Introduction

Teenagers are known to engage in more risky behavior than children or adults: adolescents are more likely than older or younger individuals to binge drink, smoke cigarettes, have casual sex partners, engage in violent and other criminal behavior, and to be involved in fatal or serious automobile crashes, the majority of which are caused by risky driving or driving under the influence of alcohol (Steinberg, Albert, Cauffman, Banich, Graham & Woolard, 2008). Many experts agree that these preventable behaviors present the greatest threat to the well-being of young people in industrialized societies.

Significantly, adolescent risk taking differs from that of adults in its social context as well as its incidence. One of the hallmarks of adolescent risk taking is that it is much more likely than that of adults to occur in the presence of peers, as evidenced in studies of reckless driving (Simons-Morton, Lerner & Singer, 2005), substance abuse (Chassin, Hussong & Beltran, 2009), and crime (Zimring, 1998). Relatively greater adolescent risk taking in the presence of peers could be explained simply by the fact that adolescents spend more time with friends than do adults. However, recent experimental evidence (Gardner & Steinberg, 2005; O'Brien, Albert, Chein & Steinberg, in press) indicates that adolescents’ decisions are directly influenced by the mere presence of peers. Gardner and Steinberg (2005), for instance, examined risk taking in adolescents, college undergraduates, and adults who were randomly assigned to engage in a simulated driving task alone or in the presence of two friends. They found that adolescents (and undergraduates to a lesser extent), but not adults, took a substantially greater number of risks when observed by peers.

Many research groups (Casey, Getz & Galvan, 2008; Luna, Padmanabhan & O'Hearn, 2010; Somerville, Jones & Casey, 2010; Steinberg, 2008; Van Leijenhorst, Moor, de Macks, Rombouts, Westenberg & Crone, 2010a; see also Ernst, Pine & Hardin, 2006) have posited that adolescents’ relatively greater propensity toward risky behavior reflects the joint contribution of two brain systems that affect decision-making: (i) an incentive processing system involving the ventral striatum (VS; including the nucleus accumbens, NAcc) and the orbitofrontal cortex (OFC), among other regions, which biases decision-making based on the valuation and prediction of potential rewards and punishments; and (ii) a cognitive control system, including the lateral prefrontal cortex (LPFC), which supports goal-directed decision-making by keeping impulses in check and by providing the mental machinery needed for deliberation regarding alternative choices.

Neuroimaging studies conducted in both adult and adolescent populations show that these systems contribute to decision-making in an interactive fashion, with impulsive or risky choices often coinciding with the increased engagement of incentive processing regions (Ernst, Nelson, McClure, Monk, Munson, Eshel, Zarahn, Leibenluft, Zametkin, Towbin, Blair, Charney & Pine, 2004; Hare, Camerer & Rangel, 2009; Kuhnen & Knutson, 2005; Matthews, Simmons, Lane & Paulus, 2004; McClure, Laibson, Loewenstein & Cohen, 2004) and the decreased involvement of cognitive control activity (e.g. Eshel, Nelson, Blair, Pine & Ernst, 2007; Fecteau, Knoch, Fregni, Sultani, Boggio & Pasual-Leone, 2007).

Both of these brain systems undergo considerable modification during adolescence, but on different timetables. The incentive processing system evinces dramatic remodeling in early adolescence, particularly with respect to the distribution and density of dopamine receptors (Laviola, Pascucci & Pieretti, 2001). Evidence suggests that changes in the mesocorticolimbic dopamine system result in heightened sensitivity to rewards (Spear, 2009). Regions in this brain pathway (especially the NAcc), which is implicated in the valuation and prediction of potential rewards (Breiter & Rosen, 1999; Delgado, 2007; O'Doherty, 2004; Schultz, 2010), have been found in several recent functional neuroimaging studies to show especially heightened activation during adolescence in response to reward-relevant cues and reward anticipation (Ernst, Nelson, Jazbec, McClure, Monk, Leibenluft, Blair & Pine, 2005; Ernst, Romeo & Andersen, 2009; Galvan, Hare, Parra, Penn, Voss, Glover & Casey, 2006; Geier, Terwilliger, Teslovich, Velanova & Luna, 2010; Van Leijenhorst, Zanolie, Van Meel, Westenberg, Rombouts & Crone, 2010b). Compellingly, Galvan and colleagues (Galvan, Hare, Voss, Glover & Casey, 2007) found that the degree of NAcc activity during reward anticipation was correlated with adolescents’ self-reported risk taking, providing convergent evidence that adolescents’ heightened reward sensitivity contributes to elevated real-world risk taking.

Brain regions involved in cognitive control undergo comparatively gradual and protracted maturation, involving reductions in gray matter density and increases in myelination, from preadolescence through at least the mid-20s (Asato, Terwilliger, Woo & Luna, 2010; Giedd, 2008). This maturation is thought to support improvements in executive abilities such as response inhibition (Luna ), strategic planning (Luciana, Collins, Olson & Schissel, 2009), impulse regulation (Steinberg ), and flexible rule use (Crone, Donohue, Honomichl, Wendelken & Bunge, 2006).

