor Analysis
We next conducted an exploratory factor analysis using a prin-
cipal factor analysis with an oblique rotation on the items from the
PUM and the BAS, again using polychoric correlations. Using the
rotated solution, we found that four factors had eigenvalues greater
than 1, and a scree plot strongly suggested a four-factor solution.
Using parallel analysis, we found that the four factors had eigen-
values greater than both the average eigenvalue from 50 factor
analyses of random data and the 95th percentile eigenvalue from
those 50 random data factor analyses. The first factor comprised
items from the PUM; the second factor was made up of the 5 items
on the Reward Responsiveness subscale of the BAS; the third
factor consisted of the 4 items on the Drive subscale of the BAS;
PUM and UPPS–R Factor Analysis
We conducted an exploratory principal factor analysis with an
oblique rotation using the PUM and UPPS–R scales, again using
Table 3
Correlation Matrix for Confirmatory Factor Analysis
Item 1
2
3
4
5
6
7
8
9 10 11 12 13 14
1 — .60 .60 .67 .75 .60 .74 .63 .50 .69 .62 .68 .75 .69
2
— .73 .60 .61 .53 .58 .49 .37 .53 .56 .53 .55 .50
3
— .78 .73 .67 .73 .61 .45 .72 .63 .70 .67 .62
4
— .77 .73 .74 .62 .49 .71 .64 .71 .70 .64
5
— .74 .85 .75 .48 .73 .70 .79 .84 .80
6
— .73 .64 .52 .70 .62 .67 .62 .63
7
— .74 .48 .74 .72 .79 .85 .79
8
— .55 .77 .67 .75 .80 .84
9
— .64 .54 .54 .53 .54
10
— .72 .85 .78 .78
11
— .79 .75 .76
12
— .85 .81
13
— .89
14
—
Note.
Items are labeled by item number in Table 2. All correlations were significant at
p
.01.
POSITIVE URGENCY
111
polychoric correlations.
1
Using the rotated solution, we found that
five factors had eigenvalues greater than 1, and a scree plot
strongly suggested a five-factor solution. Using parallel analysis,
the five factors had eigenvalues greater than both the average
eigenvalue from 50 factor analyses of random data and the 95th
percentile eigenvalue from those 50 random data factor analyses.
As anticipated, the five factors consisted of the items measuring
positive urgency, sensation seeking, lack of perseverance, negative
urgency, and lack of deliberation.
For positive urgency, all PUM items loaded highest on the
positive urgency factor. PUM loadings ranged from .55 to .81,
with a mean loading of .68. PUM loadings were at least .22 higher
than their loadings on any other factor. For the UPPS–R scales, all
the items loaded most highly on their assigned scale. The negative
urgency item loadings ranged from .64 to .94, with a mean loading
of .75. Negative urgency items tended to have secondary loadings
on positive urgency but no primary loadings on the PUM factor.
The mean secondary loading of negative urgency items on the
PUM factor was .11. Sensation-seeking item loadings ranged from
.62 to .91, with a mean loading of .76. Lack of perseverance item
loadings ranged from .76 to .97, with a mean loading of .86. The
lack of deliberation item loadings ranged from .46 to .92, with a
mean loading of .62. The positive urgency factor’s correlations
with the other four factors were sensation seeking,
r
.21; lack of
perseverance,
r
.22; negative urgency,
r
.37; and lack of
deliberation,
r
.28 (
p
.01 in each case).
Budde & Testa, 2005). Thus, college student samples may reason-
ably reflect the rates of risky behavior characterizing late adoles-
cents in general. Because the rates of risky behaviors are quite high
among college students, risk-related phenomena can be studied
and are of clinical interest in this population (Hingson, Heeren,
Winter, & Wechsler, 2005). In addition, college students’ risky
behavior appears often to be associated with celebrations and good
moods: It tends to occur on weekends, college breaks, and times
without heavy school demands (Del Boca et al., 2004).
