Peer Influence on Risk Taking, Risk Preference, and Risky Decision Making in Adolescence and Adulthood: An Experimental Study

Abstract

In this study, 306 individuals in 3 age groups--adolescents (13-16), youths (18-22), and adults (24 and older)--completed 2 questionnaire measures assessing risk preference and risky decision making, and 1 behavioral task measuring risk taking. Participants in each age group were randomly assigned to complete the measures either alone or with 2 same-aged peers. Analyses indicated that (a) risk taking and risky decision making decreased with age; (b) participants took more risks, focused more on the benefits than the costs of risky behavior, and made riskier decisions when in peer groups than alone; and (c) peer effects on risk taking and risky decision making were stronger among adolescents and youths than adults. These findings support the idea that adolescents are more inclined toward risky behavior and risky decision making than are adults and that peer influence plays an important role in explaining risky behavior during adolescence.

Full text

Peer Influence on Risk Taking, Risk Preference, and Risky Decision
Making in Adolescence and Adulthood: An Experimental Study
Margo Gardner and Laurence Steinberg
Temple University
In this study, 306 individuals in 3 age groups—adolescents (13–16), youths (18 –22), and adults (24 and
older)— completed 2 questionnaire measures assessing risk preference and risky decision making, and 1
behavioral task measuring risk taking. Participants in each age group were randomly assigned to
complete the measures either alone or with 2 same-aged peers. Analyses indicated that (a) risk taking and
risky decision making decreased with age; (b) participants took more risks, focused more on the benefits
than the costs of risky behavior, and made riskier decisions when in peer groups than alone; and (c) peer
effects on risk taking and risky decision making were stronger among adolescents and youths than adults.
These findings support the idea that adolescents are more inclined toward risky behavior and risky
decision making than are adults and that peer influence plays an important role in explaining risky
behavior during adolescence.
Keywords: adolescents, risk taking, peer influence, risk preference, decision making
It is well documented that adolescents are more likely than
adults to engage in risky behavior. For example, adolescents are
more likely than adults to drive recklessly, to drive while intoxi-
cated, to use varied illicit substances, to have unprotected sex, and
to engage in both minor and more serious antisocial behavior
(Arnett, 1992). However, despite clinical and anecdotal evidence
of heightened real-world risk taking during adolescence, labora-
tory studies of age differences in risk preference, risk perception,
and risky decision making have not yielded consistent evidence
that adolescents are actually less risk averse than are their elders.
In fact, it is often asserted that, by midadolescence, teens’ capac-
ities for understanding and reasoning in risky decision-making
situations roughly approximate those of adults (Fischhoff, 1992;
Furby & Beyth-Marom, 1992). This assertion has been used to
argue both for protecting adolescents’ rights to make autonomous
decisions about their reproductive health and for holding adoles-
cents to adult standards of criminal blameworthiness (see Stein-
berg & Scott, 2003, for a discussion).
However, as several writers have recently argued, extant studies
suggesting equivalent orientations toward risk among adolescents
and adults are only modestly useful in understanding how adoles-
cents compare with adults in real-world decision making. These
authors suggest that typical laboratory studies of risky decision
making fail to consider the emotional and social contexts in which
risk taking actually occurs (Cauffman & Steinberg, 2000; Scott,
Reppucci, & Woolard, 1995; Steinberg, 2004; Steinberg & Cauff-
man, 1996). In such studies, individual adolescents are presented
with hypothetical dilemmas under conditions of low emotional
arousal and are then asked to make and explain their decisions. In
the real world, however, adolescents’ decisions are not hypothet-
ical, they are generally made under conditions of emotional arousal
(whether negative or positive), and they are usually made in peer
groups. Whether the risky decision making of adolescents is truly
comparable to that of adults under real-world conditions remains
an open and unstudied question.
A number of explanations have been advanced to account for
differences between adolescents and adults in real-world, as op-
posed to laboratory-based, risk taking. Some have argued that age
differences in psychosocial capacities such as impulse control or
sensation seeking play an important role (see Steinberg & Cauff-
man, 1996). Consistent with this, Cauffman and Steinberg (2000)
reported that once differences in psychosocial maturity between
adolescents and adults are accounted for, age differences in risky
decision making disappear. An alternative and entirely compatible
account of age differences in risky behavior emphasizes the role of
peers and, more specifically, peer influence. That is, adolescents
may engage in more risky behavior than do adults because they are
more susceptible to the influence of their similarly risk-prone
peers. Support for this latter explanation comes, in part, from the
criminology literature. There is a small but compelling body of
evidence to suggest that when adolescents commit crimes—acts
that are inherently risky—they generally do so with their peers
(Erickson & Jensen, 1977; Zimring, 1998). For example, adoles-
cents are usually accompanied by one or more persons when
committing crimes that range in seriousness from vandalism and
drug use (Erickson & Jensen, 1977) to rape and homicide (Zim-
Margo Gardner and Laurence Steinberg, Department of Psychology,
Temple University.
This study was supported by a grant from the John D. and Catherine T.
MacArthur Foundation Network on Adolescent Development and Juvenile
Justice. We thank Rebecca Davis, Rebecca Garrett, Lauren Guttshall,
Nermine Salama, Benjamin Steinberg, and Erica Weitz for help with data
collection. We also thank He Len Chung, Alex Piquero, and Jennifer Silk
for their help with the selection and implementation of the analytic method
used.
Correspondence concerning this article should be addressed to Margo
Gardner at the Department of Psychology, Temple University, Philadel-
phia, PA 19122. E-mail: mnoel002@temple.edu
Developmental Psychology Copyright 2005 by the American Psychological Association
2005, Vol. 41, No. 4, 625–635 0012-1649/05/$12.00 DOI: 10.1037/0012-1649.41.4.625
625

ring, 1998). This is not, however, true of adults; when adults
commit crimes, they typically do so alone (Zimring, 1998).
Although adolescent risk taking often occurs in groups, it is not
known whether the greater prevalence of group risk taking ob-
served among adolescents stems from the fact that adolescents
spend more time in peer groups than adults do (Brown, 2004) or
from the heightened levels of susceptibility to peer influence that
have been shown to characterize adolescence (Steinberg & Silver-
berg, 1986). In other words, it is not clear whether adolescents
simply have more opportunities to engage in group risk taking than
do adults or whether, when faced with behavioral decisions in a
peer group context, adolescents are more easily swayed toward
risky choices.
To our knowledge, only one study has attempted to determine
whether there are developmental differences in the effects of actual
peer presence on orientation toward risk. In a comparison of
adolescents and college students, Hensley (1977) sought to deter-
mine whether the tendency for individuals to take more risks in
groups than when alone—a phenomenon known as the risky shift
(Vinokur, 1971)—might differ across age groups. Hensley (1977)
used a hypothetical decision-making questionnaire to measure risk
acceptance and found that the magnitude of the risky shift was
greater among adolescents than it was among college students.
However, the study sample was very small (22 college students
and 18 adolescents), and these results have not, to our knowledge,
been replicated with performance (as opposed to hypothetical)
measures of risk taking.