We propose that adolescents’ especially heightened propensity to take risks when with peers may derive from the maturational imbalance between these competing brain systems. Specifically, given the elevated reward value of peer interactions in adolescence (Blakemore, 2008; Spear, 2009), the presence of peers may sensitize the incentive processing system to respond to cues signaling the potential rewards of risky behavior. In the context of an immature capacity to down-regulate reward system outputs through control signaling, this reward-sensitive motivational state may bias adolescents’ decisions toward greater risk taking. At the neural level, the influence of peers on adolescents’ decisions may be manifested in the heightened activation of regions associated with reward valuation. Alternatively, peer presence may have a direct effect on cognitive control processes, and hence be reflected as altered activity within regions associated with impulse regulation.

To test these alternative predictions, we measured brain activity in adolescent, young adult, and adult participants as they made a series of decisions in a simulated driving game. In the game, participants rendered decisions about whether to stop at a given intersection, or to run through the intersection and chance a collision with another vehicle, with the goal of reaching the end of a track as quickly as possible to maximize a monetary reward. Risky decisions offered the potential payoff of experiencing no delay at the intersection, but also the potentially costly consequence of a crash, which added significantly to the delay. Social context was manipulated by having each participant play the game alone and while being observed by peers.

Method

Participants

Data from 40 subjects (14 adolescents – eight female, ages 14–18 years, M = 15.7, SD = 1.5; 14 young adults – seven female, ages 19–22 years, M = 20.6, SD = 0.9; and 12 adults – six female, ages 24–29 years, M = 25.6, SD = 1.9) were included in analyses. Informed consent was obtained for each subject according to a protocol approved by the institutional review boards of Princeton and Temple Universities, and each received monetary compensation for their participation.

Procedure

Task design

The Stoplight task (Figure 1) is a simple driving task in which subjects control the progression of a vehicle along a straight track, from a driver’s point of view. Subjects completed four rounds of the task; two in the first social condition and two in the second social condition. Each round used a track with 20 intersections (treated as separate trials), which took under 6 minutes to traverse (dependent on subjects’ choices and providence). At each intersection subjects rendered a decision (by button press) about whether or not to brake as the vehicle approached a changing traffic signal (which cycled from green to yellow to red). As the vehicle approached the intersection, the traffic signal turned yellow, and the subject decided whether to chance a possible crash in the intersection (GO decision), or to brake and wait for the light to return to green (STOP decision). Importantly, both the timing of the traffic signals and the probability of a crash in the associated intersections were varied so as to be unpredictable by the participant. Risk taking (i.e. not braking for the yellow light) was encouraged by offering monetary incentives for completing the course in a timely fashion. Successfully traveling through an intersection without braking saved time, whereas braking and waiting for the signal to turn green again was associated with a time delay. However, if the participant did not brake and a crash ensued, the loss of time was even greater than if the participant were to brake and wait for the light. Behavioral data from the scanner were acquired and temporally aligned to fMRI acquisitions using E-Prime (Psychology Software Tools, Pittsburgh, PA), interfaced with an LCD display, headphones, and a key-press unit. Additional task details are provided in the online Supporting Information.

Figure 1

The Stoplight driving game. In each run of the Stoplight driving game, participants attempted to reach the end of a straight track as quickly as possible. The 20 intersections of the track were treated as separate trials, and were spaced by a variable distance (ITI). At each intersection, participants rendered a decision to either stop the vehicle (STOP) or to take a risk and run the traffic light (GO). Stops resulted in a short delay. Successful risk taking resulted in no delay. Unsuccessful risk taking resulted in a crash, and a relatively long delay. Subjects completed four runs of the task (two in each social condition).

Manipulation of social context

When reporting for the study, all participants were asked to bring two same-age (within 2 years of their own age), same-sex, friends. Social context was manipulated as a within-subjects variable, with counterbalancing for order across subjects. In an ALONE condition, participants completed the task with no observers. In a PEER condition, participants were informed that their friends were going to observe their actions from a monitor in the neighboring scanner control room. The change of social context was a surprise manipulation.1 In the break prior to the onset of the PEER condition, and in the breaks following each subsequent functional scan of this condition, the peers were asked to communicate with the scanned subject via the scanner's intercom system. In order for the interaction to be ecologically valid, the peers were permitted to speak authentically while informing the scanned participant of their presence, demonstrating their ability to observe task performance on the monitor, and communicating that they had made predictions about the scanned participant’s pending performance. The peers were carefully instructed to make these specific points during the interaction, and to avoid comments that might explicitly or intentionally bias behavior.

Self-report questionnaires

Following the fMRI session, subjects were also asked to complete a series of self-report questionnaires. Scores obtained from these questionnaires were used to assess individual differences in impulsivity (shortened form of the Barratt Impulsiveness Scale, Version 1; Patton, Stanford & Barratt, 1995), sensation seeking (assessed using a subset of six items from the Zuckerman Sensation Seeking Scale; Zuckerman, Eysenck & Eysenck, 1978), and resistance to peer influence (assessed using the RPI scale; Steinberg & Monahan, 2007). Self-report data from individual questionnaires were missing or incomplete for some subjects.