Method
Participants
Sample 1.
Sample 1 participants were a subset (
n
326) of
the 1,322 students who participated in Study 1. The 326 partici-
pants ranged from 18 to 52 years of age, with a mean age of 19.1
years. Fifty-two percent of the sample was male. African Ameri-
cans made up 4% of the sample, Caucasians 90%, Asian Ameri-
cans 2%, and other ethnicities 4%. All participants completed adult
written consent to participate in the study. Mean scores can be
found in Table 1.
Sample 2.
Sample 2 participants were 216 college students;
79% of the sample were women. Ethnicities were as follows:
Caucasian 89%, African American 8%, Asian American 2%, and
other ethnicities 1%. Participants had a mean age of 18.2 years
(range
18–31 years). See Table 1 for mean scores.
Discussion
Measures
The PUM scale appears to be content valid and unidimensional.
It represents a distinct factor from those represented by the sub-
scales of the BAS and from those represented by the four scales of
the UPPS–R. Thus, positive urgency does not appear to be repre-
sented in a psychometric representation of the behavioral activa-
tion system, nor does it appear to be represented among the four
impulsivity-like facets of the NEO-PI–R. However, the current
study did not address external correlates of positive urgency; this
was the aim of Study 2.
PUM.
The PUM was described above. Sample 1 and Sample
2 participants completed the PUM and had an internal consistency
of .94 in Sample 1 and .95 in Sample 2.
UPPS–R.
The UPPS–R was described above; it was com-
pleted by both samples and had an overall internal consistency of
.75 in Sample 1 and .89 in Sample 2.
Negative Outcome Scale (NEGO).
This scale assesses the fre-
quency of participating in risk-taking activities that are likely to
have a negative outcome. The scale proved successful in identify-
ing risky activities, and it related to multiple forms of impulsivity
as hypothesized (Fischer & Smith, 2004). It consists of 10 items
measuring risky or impulsive behaviors. The items ask the indi-
vidual to indicate how many times in the past year they have
participated in a range of activities, with answers ranging from 1
(
0 times in the past year
)to5(
16 or more times in the past year
).
Sample items include stealing something valued less than $100,
trespassing, and having sex with someone who was involved with
someone else. In the current sample, the scale had an internal
consistency of
.70, with a mean endorsement level of 1.59.
Sample 1 participants completed this measure.
South Oaks Gambling Screen (SOGS).
The SOGS (Lesieur &
Blume, 1988) is a 20-item inventory designed to screen for
Diag-
Study 2: External Correlates of Positive Urgency
The primary aim of Study 2 was to begin to test whether positive
urgency plays a different role in explaining risky behaviors from
that of other forms of impulsivity. By risky behaviors, we mean
ill-considered actions that increase the risk of harm to the individ-
ual (e.g., trespassing, having sex with someone involved with
someone else). We thus tested (a) the hypothesis that positive
urgency would explain variance in risky behavior participation not
explained by other forms of impulsivity, and (b) the hypothesis
that positive urgency would uniquely explain variance in risky
behaviors likely to stem from an existing positive mood and, thus,
would play a different role from other forms of impulsivity.
In this study, we examined positive urgency’s relation to risky
behaviors in college students for these reasons. When adolescents
leave home, the rates of at least some forms of risky behavior tend
to increase beyond their already high adolescent levels (Budde &
Testa, 2005; Kelley, Schochet, & Landry, 2004). Late adolescent
rates of risky behavior appear not to differ as a function of college
attendance (rather, what matters is leaving adult supervision;
1
We replicated each factor analysis using Pearson product–moment
correlations. In each case, the pattern of results was the same, and all
conclusions drawn were identical. The only difference between the two
methods was that the factor loadings estimated using Pearson product–
moment correlations were consistently slightly lower than those estimated
using polychoric correlations.
[ Pobierz całość w formacie PDF ]