There are other findings that indirectly support the notion that
adolescents may be more easily swayed toward risky behavior than
adults. Compared with adults, adolescents have limited abilities in
areas of psychosocial functioning, such as self-reliance, which
likely interfere with the ability to act independently of the influ-
ence of others (Cauffman, 1996; Cauffman & Steinberg, 2000;
Steinberg & Cauffman, 1996). Not surprisingly, several studies
have found a curvilinear relation between age and peer conformity
on responses to hypothetical dilemmas about antisocial decision
making, with conformity increasing throughout childhood and into
midadolescence and decreasing thereafter (Berndt, 1979; Brown,
Clasen, & Eicher, 1986; Steinberg & Silverberg, 1986). Although
researchers have not examined the developmental pattern of resis-
tance to peer influence beyond late adolescence, there is some
evidence that peer influence remains an important predictor of
participation in risky behavior even during young adulthood (An-
drews, Tildesley, Hopps, & Li, 2002; Horvath & Zuckerman,
1993). Thus, when confronted with risky decisions in the context
of a peer group, adolescents, and perhaps even young adults, may
be less able than older adults to resist the influence of their
risk-prone age mates.
Further support for the idea of heightened peer effects on risky
behavior during adolescence comes from additional findings on
the risky-shift. Although a number of researchers have found that
risk-taking tendencies are greater when individuals are in groups
than when alone (e.g., Blascovich & Ginsburg, 1974; Blascovich,
Ginsburg, & Howe, 1975; Blascovich, Veach, & Ginsburg, 1973;
Kogan & Wallach, 1967; Lamm, 1967; Lamm, Trommsdorff, &
Rost-Schaude, 1972; Pruitt & Teger, 1969; Vidmar, 1970; Wallach
& Kogan, 1965; Yinon, Jaffe, & Feshbach, 1975), several inves-
tigations have found the reverse to be true. Indeed, in some cases,
individuals demonstrate a conservative shift and are actually more
risk averse when in groups than when alone (e.g., Cohen & Ruis,
1974; Pilkonis & Zanna, 1973; Zaleska, 1974). Accordingly, social
psychologists have advanced an alternative theoretical framework
for understanding group risk taking. Whereas proponents of the
risky shift theory assert that the presence of others should always
lead to increased risk taking, advocates for the more recent group
polarization theory suggest that the direction of group effects on
risk taking depends on the risk-taking tendencies of the group
members (Hogg, Turner, & Davidson, 1990). According to this
theory, relatively conservative individuals should become even
more conservative when grouped together, whereas individuals
who are inclined to take risks should make even more risky
choices (Hogg et al., 1990). Given this theoretical framework,
adolescents’ generally greater inclination toward risky behavior as
individuals, in combination with their greater susceptibility to peer
influence should, in theory, result in a larger effect of peer pres-
ence on risky behavior among adolescents than among adults.
1
The
goal of the present study is therefore to examine whether adoles-
cents, relative to adults, are more likely to take risks when their
peers are present.
The Present Study
In the present study, we examined the differential effects of the
presence of peers on risk taking, risk preference, and risky decision
making among adolescents (Mage ⫽14), youths (Mage ⫽19),
and adults (Mage ⫽37). Our three primary hypotheses were as
follows:
Hypothesis 1. Risk taking, risk preference, and risky decision
making will decrease with age.
Hypothesis 2. On average, individuals will demonstrate more
risk taking, greater risk preference, and more risky decision
making when in the company of their peers than when alone.
Hypothesis 3. The difference between levels of risk taking,
risk preference, and risky decision making with and without
the presence of peers will decrease with age. That is, group
effects on risk orientation will be greater among adolescents
than among youths, and greater among youths than among
adults.
Method
Sample
Our sample included 106 adolescents (54 girls and 52 boys), ages 13 to
16 (Mage ⫽14.01, SD ⫽1.02), 105 youths (53 women and 52 men), ages
18 to 22 (Mage ⫽18.78, SD ⫽1.07), and 95 adults (48 women and 47
men), ages 24 and older (Mage ⫽37.24, SD ⫽12.37). All participants
1
In keeping with group polarization theory, we are not suggesting that
all adolescents should demonstrate shifts toward increased risk taking
when in the presence of peers. It is conceivable that some adolescents,
when placed in a group with risk-averse peers, might shift toward de-
creased risk taking. However, we expect that, given generally greater
propensities for risk taking among adolescents, adolescents should, on
average, be more likely than adults to demonstrate group-induced shifts
toward greater risk taking.
626 GARDNER AND STEINBERG

were recruited from areas in and around a major urban center. The
adolescents were recruited from middle schools, day camps, and commu-
nity centers; the youths were recruited from undergraduate introductory
psychology courses at a large urban university; and the adults were re-
cruited through fliers posted on urban university and community college
campuses, advertisements distributed to community organizations, and
word of mouth.
The adolescent sample was composed of 50.9% girls and 49.1% boys;
the youth sample was composed of 50.5% women and 49.5% men; and the
adult sample was composed of 50.5% women and 49.5% men. These three
groups did not differ significantly with respect to gender composition,
␹
2
(2) ⫽.005, p⫽.997. The three age groups were also very similar in
terms of their ethnic composition. The majority of the participants were
either White (48.7%) or African American (38.2%). Given the very small
percentage of participants from other ethnic groups (the sample included
only 1% Native Americans, 7.2% Asian Americans, 3.9% Latinos, and
0.7% others), the three age groups were compared only with respect to the
percentages of White versus non-White participants. The adolescent sam-
ple included 44.8% White participants and 55.2% non-White participants;
the youth sample included 53.3% White participants and 46.7% non-White
participants; and the adult sample included 48.4% White participants and
51.6% non-White participants. These three groups did not differ signifi-
cantly with respect to ethnic composition,
␹
2
(2) ⫽1.554, p⫽.460. Finally,
on a participant-report parent education rating scale (a proxy for socioeco-
nomic status) where 1 ⫽less than high school diploma,2⫽high school
diploma/GED,3⫽some college/vocational school,4⫽college graduate,
and 5 ⫽graduate or professional school, all three age groups reported a
mean level of parent education within the range of some college or
vocational training. Thus, the three age groups did not differ substantially
with respect to gender, ethnicity, or socioeconomic status (as indexed by
parent education).
Recruitment and Procedure
Slightly different recruitment procedures were used for the adolescents
versus the college students and adults. Each college undergraduate and
adult participant was asked to invite to the session 2 people of his or her
gender that he or she knew. These groups of 3 were then randomly assigned
to a group condition (in which all 3 participants completed the battery of
measures at the same time, in the same room, and communicated with each
other while completing the tasks and measures) or to a sole participant
condition (in which each of the 3 participants completed the battery of
measures alone while the other 2 participants waited outside the room
where the session took place).
Requirements concerning the need for parental consent among the
adolescent participants necessitated a slight modification in this procedure.
We found it difficult to get predetermined groups of 3 adolescents to
appear for scheduled appointments together and to arrive with all 3 signed
parent consent forms in hand. Thus, after first obtaining parental consent
from large numbers of adolescents at different recruitment sites, we ran-
domly assigned half of these adolescents within each recruitment site to the
group condition (in which they completed the measures with two other
teenagers of the same gender from their recruitment site) and half to the
sole participant condition (in which they completed the measures alone).