fMRI Data Acquisition

Subjects were scanned using a head-only 3 Tesla Siemens (Erlangen, Germany) Allegra magnet located at Princeton University. A T1-weighted magnetization-prepared rapid-acquisition gradient echo (MPRAGE) scan collected in the sagittal plane provided high resolution (1mm³) 3D structural imaging of the whole head, for use in subject coregistration. Each functional scan of Stoplight task performance included 195 acquisitions collected with a whole brain T2*-weighted echoplanar imaging (EPI) sequence (TR = 2.0s, TE = 30ms, flip = 70°, 33 slices, 3 mm slice thickness with 1 mm gap, 3 × 3mm in-plane resolution).2

fMRI Data Analysis

fMRI data analysis was performed using AFNI (Cox, 1996). Preprocessing of functional data consisted of several stages, beginning with a six-parameter rigid-body motion correction in three dimensions, and coregistration of the corrected functional and structural images. FMRI acquisitions requiring motion correction greater than 4 mm of translation or 4 degrees of rotation in any dimension were censored out of the dataset. Data were then interpolated to correct for slice acquisition order effects, normalized to Montreal Neurological Institute (MNI) coordinates, and smoothed with a 6 mm full-width at half maximum Gaussian kernel. The corrected fMRI data from each individual subject were analyzed in an event-related fashion using a general linear model (GLM). To explore the neurobiological correlates of age differences in the impact of social context on risky decision-making, we examined BOLD fMRI activity associated with the decision-making period of the Stoplight task.3 Specifically, event-related time-series indexing the moment when the traffic signal at each intersection cycled from green to yellow were created, and the resulting time-series were convolved with a canonical model of the hemodynamic response function (Boynton, Engel, Glover & Heeger, 1996). These event-related time-series were combined according to the social context in which they occurred to form separate PEER and ALONE condition regressors. These two regressors were entered into a single GLM equation to identify voxels exhibiting fMRI signal changes at the point of decision-making in each social context. The GLM equation also included covariates of non-interest that captured noise due to variation in run-based means, linear and quadratic scanner drift, and estimated motion. To further reduce noise, voxel-wise parameter estimates obtained from individual subjects were further subjected to outlier detection (> 2.5 SD) and removal prior to group testing. The voxel-wise parameter estimates (beta coefficients) obtained from individual subjects were entered into a group random-effects analysis in order to identify regions exhibiting main and interactive effects for age and social context. These group analyses were based on a two-way repeated measures ANOVA, treating age group as a between-subjects factor and social context as a within-subjects (repeated) factor. Additional planned, pair-wise contrasts were conducted to further clarify the differences driving significant main and interactive effects. Group-wise statistical maps obtained for all effects were constrained to an anatomical mask including cortical and subcortical gray matter, and were corrected for multiple comparisons using a voxel-wise probability threshold (p < .005) and contiguity requirement (seven adjacent voxels) that resulted in a family-wise error (FWE) rate below .05, based on Monte Carlo simulations.

Results

Behavioral results

We assessed behavioral sensitivity to social context by contrasting decision-making in the PEER and ALONE conditions. As in Gardner and Steinberg (2005), we found that adolescents and older participants behaved comparably when tested alone, but that performance in the adolescent group was sensitive to social context. Although the age by social context interaction did not reach statistical significance at this sample size [F(2, 38) = 2.66, p = .084), only adolescents took significantly more risks when observed by peers than when alone (Figure 2), as evidenced by a significantly increased number of GO decisions [t(13) = 2.16, p = .025, one-tailed] and subsequent crashes [t(13) = 4.06, p < .001, one-tailed]. Additional behavioral results are provided in the online Supporting Information.

Figure 2

Stoplight task performance. Mean (a) percentage of risky decisions and (b) number of crashes for adolescent, young adult, and adult participants when playing the Stoplight task alone and with a peer audience. Error bars indicate standard error of the mean.

In order to assess the construct validity for the scanner implementation of the Stoplight task, we examined correlations between task performance and subject’s self-report responses. Variation in Stoplight performance may reflect inter-subject differences in both sensation seeking (i.e. by impelling a player to take risks) and inhibitory control (i.e. by moderating a player’s ability to regulate braking). However, a previous large-scale individual differences study found that variation in Stoplight task performance was significantly predicted by self-reported sensation seeking, but not self-reported impulsivity (Steinberg ). Despite the much smaller sample size, the present results replicate these earlier findings. As in the larger-scale study, we conducted a regression analysis in which self-report measures of sensation seeking and impulsivity were entered as simultaneous predictors of risky driving in the Stoplight task (ALONE condition), and found that behavior was significantly predicted by sensation seeking (s = .438, t = 2.40, p = .024), but not impulsivity (s = −.009, t = .049, ns). These findings provide further validation for the scanner implementation of the task, and suggest that differences in individuals’ reward- or thrill-seeking biases have an especially strong influence on task performance.

fMRI results

Regions exhibiting significant main and interactive effects of age and social context are shown in Table 1. In light of our neurodevelopmental framework, we focused subsequent planned analyses on regions showing either a main effect of age, or an age by social context interaction. Pair-wise contrasts between age groups (adolescents vs. young adults, adolescents vs. adults, young adults vs. adults) indicated that for all of the regions exhibiting a main effect of age, the effect was driven by significantly stronger engagement in adults relative to adolescents (no significant clusters were present for the other pair-wise comparisons). Notably, this pattern of greater regional activation for adult participants was observed in several left LPFC sites (Figure 3a), with young adults demonstrating an intermediate (not significantly different from either adolescents or adults) level of LPFC engagement (Figure 3c, left).

Figure 3

Regions showing a main effect of age and an age by social condition interaction. (a) Regions showing a main effect of age, including the left lateral prefrontal cortex (LPFC, MNI peak coordinates: x = –46, y = 11, z = 26, BA 46), (b) Regions exhibiting an age × social condition interaction, including the right ventral striatum (VS, MNI peak coordinates: x = 9, y = 12, z = −8) and left orbitofrontal cortex (OFC, MNI peak coordinates: x = −22, y = 47, z = −10), and (c) Mean estimated BOLD signal change (beta coefficients) from the four peak voxels of the LPFC (left), VS (middle), and OFC (right) in adolescents (adols.), young adults (YA), and adults under ALONE and PEER conditions. Error bars indicate standard error of the mean. Brain images are shown by radiological convention (left on right), and thresholded at p < .01 for presentation purposes.

Table 1

Regions showing significant (FWE < .05) main and interactive effects of age and social condition in association with Stoplight Task decision-making

BA = Brodmann's Areax,y,z = MNI coordinates

Consistent with the prediction that peer presence especially sensitizes incentive processing in adolescents, significant age by social context interactions were found selectively in the VS and OFC – regions known to be involved in reward prediction and valuation (Figures 3b and 3c). Planned within-group contrasts of PEER versus ALONE condition activity indicated significantly greater PEER condition activation of the VS and OFC among adolescents, but not in the other two groups. Furthermore, direct comparison of age groups within each social condition indicated greater activation in these incentive processing regions among adolescents, relative to adults, in the PEER but not the ALONE condition. These results were further corroborated by an independent voxel-wise test for regions showing a correlation between age and the magnitude of the neural peer effect (i.e. the difference between activity in the PEER and ALONE conditions), which indicated significant inverse correlations [r(38) < −.40, p < .01] between age and context-dependent activation in both the VS and OFC.

We additionally examined trial-based variation in the magnitude of decision-related activity to determine if it could explain the riskiness of the subsequent decision (GO vs. STOP). Specifically, we treated the obtained LPFC, VS, and OFC clusters as regions-of-interest (ROI) and compared activity in these regions during GO versus STOP trials (collapsing across social context). Among adolescent subjects, greater activity in both the VS (Figure 4a) and OFC was associated with risky decision-making, as indicated by significantly increased activity in these regions for GO relative to STOP trials. No decision-dependent differences were found in these regions for older age groups, and activity in the LPFC was statistically equivalent for GO and STOP trials regardless of age (additional detailing of activation in the VS, OFC, and LPFC clusters is provided with the online Supporting Information).Figure 4Activity in the right ventral striatum (VS). Estimated activity was extracted from an average of the four peak voxels in the VS ROI. (a) Estimated VS activity for all GO and STOP trials in adolescents (adols.), young adults (YA), and adults. Significantly different VS activity for GO relative to STOP trials was found for only the adolescents. Error bars indicate standard error of the mean. (b) Scatterplot of activity in the VS indicating an inverse linear correlation between self-reported resistance to peer influence (RPI) and the neural peer effect (β_peer − β_alone).

Additional evidence of the relationship between anticipatory incentive processing and the peer influence on adolescent risk taking was obtained by testing the correlations between activity in each ROI and self-reported sensation seeking, impulsivity, and resistance to peer influence. Whereas individual differences in self-reported sensation seeking and impulsivity were not predicted by the activity patterns observed in any ROI, self-reported resistance to peer influence (RPI) correlated significantly with individual variation in the neural peer effect (Peer vs. Alone) exhibited by the VS [r(28) = −.54, p < .01; Figure 4b]. This relationship remained significant even when age was controlled [r(28) = −.51, p < .01]. In other words, participants’ perception of their susceptibility to peer influence was predicted by the sensitivity of VS output to social context, and this relationship was not diminished when the relationship between age and these two variables was taken into account.

Discussion

As expected, we found that adolescents, but not adults, exhibited increased risk taking when observed by their peers. This behavioral outcome replicates past findings (Gardner & Steinberg, 2005) despite the unique manipulation of social context that was required to accommodate the fMRI environment. Indeed, these behavioral findings have intrinsic value (even without the complementary fMRI results) in further explicating the origins of the peer effect on adolescent decision-making. Since peers were located in a separate room and were prevented from interacting with participants during the decision-making task, adolescents’ heightened inclination to take risks when watched by their friends cannot be explained, at least in this study, by greater explicit encouragement from their peers to engage in risky behavior. In other words, the observed peer effect was not due to overt ‘peer pressure’.

We posited that the risk-promoting effect of peer presence on adolescent decision-making could arise from a neural ‘vulnerability’ that emerges due to the discordant maturation of the brain systems that support decision-making. Accordingly, we sought to determine whether the peer effect might result from alterations to the activity of neural systems underlying cognitive control, incentive processing, or both. As in several prior studies of age-dependent effects on reward processing, we observed differential anticipatory activity in the VS and OFC across age groups. However, our key finding is that this age difference in reward system activity was dependent on social context, consistent with the hypothesis that the presence of peers differentially sensitizes adolescents to the reward value of risky choices. Specifically, relative to adults, adolescents demonstrated significantly greater activation of VS and OFC as they rendered decisions about risk, but only when they were aware that their friends were watching them. Meanwhile, adults showed no differences in the activation of these regions as a function of social context. Indeed, among adult participants, activity in these known reward-sensitive regions was not above baseline in either social context, and in fact skewed negatively in the PEER condition. This pattern of activity may simply indicate that older participants did not perceive task events as potentially rewarding, despite the contingency to a later monetary incentive. However, an alternative explanation is provided within our neurodevelopmental account, which posits that the presence of peers is not rewarding for adults, and that adults are better able to recruit the LPFC to actively suppress reward system outputs as they enact strategic decision-making.