Thus, while the adolescent triads in the group condition were always
composed of individuals who knew and were familiar with each other (i.e.,
individuals within each triad were from the same camp, classroom, or
community-center summer or afterschool program), they were not neces-
sarily composed of individuals who had selected one another as partners
for the experiment. Although the adolescent and two older groups differed
in this respect, in the real world, adolescents often find themselves in
groups with other teenagers whom they know but have not necessarily
selected as companions (i.e., in classrooms, on sports teams, in extracur-
ricular activities, etc.). Moreover, to the extent that friends have been
shown to exert more pressure on each other than acquaintances (e.g.,
McPhee, 1996), the way in which adolescents were recruited resulted in a
more conservative test of the hypothesis that, relative to adults, risk
orientation among adolescents is more influenced by peer pressure.
Among the adults, 54 participants (27 men and 27 women) were as-
signed to the group condition, and 41 participants (20 men and 21 women)
were assigned to the sole participant condition.
2
Among the youths, 54
participants (27 men and 27 women) were assigned to the group condition,
and 51 (25 men and 26 women)
3
were assigned to the sole participant
condition. Among the adolescents, 54 participants (27 boys and 27 girls)
were assigned to the group condition, and 52 (25 boys and 27 girls) were
assigned to the sole participant condition.
All participant triads were composed of individuals of the same gender.
However, the ethnic composition of the triads was not constricted in this
manner. Among the adults, 37.5% of the triads consisted of all White
participants, 43.8% consisted of all non-White participants, and 18.8% of
the triads consisted of White and non-White individuals. Among the
youths, 45.9% of the triads consisted of all White participants, 37.8%
consisted of all non-White participants, and 16.2% consisted of White and
non-White participants. Among the adolescents, 19.4% of the triads con-
sisted of all White participants, 25% consisted of all non-White partici-
pants, and 55.6% consisted of White and non-White participants.
All participants completed three measures of risk orientation that were
part of a battery of measures administered for a larger study of psychos-
ocial development. The entire battery of measures took approximately 1 hr
to complete. Each adolescent and adult participant was compensated $20,
and each undergraduate participant was given the choice between either a
$20 payment or research credit in an introductory psychology course.
Measures
Risk taking. Risk taking was assessed with a video game called
“Chicken” (Sheldrick, 2004). Chicken is played on a laptop computer and
requires participants to make decisions about whether to stop a car that is
moving across the screen once a traffic light turns from green to yellow.
The appearance of the yellow light signals the impending appearance of a
red traffic light, as well as a potential crash if the car is still moving when
the red light appears. Chicken was selected because it measures risk taking
in the moment rather than the more deliberative form of risk taking
assessed in many studies, in which participants have unlimited time to
consider and evaluate all potential decisions and outcomes. Additionally,
Chicken requires participants to make actual decisions in a risky situation,
rather than simply requiring participants to report what they would do in a
hypothetical risky situation.
The game is played from a third-person, side-view perspective (see
Figure 1) and consists of 15 trials. In each trial, participants watched an
animated car move across the screen for a predetermined amount of time
until a yellow traffic light appeared. Before the first trial, players were
informed that at some unknown point after the yellow light appeared, the
traffic light would turn red and a wall would pop up in front of the car.
Players were told that the object of the game was to allow the car to move
as far as possible without crashing into the wall. Players controlled whether
the car was moving or stopped but not the speed of the car. Participants
accumulated more points the further the car moved without crashing but
2
One person was dropped from one of the adult sole participant triads.
At the age of 79, we believed this individual to be developmentally
dissimilar from the rest of the adult sample.
3
One person was missing from three of the young adult triads assigned
to the sole participant condition. In the first two cases, the 3rd member of
the group left before it was his or her turn to complete the session. In the
third case, data for the 3rd member of the peer group were excluded
because it was later determined that the participant had already completed
the study several months prior.
627
RISK TAKING IN ADOLESCENCE AND ADULTHOOD

lost any points that had been accumulated on that trial if the car crashed.
If a player stopped the car before it crashed, the player had the option of
restarting the car and allowing it to move further, or leaving the car where
it was and accepting the amount of points accumulated. Thus, when the
yellow light appeared, players had to decide how much further to allow the
car to move, balancing their desire to accumulate points against the
possibility of crashing the car into the wall. The latency between the
beginning of the trial and the appearance of the yellow light, and between
the appearance of the yellow light and the appearance of the wall varied
across trials, such that the participants did not know whether the wall
would appear suddenly or after some delay. The objectives of the game and
the potential positive and negative outcomes (earning points vs. crashing
and failing to accumulate points, respectively) were explained to partici-
pants during a demonstration. In order to ensure that all participants were
equally familiar with the potential consequences of driving through a
yellow traffic light, the demonstration round included a depiction of the
animated car both driving safely through the yellow light without crashing,
as well as driving through the yellow traffic light and crashing into the pop
up wall.
The computer recorded the amount of time that the car was in motion
between the onset of the yellow light and the car’s final stop, as well as the
number of car restarts per round. Mean scores for the number of car restarts
per round, and the percentage of time the car was in motion were calculated
for each participant. Longer moving times and more restarts indicated
greater risk taking. Scores on these two indices of risk taking were highly
correlated (r⫽.61, p⬍01) and were therefore standardized and averaged
in order to compute a composite indicator of risk taking on the Chicken
game.
Participants in the sole participant condition completed the task as
described. Participants in the group condition took turns playing the game,
but all of them completed 15 trials in a row, as did participants in the sole
participant condition. In the group condition, while one participant was
playing the game, the other two were told that they could call out advice
about whether to allow the car to keep moving or to stop it. The player was
instructed that he or she could choose whether to follow the advice of his
or her peers.
Risk preference. A shortened, modified version of the Benthin Risk
Perception Measure (BRPM; Benthin, Slovic, & Severson, 1993) was used
to assess risk preference. This measure assesses both risk perception (the
extent to which one perceives a given activity as carrying the potential for
adverse consequences) and risk preference (whether one believes the
benefits inherent in an activity outweigh the costs, or vice versa). Only data
from the scale reflecting cost– benefit consideration are used in the present
analyses. We chose not to include data from the risk perception scale
because prior studies have failed to find age differences in performance on
this scale (Steinberg, 2004). Similarly, evidence from the research of
Beyth-Marom and colleagues (Beyth-Marom, Austin, Fischoff, Palmgren,
& Jacobs-Quadrel, 1993) suggests that adolescents and adults are relatively
equal in terms of their awareness of the potential for adverse consequences
in risky situations. However, as argued by Furby and Beyth-Marom (1992),
adolescents and adults may differ in terms of the relative weights or values
that they attach to the potential costs and benefits of risky activities. Studies
of the relation between risk taking and cost versus benefit consideration
suggest that those who give lesser consideration to costs and greater
consideration to benefits are more likely to engage in risky behavior (e.g.,
Fromme, Stroot, & Kaplan, 1993; Goldberg & Fischhoff, 2000; Horvath &
Zuckerman, 1993; Lavery, Siegel, Cousins, & Rubovits, 1993; McBride,
Weatherby, Inciardi, & Gillespie, 1999; Singer, Dai, Weeks, & Malave,
1998; Thorton, Gibbons, & Gerrard, 2002). Thus, differential consideration
or weighting of potential costs versus benefits among adolescents and
adults may partially account for observed age differences in risky behavior.
In completing the Risk Preference Scale, participants were presented
with five hypothetical scenarios involving risky behavior. These scenarios
included having sex without a condom, riding in a car driven by someone
Figure 1. An image from the Chicken video game. In this frame, the traffic light has just turned red. The car
was still moving when the light turned red; consequently, a brick wall appeared in front of the car, resulting in
a crash.