Consistent with this account, and with previously observed differences between adolescents and adults in the recruitment of the cognitive control system during decision-making, we found that adults engaged multiple LPFC sites more robustly than did adolescents when making decisions in the Stoplight task. Importantly, this age difference was evinced regardless of the presence of peers (i.e. it did not interact with social context). These results indicate that while there are age-dependent differences in the recruitment of cognitive control processes, the behavior of this system is insensitive to the manipulation of peer context, and does not principally account for the peer influence on risk taking seen during adolescence but not adulthood. Given that LPFC activation has been associated with calculation of outcome probabilities in support of strategic approaches to risky decision-making (Heekeren, Marrett, Ruff, Bandettini & Ungerleider, 2006; Tobler, Christopoulos, O'Doherty, Dolan & Schultz, 2008; Venkatraman, Payne, Bettman, Luce & Huettel, 2009), we speculate that the stronger activation of LPFC by adults in our study corresponds to a greater reliance on a deliberative strategy to guide decision-making.

Age differences in the context-dependent activation of VS and OFC suggest that adolescents’ heightened susceptibility to peer influence is due more to the impact of peers on incentive processing than on cognitive control. Evidence that VS and OFC activation tracked with the riskiness of subsequent decision-making (GO vs. STOP) in the driving task further suggests that the reward sensitization effect of peer presence translates into an observable increase in risky behavior. Moreover, the significant negative correlation between VS sensitivity to social context and self-reported resistance to peer influence (RPI) indicates that the brain – behavior link observed in our fMRI paradigm reflects an ecologically relevant process that is implicated in risky decision-making in the real world.

Importantly, the social influence on adolescent decision-making was indexed as a phasic interaction between social information processing and reward-related processing, and not as a state-based effect of social context. Accordingly, the significant age by social context interactions are not readily explained as simple age-related differences in the degree of social information processing, which would more likely be reflected as sustained activation differences between PEER and ALONE runs. Rather, these interactions were manifest in the differential magnitude of the transient BOLD response associated with the moment of decision-making, and the magnitude of this transient response forecast subsequent decision-making. It is thus unsurprising that other known components of the social brain (Adolphs, 2003) (e.g. ventromedial PFC, superior temporal gyrus) were not implicated in the interaction. There is evidence that changes in the density and distribution of receptors for dopamine and oxytocin within regions critical to incentive processing take place around the time of puberty (Spear, 2009), and that these changes coincide with a dramatic elevation in the salience of peer interactions (Csikszentmihalyi, Larson & Prescott, 1977). Relative to children and adults, adolescents show heightened activation within incentive processing regions in response to a variety of social stimuli, such as facial expressions and social feedback (Blakemore, 2008). Given the elevated salience and reward value of peer interaction in adolescence (Spear, 2009), awareness of peer observation could act as a tonic stimulus to incentive processing circuits, which are subsequently sensitized to respond to the potential rewards of risky choices.

The present work supports this speculation, and represents an important step toward characterizing the neurobiological mechanisms that are instrumental in adolescent risk taking. However, our task design precludes delineation of the specific reward processes that might be affected by peer context (e.g. reward valuation vs. prediction). Recent work dissociating the functions of ‘decision value’ versus ‘prediction error’ calculations in adolescents’ reward seeking behavior (Cohen, Asarnow, Sabb, Bilder, Bookheimer, Knowlton & Poldrack, 2010) suggests that it may be possible to separately evaluate these reward-specific processes, and we believe this approach could be usefully extended into future investigations of the peer effect. Moreover, it should be noted that our risk-taking task necessarily confounded decision-making and reward-anticipation processes. We accordingly acknowledge that although the primary effect of interest (the age by social context interaction) was observed in brain areas with known involvement in reward processing, we cannot completely rule out alternative explanations. For instance, increased adolescent VS output in the peer condition could reflect an impact of social context on uncertainty during decision-making (rather than on reward processing per se), greater vulnerability to distraction (of any kind) among this age group, or simply differences in motoric output (number of button presses) across age groups and social context. We believe, however, that the involvement of these alternative processes would have been signaled by additional activity differences in brain areas known to be influenced by these factors, and no such patterns were evident.

The present findings expand our understanding of the mechanisms through which teenagers are influenced by their friends to engage in health-compromising behavior, and thus have the potential to inform strategies for intervening to reduce adolescent risk taking. The results provide evidence indicating that awareness of peers selectively amplifies activity in the adolescent brain’s incentive processing system, which in turn influences subsequent decisions about risk. These alterations in brain activity occur even in the absence of direct interactions between the adolescent participants and their peers, and are thus not easily explained as the result of explicit peer pressure to engage in risky activity. The findings extend recent efforts to connect adolescent decision-making to patterns of brain maturation by showing how the neural substrates of adolescent decision-making can be altered by the context in which it occurs. Investigation of the neural pathways through which these contextual factors influence behavior may provide the cornerstone needed to link existing behavioral and neuroscientific findings, and to ultimately understand the decisions that adolescents make when faced with potentially risky choices.