628 GARDNER AND STEINBERG

who has been drinking, trying a new drug that one does not know anything
about, breaking into a store at night and stealing something that one really
wants, and driving over 90 mph on the highway at night. They were then
asked to rate on a 4-point scale ranging from 1 (risks are much greater than
benefits)to4(benefits are much greater than risks) how the risks com-
pared with the benefits of the activity. A mean risk– benefit consideration
score was then calculated for each participant by averaging responses
across the five scenarios (
␣
⫽.68).
Individuals in the sole participant condition read the scenarios from
index cards and indicated their choices on a response card displaying the
4-point scale. Group condition participants followed the same procedure
but were told that they could discuss each question. However, they were
instructed that they need not reach a consensus and that each could make
a final decision at any time. Each participant had his or her own set of
response cards and had an unobstructed view of the others’ response cards.
The administrator recorded individuals’ responses.
Risky decision making. Risky decision making was assessed via the
Youth Decision-Making Questionnaire (YDMQ; Ford, Wentzel, Wood,
Stevens, & Siesfeld, 1990). Participants were presented with five hypo-
thetical dilemmas, each involving a risky decision. The dilemmas included
decisions about allowing friends to bring drugs into one’s home, stealing a
car, cheating on an exam, shoplifting, and skipping work without an
excuse, all of which adolescents, college undergraduates, and adults po-
tentially could have done. Decisions about each dilemma were made within
the context of three different scenarios. In the first scenario, participants
were informed that no matter what their decision, no negative conse-
quences would result. The second scenario—introduced by Cauffman and
Steinberg (2000)—stated that negative consequences might result if the
risky course of action were taken. The final scenario stated that negative
consequences would definitely occur if the risky course of action were
taken.
Only responses from the second decision-making scenario (i.e., negative
consequences might result) were included in the analyses for the present
study, as this was the only scenario that involved some degree of uncer-
tainty or risk. For each dilemma, participants were asked to decide what
they would do “if they were really in that situation” on a 4-point scale that
ranged from 1 (definitely making the risky decision)to4(definitely not
making the risky decision). Scores were reverse coded, such that higher
scores indicated higher risk-taking tendencies. A mean risky decision-
making score was calculated for each participant by averaging the scores
across the five dilemmas (
␣
⫽.65).
The YDMQ was presented on a laptop computer. Individuals in the sole
participant condition were asked to indicate their desired choices on cards
displaying the 4-point response scale. Participants in the group condition
followed the same procedure but were told that they could discuss each
situation. They were also informed that they did not need to reach a
consensus and that they could each make a final decision at any time. Each
group condition participant had his or her own set of response cards, and
each had an unobstructed view of the others’ cards. The administrator
recorded participants’ responses. Means, standard deviations, and intercor-
relations among the study variables are presented in Table 1.
Results
Data Analyses
Because participants were recruited in groups of 3, scores for
participants within each triad could not be treated as independent.
4
In order to accommodate the nested structure of the data, all
analyses were performed with the linear mixed model (LMM)
procedure in the Statistical Package for Social Sciences 11.5
(SPSS; SPSS, Inc., 2005). Unlike the general linear model (GLM)
procedure, which assumes that all observations are independent of
one another, the LMM procedure allows for correlated variability
among observations. Because the LMM procedure does not permit
the simultaneous analysis of multiple dependent variables, separate
LMM analyses were performed for each of the three dependent
variables (Chicken, BRPM, YDMQ). Prior to entering the inde-
pendent and dependent variables for each analysis, the structure of
the data—individuals nested within triads—was specified. Then,
for each analysis, chronological age was entered as a continuous
independent variable, and condition (group vs. sole participant)
was entered as a fixed factor. Additionally, gender and ethnicity
(White vs. non-White) were entered as fixed variables in order to
determine whether these variables moderated age, condition, or
Age ⫻Condition effects.
Age differences in risk taking, risk preference, and risky deci-
sion making. The effect of chronological age on risk taking and
risky decision making was significant, F(1, 284) ⫽18.79, p⬍
.0001, r
effect size
⫽.249, and, F(1, 288) ⫽24.599, p⬍.0001,
r
effect size
⫽.281, respectively. During the risk-taking game,
younger individuals allowed the car to move forward for longer
periods of time after the appearance of the yellow light and were
more likely to restart the car after stopping it. Similarly, younger
individuals were more likely than older participants to select the
risky course of action on the risky decision-making questionnaire.
The effect of chronological age on risk preference was not signif-
icant, however, F(1, 288) ⫽.563, p⫽.465.
Effect of peer presence on risk taking, risk preference, and risky
decision making. We found significant effects of peer presence
on all three measures of risk orientation. Specifically, compared
with those who completed the measures by themselves, partici-
pants who completed the same measures with peers present took
more risks during the risk-taking game, F(1, 284) ⫽15.05, p⬍
4
The adolescent sole participants were not recruited in groups of 3. How-
ever, in order to structure the data from the three age groups as similarly as
possible, triads of adolescent sole participants were created for purposes of
data analyses. The adolescent sole participant sample was subdivided by data
collection site and then further subdivided by gender, such that a female from
a particular community center could only be grouped with another female from
that same community center, or a male from a particular middle school could
only be grouped with another male from that same middle school. This was
done under the assumption that adolescents of the same gender from the same
site would most likely know one another, thus making the triads of adolescent
sole participants as similar to those of undergraduate and adult sole participant
triads as possible.
Table 1
Intercorrelations, Means, and Standard Deviations of Study
Variables
Variable 1 2 3 4 MSD
1. Chronological age — ⫺.243** ⫺.279** ⫺.091 22.77 11.98
2. Risk taking
(Chicken) — .147* .127* 0.00 0.90
3. Risky decision
making (YDMQ) — .331** 2.01 0.56
4. Risk preference
(BRPM) — 1.50 0.50
Note. YDMQ ⫽Youth Decision-Making Questionnaire; BRPM ⫽
Benthin Risk Perception Measure.
*p⬍.05. ** p⬍.01.
629
RISK TAKING IN ADOLESCENCE AND ADULTHOOD

.0001, r
effect size
⫽.224; gave greater weight to the benefits rather
than the costs of risky activities, F(1, 288) ⫽3.662, p⫽.057,
reffect size ⫽.112; and were more likely to select risky courses of
action in the risky decision-making situations, F(1, 288) ⫽6.308,
p⬍.05, r
effect size
⫽.146.
Differential effects of peer presence on risk taking, risk prefer-
ence, and risky decision making as a function of age. The effects
of peer presence varied as a function of age on the risk-taking
measure, F(1, 284) ⫽4.801, p⬍.05, r
effect size
⫽.129, and the
risky decision-making measure, F(1, 288) ⫽4.943, p⬍.05,
r
effect size
⫽.130, but not on the Risk Preference Scale, F(1, 293) ⫽
.284, p⫽.594. As Figure 2 indicates, for example, the magnitude
of the group effect on risk taking was greater among younger
rather than older participants (see Table 2 for means and standard
deviations). The pattern of results was similar with respect to risky
decision making.
The Effects of Gender and Ethnicity
The effects of gender and ethnicity on risk orientation were not
focal issues in the present study. Thus, no hypotheses on the effects of
these variables were generated. Nonetheless, gender and ethnicity
were included in the model in order to determine whether the age,
condition, or Age ⫻Condition interaction effects differed across
males and females, or between White and non-White individuals.
We found few significant gender effects. There were no differ-
ences between males and females on risk taking or risky decision
making, nor were there any significant two-way interaction effects
involving gender on measures of these constructs. Additionally,
we failed to find significant Age ⫻Condition ⫻Gender interac-
tions on any measure of risk orientation. Nevertheless, we did find
main effects of gender and gender differences in age and condition
effects on the measure of risk preference. First, males gave sig-
nificantly greater weight to the benefits of risky decisions than did
females, F(1, 288) ⫽19.961, p⬍.0001, r
effect size
⫽.255. Second,
we found that males weighted the benefits of risky activities more
heavily when in a group than when alone, but that cost– benefit
consideration did not differ substantially between the group and
sole participant conditions among females, F(1, 288) ⫽6.058, p⬍
.05, r
effect size
⫽.144. Finally, we found that among younger
individuals, males weighted the benefits of risky decisions more
heavily than did females but that among older individuals males
and females gave comparable weights to the benefits of risky
decisions, F(1, 288) ⫽11.089, p⬍.01, r
effect size
⫽.193.
In contrast to these limited gender differences, a number of
significant ethnicity effects were identified. First, we found sig-
nificant differences between White and non-White participants
on the measures of risk taking, F(1, 284) ⫽11.67, p⬍.01,
r
effect size
⫽.199, and risky decision making, F(1, 288) ⫽6.645,
p⬍.01, r
effect size
⫽.150. However, the direction of these effects
differed. Although non-White participants engaged in greater risk
taking than did White participants, White participants made more
risky decisions than did non-White participants.
Second, the effects of age on risk taking, risky decision making,
and risk preference differed across White and non-White individ-
Figure 2. Age ⫻Condition interaction on Chicken game, where higher scores indicate more risk taking.
630 GARDNER AND STEINBERG

uals, F(1, 284) ⫽9.03, p⬍.01, r
effect size
⫽.176; F(1, 288) ⫽
4.289, p⬍.05, r
effect size
⫽.121; and, F(1, 288) ⫽3.922, p⬍.05,
r
effect size
⫽.116, respectively (see Table 2). Overall, as noted
earlier, adolescents and youths were more oriented toward risk
than were adults (although the age effect on risk preference mea-
sure did not reach significance). However, within the age groups,
risk taking, risk preference, and risky decision making varied
somewhat as a function of ethnicity. For instance, although there
were negligible differences between White and non-White adults
in risk taking and risk preference, among adolescents non-White
individuals took more risks and demonstrated a greater preference
for risk than did White individuals. Conversely, whereas non-
White adults were slightly more likely to make risky decisions
than were White adults, White adolescents were slightly more
likely to make risky decisions than were non-White adolescents.
Third, White and non-White participants differed in their re-
sponse to peer presence on the measures of risk taking, F(1,
284) ⫽4.383, p⬍.05, r
effect size
⫽.123, and risk preference, F(1,
288) ⫽6.517, p⬍.05, r
effect size
⫽.149 (see Table 2). Peer
presence was associated with greater risk taking and risk prefer-
ence among both White and non-White participants. However,
condition (group vs. sole participant) differences in mean risk-
taking scores were greater among non-White than were those
among White participants. On the risk-preference measure, effect
sizes for condition differences mirrored this pattern (r
effect sizes
⫽
.203 and .108 for non-White and White participants, respectively).
However, effect size estimates indicated the opposite pattern for
risk taking (r
effect sizes
⫽.196 and .233 for non-White and White
participants, respectively). Effect size estimates are reduced by
variability, and although condition differences in mean risk taking
were greater among non-White than White participants, scores
among non-White participants were more variable (see Table 2 for
standard deviations).
Finally, we found significant Age ⫻Condition ⫻Ethnicity
interaction effects on risk taking, F(1, 284) ⫽4.011, p⬍.05,
r
effect size
⫽.118, and risk preference, F(1, 288) ⫽5.961, p⬍.05,
r
effect size
⫽.142; see Table 2). As noted earlier, peer presence had
a greater impact on risk orientation among adolescents and youths
than among adults (although the two-way Age ⫻Condition inter-
action for risk preference did not reach significance). However, the
effects of peer presence on the risk-taking and risk-preference
tendencies of individuals within some, but not all, of the age
groups varied as a function of ethnicity. For instance, among
adults, there were no differences between White and non-White
participants in the effects of peer presence on risk preference.
However, among adolescents, peer effects on risk preference were
greater for non-White than for White participants. A similar pat-
tern emerged on the risk-taking measure. That is, peer effects on
risk taking were greater among non-White than among White
adolescents. However, on this measure, peer effects on adult
risk-taking tendencies were greater for White than for non-White
participants (see Figure 3).
Table 2
Descriptive Statistics for Group and Sole Participant Conditions by Age Group
Age Condition
White Non-White Total
MSDNMSDNMSDN
Risk taking (Chicken)
Adolescent Sole ⫺.164 .612 22 ⫺.035 .722 29 ⫺.097 .669 52
Group .140 .886 25 .907 1.300 29 .552 1.182 54
Youth Sole ⫺.258 .729 24 ⫺.091 .666 27 ⫺.170 .694 51
Group .139 .848 30 .289 .893 20 .199 .860 50
Adult Sole ⫺.367 .387 22 ⫺.316 .620 19 ⫺.343 .502 41
Group ⫺.080 .566 24 ⫺.335 1.167 30 ⫺.221 .949 54
Risky decision making (YDMQ)
Adolescent Sole 2.127 .511 22 1.862 .571 29 1.977 .551 52
Group 2.072 .519 25 2.021 .522 29 2.044 .516 54
Youth Sole 2.142 .564 24 2.074 .448 27 2.106 .502 51
Group 2.506 .477 32 2.282 .358 22 2.428 .455 50
Adult Sole 1.655 .339 22 1.962 .758 19 1.781 .581 41
Group 1.625 .397 24 1.720 .497 30 1.678 .453 54
Risk preference (BRPM)
Adolescent Sole 1.391 .335 22 1.372 .286 29 1.385 .304 52
Group 2.072 .519 25 2.021 .522 29 1.593 .764 54
Youth Sole 2.142 .564 24 2.074 .448 27 1.451 .447 51
Group 1.819 .600 32 1.555 .365 22 1.724 .536 50
Adult Sole 1.427 .317 22 1.368 .354 19 1.400 .332 41
Group 1.425 .355 24 1.393 .358 30 1.407 .354 54
Note. Chicken means are based on standardized scores. YDMQ ⫽Youth Decision-Making Questionnaire;
BRPM ⫽Benthin Risk Perception Measure.
631
RISK TAKING IN ADOLESCENCE AND ADULTHOOD

Discussion
Between adolescence and adulthood there is a significant de-
cline in both risk taking and risky decision making. In addition, our
findings suggest that, in some situations, individuals may take
more risks, evaluate risky behavior more positively, and make
more risky decisions when they are with their peers than when they
are by themselves. Most importantly, the effects of peer presence
on both risk taking and risky decision making vary as a function of
age. That is, although the sample as a whole took more risks and
made more risky decisions in groups than when alone, this effect
was more pronounced during middle and late adolescence than
during adulthood. Thus, relative to adults, adolescents are more
susceptible to the influence of their peers in risky situations.