Link to Article

How Peer Pressure Affects Adolescent Brain and Risky Behavior

Introduction

Adolescents are more prone to risky behaviors, such as substance abuse, reckless driving, and criminal activity, compared to children and adults (Steinberg et al., 2008). This heightened risk-taking during adolescence poses significant threats to their well-being. Notably, adolescent risk-taking frequently manifests in social contexts, particularly in the presence of peers (Chassin et al., 2009; Simons-Morton et al., 2005; Zimring, 1998). Experimental studies have demonstrated that the mere presence of peers can directly influence adolescents' decisions towards greater risk-taking (Gardner & Steinberg, 2005; O'Brien et al., in press).

This study investigates the neural underpinnings of this peer influence on adolescent risk-taking. We hypothesize that this phenomenon stems from the differential developmental trajectories of two key brain systems involved in decision-making:

1. Incentive Processing System: This system, encompassing regions like the ventral striatum (VS; including the nucleus accumbens, NAcc) and the orbitofrontal cortex (OFC), evaluates and predicts potential rewards and punishments, influencing decision-making based on these valuations.

2. Cognitive Control System: This system, primarily involving the lateral prefrontal cortex (LPFC), regulates impulsive behaviors and supports goal-directed decision-making by facilitating deliberation among choices.

The incentive processing system undergoes significant remodeling during early adolescence, characterized by changes in dopamine receptor distribution and density, leading to heightened reward sensitivity (Laviola et al., 2001; Spear, 2009). Conversely, the cognitive control system matures gradually from preadolescence through adulthood, marked by gray matter reduction and increased myelination, contributing to enhanced executive functions (Asato et al., 2010; Crone et al., 2006; Giedd, 2008; Luna et al., 2009; Steinberg, 2008).

We propose that the presence of peers, holding high reward value for adolescents (Blakemore, 2008; Spear, 2009), might amplify the activity of the incentive processing system when evaluating potential rewards associated with risky behaviors. This amplified reward sensitivity, coupled with an underdeveloped cognitive control system, could explain the heightened risk-taking observed in adolescents when with peers.

To examine this hypothesis, we employed functional magnetic resonance imaging (fMRI) to measure brain activity in adolescents, young adults, and adults while they engaged in a simulated driving game involving risky decisions. We manipulated the social context by having participants play the game both alone and while being observed by their peers.

Method

Forty participants were recruited for this study:

• 14 adolescents (8 female; mean age = 15.7 years, SD = 1.5)

• 14 young adults (7 female; mean age = 20.6 years, SD = 0.9)

• 12 adults (6 female; mean age = 25.6 years, SD = 1.9)

Informed consent was obtained from all participants following institutional review board approval from Princeton and Temple Universities. Participants received monetary compensation for their participation.

Procedure

Task Design

Participants completed the "Stoplight" task, a simulated driving game where they controlled a vehicle along a straight track with 20 intersections (Figure 1). At each intersection, a traffic signal cycled from green to yellow to red. Participants decided whether to stop at the yellow light (STOP decision) or risk running it (GO decision).

Successful GO decisions saved time, while STOP decisions incurred a short delay. However, an unsuccessful GO decision resulted in a crash and a significantly longer delay. Faster completion of the track yielded higher monetary rewards, encouraging risk-taking. The timing of traffic signals and crash probabilities were unpredictable, preventing participants from anticipating outcomes.

Manipulation of Social Context

Participants completed four rounds of the Stoplight task: two alone (ALONE condition) and two while being observed by two same-age, same-sex friends (PEER condition). The PEER condition was a surprise manipulation introduced after the ALONE condition. During breaks in the PEER condition, friends could communicate with the participant via intercom, informing them of their presence, their observation of the task, and making predictions about the participant's performance. These interactions were designed to be ecologically valid while avoiding explicit encouragement of risky behavior.

Self-Report Questionnaires

Post-fMRI, participants completed self-report questionnaires assessing impulsivity (Barratt Impulsiveness Scale, short form; Patton et al., 1995), sensation seeking (subset of Zuckerman Sensation Seeking Scale; Zuckerman et al., 1978), and resistance to peer influence (RPI scale; Steinberg & Monahan, 2007).

fMRI Data Acquisition and Analysis

fMRI data were acquired using a 3 Tesla Siemens Allegra scanner at Princeton University. Structural and functional scans were acquired using standard protocols.

fMRI data were preprocessed and analyzed using AFNI (Cox, 1996) following standard procedures for motion correction, coregistration, normalization, and smoothing. A general linear model (GLM) was used to identify brain regions exhibiting significant BOLD signal changes at the decision-making point (yellow light onset) in both social contexts (PEER and ALONE). Age group and social context were included as factors in a two-way repeated measures ANOVA. Planned contrasts and correlation analyses were conducted to further investigate significant effects.