The methodological strengths of this study provide good reason
to feel confident about the internal validity of the findings. First, as
an experiment that uses random assignment, we were able to
control individuals’ exposure to peers. Second, whereas the only
previous developmental comparison of peer effects on risk taking
(Hensley, 1977) relied on a hypothetical decision-making ques-
tionnaire and used risk acceptance as a proxy for risk taking, the
battery of measures in the present study included not only hypo-
thetical decision-making questionnaires but also a behavioral mea-
sure of risk taking that required participants to make actual deci-
sions about how much risk to take in a situation that closely
mirrors one faced in everyday life—whether to “run” a yellow
light and continue through an intersection. Third, the use of
friends, or at least familiar individuals (in the case of the adoles-
cents), helped to create a more ecologically valid social context
than those of many studies of group behavior (e.g., Vidmar, 1970;
Wallach & Kogan, 1965; Yinon et al., 1975). Everyday group
decision-making situations generally involve friends or acquain-
tances, and laboratory studies that do not use such groups may not
capture the dynamics of real-life group decision making.
It is also necessary to recognize several of the study’s limita-
tions. First, although the driving game Chicken is a closer approx-
imation to real-life risk-taking situations than are the typical
decision-making questionnaires used in most research of this sort,
no laboratory task can adequately simulate real life. No matter how
realistic the task, it is difficult to determine whether participants’
performance in the laboratory is an accurate representation of their
real-world behavior.
Second, different recruitment procedures were used for the
adolescents versus the youths and adults. The youths and adults
came to the sessions in groups of 3 friends, but the adolescents
were assigned to groups of 3 (although all groups of 3 were made
up of individuals who were from the same classroom, camp, or
community-center program, and who were familiar with one an-
other). Thus, it is conceivable that the older individuals knew one
another better than did the adolescents. However, we believe that
differences in familiarity among the age groups were minimal and
that any effects of peer familiarity of behavior resulted in a more
conservative test of our central hypothesis (i.e., that, relative to
adults, adolescents should be more easily swayed toward risky
behavior by their friends). Although some contradictory findings
do exist (e.g., Leary et al., 1994), there is evidence to suggest that
the effects of peer familiarity on behavior may either be negligible,
Figure 3. Age ⫻Condition ⫻Ethnicity Interaction on Chicken game, where higher scores indicate more risk
taking. W⫽White participants; NW ⫽non-White participants.
632 GARDNER AND STEINBERG

or may be stronger when in the company of friends versus ac-
quaintances or strangers. For example, in a study of impression
management among young adults, Bohra and Pandey (1984) found
very few differences in the attempts of participants to manage the
impressions of friends versus strangers. Moreover, in the few cases
in which differences were found (e.g., use of other enhancement
strategies), interactions with friends were generally more likely to
elicit the use of impression management strategies than were
interactions with strangers. Similarly, Gardner and Martinko
(1988) found that participants in a study of school principals were
more likely to use impression management strategies (e.g., other
enhancement, apologies) when interacting with more familiar, as
opposed to less familiar, individuals. Finally, in a study that
examined the relation between familiarity and willingness to exert
peer pressure (both antisocial and prosocial) among adolescents,
McPhee (1996) found that participants were more likely to exert
pressure on friends than on acquaintances. Thus, if differences
among the age groups in triad familiarity affected our results in
any measurable way, we believe that the adolescent participants,
who completed the battery of measures with acquaintances (as
opposed to self-selected friends), should have demonstrated group
induced shifts toward risk taking that were no greater than those
observed among the adults. But, overall, this was not the case.
Finally, it is conceivable that members of the three age groups
differed in their prior experience with the subject matter of the
risk-orientation measures. Thus, age differences in performance on
the risky decision-making and risk-taking measures might be con-
strued as an artifact of differences between the age groups in prior
experience. However, any differences in experience with the con-
tent of the risky decision-making questionnaire were likely limited
to the individual dilemmas. Thus, although a given dilemma might
have been relatively more familiar to a particular age group,
overall, the items were balanced in such a way that no one age
group should have been more familiar with all five dilemmas than
any other age group. With respect to the risk-taking measure,
Chicken, it is likely that the youths and adults had more first hand
experience with driving than the adolescents. However, Chicken is
a video game played from a third-person perspective not a driving
simulation experienced from a first-person perspective. Although
adolescents may have limited experience with driving, they have
ample experience with video and computer games Additionally, by
adolescence, individuals have spent a great deal of time riding in
cars and are surely familiar with the potential consequences of
failing to follow traffic signals. Nonetheless, in order to ensure
equal familiarity with the potential consequences of running a
yellow light, all participants observed a demonstration round prior
to playing the game in which they watched the animated car crash
after running a yellow light. Thus, we believe that differences in
first-hand driving experience had minimal impact on performance.
It is also important to note the few instances in which our
hypotheses were not supported. Specifically, we failed to find
significant age main effects or two-way Age ⫻Condition inter-
action effects on the risk preference measure. In conceptualizing
the study, we assumed that those who gave greater weight to the
benefits versus the costs of risky decisions should be more likely
to take risks. As noted earlier, there are a number of studies that
have found strong correlations between cost– benefit consideration
and risk taking (e.g., Goldberg & Fischoff, 2000; Horvath &
Zuckerman, 1993; Thorton, Gibbons, & Gerrard, 2002). However,
the samples in these studies were composed primarily of adults,
and there is some evidence to suggest that measures of risk
preference may not predict risk taking in the same way among
adolescents as among adults. Indeed, despite findings that adoles-
cents are more likely than are adults to engage in risky behavior,
several studies suggest that adolescents are relatively similar to
adults in their ability to recognize the risks and benefits of their
actions. For example, Beyth-Marom et al. (1993) found few dif-
ferences between adolescents and adults in the spontaneous men-
tion of the costs and benefits associated with several risky actions.
This has prompted some to argue that age differences in risky
behavior may be better accounted for by differences in psycho-
social functioning than by differences in more cognitive aspects of
risk orientation, such as risk preference (Cauffman, 1996; Cauff-
man & Steinberg, 2000; Steinberg & Cauffman, 1996). In this
respect, our failure to find age-related differences in individuals’
cost– benefit appraisals is not entirely surprising.
We did find some interesting gender differences in risk prefer-
ence, however. Specifically, males, particularly at younger ages,
were more likely than were females to weigh the benefits of risky
activities over the costs. Additionally, peer effects on benefit
versus cost consideration were greater among males than among
females. Although we did not explicitly predict these gender
differences, our findings are consistent with several previous stud-
ies. For instance, Parsons, Halkitis, Bimbi, and Borkowski (2000)
found that, among young adults, males reported more benefits and
fewer risks when asked about the consequences of risky behaviors.
Additionally, Brown et al. (1986) found that, at least among
adolescents, males are more susceptible to peer influence than are
females in antisocial or risky situations. Nonetheless, it is inter-
esting that these gender-related differences in risk– benefit consid-
eration did not translate into gender differences on the more direct
measures of risk taking or risky decision making.
We also found differences in risk orientation as a function of
ethnicity. First, we found differences between White and non-
White participants in risk taking, risk preference, and risky deci-
sion making—particularly among adolescents (ethnic differences
in risk orientation among adults were small to negligible). How-
ever, the direction of these ethnic group differences varied across
measures. Whereas non-White adolescents demonstrated greater
risk taking and risk preference than did White adolescents, White
adolescents demonstrated greater risky decision making than did
non-White adolescents. This is not entirely surprising given that
prior studies have identified differences in the direction of ethnic-
ity effects for different risk behaviors. For example, there is
evidence to suggest that minority adolescents (particularly African
Americans) are more likely than are White adolescents to engage
in risky sexual behavior (e.g., Koniak-Griffin & Brecht, 1995;
Neumark-Sztainer et al., 1996; Santelli, Lowry, Brener, & Robin,
2000) and to participate in delinquent activities (e.g., Blum et al.,
2000; Hawkins, Laub, & Lauritsen, 1998; Piquero & Buka, 2002).
However, there is also evidence to suggest that White adolescents
may take more risks than may non-White adolescents when sub-
stance use is the behavior of interest. Specifically, a number of
studies have found that White adolescents engage in more alcohol
and tobacco use than adolescents from many non-White ethnic
groups (Best et al., 2001; Blum et al., 2000; Brannock, Schandler,
& Oncley, 1990; Douglas & Collins, 1997). Accordingly, re-
searchers studying adolescent risk taking must exercise caution in
633
RISK TAKING IN ADOLESCENCE AND ADULTHOOD

asserting that ethnic group differences on particular risk measures
reflect more general patterns of ethnic group differences in risk
taking overall.
Second, we found that peer effects on risk orientation varied
across ethnic groups. Specifically, we found that the effects of peer
presence on risk preference were greater among non-White than
among White participants. Though effect size estimates were not
entirely consistent, inspection of mean scores suggests that the
same pattern of ethnic group differences may also exist for risk
taking. However, if it is, in fact, the case that non-White, relative
to White, individuals are more susceptible to peer influence in
risky situations, this ethnic group difference appears to be largely
limited to adolescence. Although peer effects on risk taking and
risk preference were greater among non-White than among White
adolescents, non-White adults demonstrated levels of resistance to
peer influence that were equal to or greater than those demon-
strated by White adults. Though few studies have examined ethnic
group differences in the development of resistance to peer influ-
ence, there is tentative evidence to support the finding that, relative
to White adolescents, non-White adolescents may be more suscep-
tible to the influence of others when in risky situations. For
instance, Zimmerman, Sprecher, Langer, & Holloway (1995)
found that African American and Hispanic adolescent females
were slightly less confident in their ability to refuse unwanted sex
than White adolescent females. However, to our knowledge, no
prior research has directly examined the possibility that ethnic
group differences in susceptibility to peer influence in risky situ-
ations may diminish as individuals move into adulthood. Thus,
further research is necessary in order to both replicate and explain
this finding.
In conclusion, it appears that differences in rates of group risk
taking among adolescents versus adults are not simply the product
of differences in the amount of time teenagers and adults spend
with peers but are instead the result of age differences in individ-
uals’ orientation toward risky behavior when in the presence of
friends. Moreover, our results suggest that the psychosocial capac-
ities that undergird the ability to resist peer pressure may continue
to develop throughout late adolescence and into early adulthood.
Thus, interventions aimed at reducing risky behavior among ado-
lescents and young adults—particularly those from ethnic minority
groups— ought to focus some attention on increasing individuals’
resistance to peer influence. For reasons not yet understood, the
presence of peers makes adolescents and youth, but not adults,
more likely to take risks and more likely to make risky decisions.
References
Andrews, J. A., Tildesley, E., Hops, H., & Li, F. (2002). The influence of
peers on young adult substance use. Health Psychology, 21, 349 –357.
Arnett, J. (1992). Reckless behavior in adolescence: A developmental
perspective. Developmental Review, 12, 339 –373.
Benthin, A., Slovic, P., & Severson, H. (1993). A psychometric study of
adolescent risk perception. Journal of Adolescence, 16, 153–168.
Berndt, T. J. (1979). Developmental changes in conformity to peers and
parents. Developmental Psychology, 15, 608 – 616.
Best, D., Rawaf, S., Rowley, J., Floyd, K., Manning, V., & Strang, J.
(2001). Ethnic and gender differences in drinking and smoking among
London adolescents. Ethnicity & Health, 6, 51–57.
Beyth-Marom, R., Austin, L., Fischoff, B., Palmgren, C., & Jacobs-
Quadrel, M. (1993). Perceived consequences of risky behaviors: Adults
and adolescents. Developmental Psychology, 29, 549 –563.
Blascovich, J., & Ginsburg, G. P. (1974). Risky shifts and gambling:
What’s at stake? Personality and Social Psychology Bulletin, 1, 246 –
248.
Blascovich, J., Ginsburg, G. P., & Howe, R. C. (1975). Blackjack and the
risky shift, II: Monetary stakes. Journal of Experimental Social Psychol-
ogy, 11, 224 –232.
Blascovich, J., Veach, T. L., & Ginsburg, G. P. (1973). Blackjack and the
risky shift. Sociometry, 36, 42–55.
Blum, R. W., Beuhring, T., Shew, M. L., Bearinger, L. H., Sieving, R. E.,
& Resnick, M. D. (2000). The effects of race/ethnicity, income, and
family structure on adolescent risk behaviors. American Journal of
Public Health, 90, 1879 –1884.
Bohra, K. A., & Pandey, J. (1984). Ingratiation toward strangers, friends,
and bosses. The Journal of Social Psychology, 122, 217–222.
Brannock, J. C., Schandler, S. L., & Oncley, P. R. (1990). Cross-cultural
and cognitive factors examined in groups of adolescent drinkers. Journal
of Drug Issues, 20, 427– 442.
Brown, B. (2004). Adolescents’ relationships with peers. In R. Lerner & L.
Steinberg (Eds.), Handbook of adolescent psychology (2nd ed., pp.
363–394). New York: Wiley.
Brown, B. B., Clasen, D. R., & Eicher, S. A. (1986). Perceptions of peer
pressure, peer conformity dispositions, and self-reported behavior
among adolescents. Developmental Psychology, 22, 521–530.
Cauffman, E. (1996). Maturity of judgment: Psychosocial factors in ado-
lescent decision making (Doctoral dissertation, Temple University,
1986). Dissertation Abstracts International, 57(9-B), 5943.
Cauffman, E., & Steinberg, L. (2000). Researching adolescents’ judgment
and culpability. In T. Grisso & R. G. Schwartz (Eds.), Youth on trial: A
developmental perspective on juvenile justice (pp. 325–343). Chicago:
The University of Chicago.
Cohen, S. L., & Ruis, C. B. (1974). Wary shift or risky shift? Bulletin of
the Psychonomic Society, 3, 214 –216.
Douglas, K. A., & Collins, J. L. (1997). Results from the 1995 National
College Risk Behavior Survey. Journal of American College Health, 46,
55– 66.
Erickson, M., & Jensen, G. F. (1977). Delinquency is still a group behav-
ior! Toward revitalizing the group premise of the sociology of deviance.
Journal of Criminal Law and Criminology, 68, 262–273.
Fischhoff, B. (1992). Risk taking: A developmental perspective. In J. F.
Yates (Ed.), Risk Taking Behavior (pp. 133–162). Oxford, England:
Wiley.
Ford, M., Wentzel, K., Wood, D., Stevens, E., & Siesfeld, G. A. (1990).
Processes associated with integrative social competence: Emotional and
contextual influences on adolescent social responsibility. Journal of
Adolescent Research, 4, 405– 425.