Results

Behavioral Results

Replicating previous findings (Gardner & Steinberg, 2005), adolescents took significantly more risks in the PEER condition compared to the ALONE condition, as evidenced by a higher percentage of GO decisions and crashes (Figure 2). This effect of social context was not observed in young adults or adults. Correlation analyses revealed that risk-taking behavior in the ALONE condition was significantly predicted by self-reported sensation seeking, but not impulsivity, consistent with previous findings (Steinberg, 2008).

fMRI Results

Main Effect of Age

Several regions within the left LPFC exhibited a significant main effect of age (Figure 3a). Pairwise comparisons revealed significantly greater LPFC engagement in adults compared to adolescents, with young adults demonstrating an intermediate level of activation (Figure 3c, left).

Age by Social Context Interaction

Significant age by social context interactions were observed in the VS and OFC (Figures 3b and 3c), regions associated with reward processing. Adolescents, but not young adults or adults, showed significantly greater activation in these regions during the PEER condition compared to the ALONE condition. Further, adolescents exhibited greater activation in these regions compared to adults in the PEER condition, but not in the ALONE condition. These findings suggest that peer presence selectively enhances the responsiveness of reward-related regions in adolescents.

Decision-Related Activity

In adolescents, greater activity in the VS and OFC was associated with risky GO decisions compared to safer STOP decisions (Figure 4a). This pattern was not observed in older groups, suggesting a specific link between reward system activity and risk-taking in adolescents.

Correlation with Resistance to Peer Influence

Interestingly, the magnitude of the neural peer effect (PEER - ALONE activity) in the VS was significantly negatively correlated with self-reported resistance to peer influence (RPI) (Figure 4b). This correlation remained significant after controlling for age, suggesting a strong link between individual differences in susceptibility to peer influence and the sensitivity of the VS to social context.

Discussion

This study provides compelling neurobiological evidence for the heightened susceptibility of adolescents to peer influence on risk-taking. Our findings suggest that this susceptibility stems from the selective amplification of activity within the incentive processing system (VS and OFC) in the presence of peers, coupled with an immature cognitive control system (LPFC).

Our data indicate that even passive observation by peers can significantly influence adolescent decision-making, highlighting the potent effect of social context on risk-taking during this developmental period. Critically, this influence appears to operate implicitly, even in the absence of direct peer pressure or encouragement.

The observed heightened activity in reward-related regions during peer observation aligns with the notion that adolescents place a high value on peer interactions and are particularly sensitive to social rewards. This heightened sensitivity, combined with a developing cognitive control system, may create a "perfect storm" for risk-taking in social contexts.

Importantly, the correlation between VS sensitivity to peer presence and self-reported resistance to peer influence strengthens the ecological validity of our findings. It suggests that the neural mechanisms we observed in a controlled experimental setting have real-world implications for adolescent behavior.

While our study elucidates the neural processes underlying peer influence on adolescent risk-taking, further research is warranted to disentangle the specific contributions of reward valuation, prediction error, and other cognitive processes involved. Additionally, investigating the longitudinal development of these neural systems and their sensitivity to peer influence could provide valuable insights into the trajectory of risk-taking behavior across adolescence.

Understanding the neural mechanisms underlying peer influence on adolescent risk-taking has significant implications for developing effective prevention and intervention strategies. Targeting interventions that address the interplay between reward sensitivity, cognitive control, and social context could prove particularly effective in reducing risky behaviors during this vulnerable developmental period.

Link to Article

Risky Decisions: How Peer Presence Affects Teenagers' Brains When Making Risky Decisions

Introduction

It's no secret that teenagers are more likely than kids or adults to take risks: drinking too much, smoking, having multiple sexual partners, getting into fights or trouble with the law, and risky driving are all behaviors that are more common in adolescence (Steinberg, et al., 2008). These preventable behaviors are a serious concern, and many experts believe they are the biggest threat to young people's well-being in developed countries.

What makes teenage risk-taking different from adult risk-taking is the social context. Studies on reckless driving (Simons-Morton, et al., 2005), substance use (Chassin, et al., 2009), and criminal activity (Zimring, 1998) all show that teenagers are far more likely to engage in these behaviors when they are with their friends. While this could be explained by teens simply spending more time with friends, recent experiments (Gardner & Steinberg, 2005; O'Brien, et al., in press) suggest that the mere presence of peers directly influences teens' choices. For example, Gardner and Steinberg (2005) found that teenagers (and, to a lesser extent, college students), took more risks in a simulated driving game when their friends were watching.

Many researchers believe this increased risk-taking in teenagers is due to the interplay of two developing brain systems (Casey, et al., 2008; Luna, et al., 2010; Somerville, et al., 2010; Steinberg, 2008; Van Leijenhorst, et al., 2010a; see also Ernst, et al., 2006):

1. Incentive Processing System: This system, which includes areas like the ventral striatum (VS, including the nucleus accumbens, or NAcc) and the orbitofrontal cortex (OFC), evaluates potential rewards and punishments, influencing decisions based on what seems most appealing.

2. Cognitive Control System: This system, which includes the lateral prefrontal cortex (LPFC), regulates impulses and supports thoughtful decision-making by considering different options.

Neuroimaging studies show that these systems interact during decision-making. Impulsive or risky decisions are often linked to greater activity in the incentive processing regions (Ernst, et al., 2004; Hare, et al., 2009; Kuhnen & Knutson, 2005; Matthews, et al., 2004; McClure, et al., 2004) and less activity in cognitive control regions (e.g. Eshel, et al., 2007; Fecteau, et al., 2007).