Fromme, K., Stroot, E., & Kaplan, D. (1993). The Comprehensive Effects
of Alcohol questionnaire: Development and psychometric evaluation of
a new expectancy questionnaire. Psychological Assessment, 5, 19 –26.
Furby, L., & Beyth-Marom, R. (1992). Risk taking in adolescence: A
decision-making perspective. Developmental Review, 12, 1– 44.
Gardner, W. L., & Martinko, M. J. (1988). Impression management: An
observational study linking audience characteristics with verbal self-
presentations. Academy of Management Journal, 31, 42– 65.
Goldberg, J., & Fischoff, B. (2000). The long-term risks in the short-term
benefits: Perceptions of potentially addictive activities. Health Psychol-
ogy, 19, 299 –303.
Hawkins, D. F., Laub, J. H., & Lauritsen, J. L. (1998). Race, ethnicity, and
serious juvenile offending. In R. Loeber & D. P. Farrington (Eds.),
Serious and violent juvenile offenders: Risk factors and successful
interventions (pg. 30 – 46). Thousand Oaks, CA: Sage
Hensley, W. E. (1977). Probability, personality, age, and risk taking. The
Journal of Psychology, 95, 139 –145.
Hogg, M. A., Turner, J. C., & Davidson, B. (1990). Polarized norms and
634 GARDNER AND STEINBERG

social frames of reference: A test of the self-categorization theory of
group polarization. Basic & Applied Social Psychology, 11, 77–100.
Horvath, P., & Zuckerman, M. (1993). Sensation seeking, risk appraisal,
and risky behavior. Personality and Individual Differences, 14, 41–52.
Kogan, N., & Wallach, M. A. (1967). Group risk taking as a function of
members’ anxiety and defensiveness levels. Journal of Personality, 35,
50 – 63.
Koniak-Griffin, D., & Brecht, M. (1995). Linkages between sexual risk
taking, substance use, and AIDS knowledge among pregnant adolescents
and young mothers. Nursing Research, 44, 340 –346.
Lamm, H. (1967). Will an observer advise higher risk taking after hearing
a discussion of the decision problem? Journal of Personality and Social
Psychology, 6, 467– 471.
Lamm, H., Trommsdorff, G., & Rost-Schaude, E. (1972). Self-image,
perception of peers’ risk acceptance and risky shift. European Journal of
Social Psychology, 2, 255–272.
Lavery, B., Siegel, A. W., Cousins, J. H., & Rubovitz, D. S. (1993).
Adolescent risk taking: An analysis of problem behavior in problem
children. Journal of Experimental Child Psychology, 55, 277–294.
Leary, M. R., Nezlek, J. B., Downs, D., Radford-Davenport, J., Martin, J.,
& McMullen, A. (1994). Self-presentation in everyday interactions:
Effects of target familiarity and gender composition. Journal of Person-
ality and Social Psychology, 67, 664 – 673.
McBride, D. C., Weatherby, N. L., Inciardi, J. A., & Gillespie, S. A.
(1999). AIDS susceptibility in a migrant population: Perception and
behavior. Substance Use and Misuse, 34, 633– 652.
McPhee, J. H. (1996). Influence strategies in young adolescent dyads.
Dissertation Abstracts International, 57(02), 1468B.
Neumark-Sztainer, D., Story, M., French, S., Cassuto, N., Jacobs, D. R., &
Resnick, M. D. (1996). Patterns of health-compromising behaviors
among Minnesota adolescents: Sociodemographic variations. American
Journal of Public Health, 86, 1599 –1606.
Parsons, J. T., Halkitis, P. N., Bimbi, D., & Borkowski, T. (2000). Per-
ceptions of the benefits and costs associated with condom use and
unprotected sex among late adolescent college students. Journal of
Adolescence, 23, 377–391.
Pilkonis, P. A., & Zanna, M. P. (1973). The choice-shift phenomenon in
groups: Replication and extension. Representative Research in Social
Psychology, 4, 36 – 47.
Piquero, A. R., & Buka, S. L. (2002). Linking juvenile and adult patterns
of criminal activity in the Providence cohort of the National Collabora-
tive Perinatal Project. Journal of Criminal Justice, 30, 259 –272.
Pruitt, D. G., & Teger, A. I. (1969). The risky shift in group betting.
Journal of Experimental Social Psychology, 5, 115–126.
Santelli, J. S., Lowry, R., Brener, N. D., & Robin, L. (2000). The associ-
ation of sexual behaviors with socioeconomic status, family structure,
and race/ethnicity among US adolescents. American Journal of Public
Health, 90, 1582–1588.
Scott, E., Reppucci, N. D., & Woolard, J. (1995). Evaluating adolescent
decisionmaking in legal contexts. Law and Human Behavior, 19, 221–
244.
Sheldrick, R. C. (2004). Social networks and degree of psychopathy among
adolescent offenders (Doctoral dissertation, Temple University, 1990).
Dissertation Abstracts International, 65(3-B), 1564.
Singer, M., Dai, H., Weeks, M. R., & Malave, D. (1998). AIDS risk
perception among women drug users in Hartford, CT. Women, Drug
Use, and HIV Infection, 27, 67– 85.
SPSS, Incorporated. (2005). Linear mixed-effects modeling in SPSS: An
introduction to the MIXED procedure (LMEMWP-0305). Chicago, IL:
Author.
Steinberg, L. (2004). Risk taking in adolescence: What changes, and why?
Annals of the NewYork Academy of Sciences, 1021, 51–58.
Steinberg, L., & Cauffman, E. (1996). Maturity of judgment in adoles-
cence: Psychosocial factors in adolescent decision making. Law and
Human Behavior, 20, 249 –272.
Steinberg, L., & Scott, E. (2003). Less guilty by reason of adolescence:
Developmental immaturity, diminished responsibility, and the juvenile
death penalty. American Psychologist, 58, 1009 –1018.
Steinberg, L., & Silverberg, S. B. (1986). The vicissitudes of autonomy in
early adolescence. Child Development, 57, 841– 851.
Thorton, B., Gibbons, F. X., & Gerrard, M. (2002). Risk perception and
prototype perception: Independent processes predicting risk behavior.
Personality and Social Psychology Bulletin, 28, 986 –999.
Vidmar, N. (1970). Group composition. Journal of Experimental Social
Psychology, 6, 153–166.
Vinokur, A. (1971). Review and theoretical analysis of the effects of group
processes upon individual and group decisions involving risk. Psycho-
logical Bulletin, 76, 231–250.
Wallach, M. A., & Kogan, N. (1965). The roles of information, discussion,
and consensus in group risk taking. Journal of Experimental Social
Psychology, 1, 1–19.
Yinon, Y., Jaffe, Y., & Feshbach, S. (1975). Risky aggression in individ-
uals and groups. Journal of Personality and Social Psychology, 31,
808 – 815.
Zaleska, M. (1974). The effects of discussion on group and individual
choices among bets. European Journal of Social Psychology, 4, 229 –
250.
Zimmerman, R. S., Sprecher, S., Langer, L. S., & Holloway, C. D. (1995).
Adolescents’ perceived ability to say “no” to unwanted sex. Journal of
Adolescent Research, 10, 383–399.
Zimring, F. E. (1998). American youth violence. Oxford, England: Oxford
University Press.
Received April 6, 2004
Revision received November 12, 2004
Accepted November 16, 2004 䡲
635
RISK TAKING IN ADOLESCENCE AND ADULTHOOD