These brain systems develop significantly throughout adolescence, but at different rates. The incentive processing system undergoes major changes in early adolescence, particularly in the number and sensitivity of dopamine receptors (Laviola, et al., 2001). These changes are thought to make teens more sensitive to rewards (Spear, 2009). Indeed, neuroimaging studies show that reward-related brain areas like the NAcc, are especially active in adolescents when they anticipate rewards (Ernst, et al., 2005; Ernst, et al., 2009; Galvan, et al., 2006; Geier, et al., 2010; Van Leijenhorst, et al., 2010b). Importantly, Galvan et al. (2007) found that higher NAcc activity during reward anticipation correlated with higher self-reported risk-taking in teens, suggesting a link between reward sensitivity and real-world risk-taking.

The cognitive control system matures more gradually from preadolescence to at least the mid-twenties, involving decreases in gray matter density and increases in myelination (Asato, et al., 2010; Giedd, 2008). This maturation is believed to support the development of executive functions like self-control (Luna), planning (Luciana, et al., 2009), impulse regulation (Steinberg), and flexible thinking (Crone, et al., 2006).

We believe that the difference in development between these two systems makes teenagers more likely to take risks, especially when their friends are around. Since spending time with friends is highly rewarding for teens (Blakemore, 2008; Spear, 2009), having peers present may heighten the activity of the incentive processing system, making teens more focused on the potential rewards of risky behavior. Combined with an underdeveloped ability to control impulses, this reward-focused state could lead to riskier choices.

This study investigates how peer presence affects brain activity during decision-making. We used a simulated driving game to measure brain activity in adolescents, young adults, and adults while they made decisions alone and while being observed by their friends. This allowed us to examine whether peers primarily affect the reward system, the cognitive control system, or both.

Method

Participants

Forty participants were included: 14 adolescents (8 female, ages 14-18), 14 young adults (7 female, ages 19-22), and 12 adults (6 female, ages 24-29). All participants provided informed consent and were compensated for their time.

Procedure

Task Design

Participants played a driving game called the "Stoplight task" (Figure 1), where they controlled a virtual car on a track. They completed four rounds, two in each social condition (alone, with peers). Each round involved navigating 20 intersections while trying to reach the end of the track as quickly as possible to earn money.

At each intersection, a traffic signal changed from green to yellow to red. Participants had to decide whether to risk running the yellow light to save time (GO) or brake and wait (STOP). Running the light successfully saved time, while braking caused a delay. However, crashing after running a red light caused a longer delay than stopping. The timing of the traffic signals and the likelihood of crashing were unpredictable, encouraging risk-taking for greater rewards.

Manipulation of Social Context

Participants were asked to bring two same-age, same-sex friends. In the ALONE condition, they played the game alone. In the PEER condition, they were told their friends were watching their performance on a monitor in another room. This change was a surprise.

During breaks in the PEER condition, friends could talk to the participant through an intercom, confirming they were watching and making predictions about the participant's performance. Friends were instructed to avoid making comments that would directly influence the participant's decisions.

Self-Report Questionnaires

Afterward, participants completed questionnaires assessing impulsivity, sensation seeking, and resistance to peer influence.

fMRI Data Acquisition and Analysis

Brain activity was measured using fMRI while participants played the Stoplight task. Data analysis focused on the decision-making period (when the light turned yellow) to see how social context (ALONE vs. PEER) affected brain responses in different age groups.

Results

Behavioral Results

As expected, only adolescents took significantly more risks when their friends were watching, making more GO decisions and crashing more often (Figure 2). This supports previous findings (Gardner & Steinberg, 2005) and shows that even without direct peer pressure, adolescents are more likely to take risks when peers are observing them.

Figure 2

fMRI Results

Brain imaging revealed significant age-related differences in brain regions associated with both cognitive control and reward processing (Table 1).

Table 1

• Cognitive Control: Adults showed greater activation in the LPFC compared to adolescents, regardless of social context (Figure 3a). This suggests that adults rely more on this area for controlled, strategic decision-making.

• Reward Processing: Activity in the VS and OFC, key areas of the reward system, showed a significant interaction between age and social context (Figure 3b). Adolescents, but not adults, showed increased activity in these areas when their friends were watching (Figure 3c). This suggests that peer presence makes adolescents more sensitive to potential rewards during risky decision-making.

Further analysis revealed that in adolescents, greater VS and OFC activity predicted riskier choices (GO vs. STOP) (Figure 4a), indicating a link between reward system activity and actual risk-taking behavior. Interestingly, how much the VS responded to the presence of peers was correlated with self-reported resistance to peer influence (Figure 4b). This means that teens whose VS activity increased significantly when their friends were watching were also more likely to say they are easily influenced by their peers in real life.

Discussion

This study shows that while adults and teens make similar decisions when alone, adolescents take more risks when their friends are watching, even without direct peer pressure. This difference in behavior is mirrored in their brains. While adults show greater activation in brain regions associated with cognitive control (LPFC), adolescents show greater activation in reward-related areas (VS and OFC) when peers are present.

The findings suggest that teenagers' increased vulnerability to peer influence when making risky decisions stems from heightened sensitivity to the potential rewards of risk-taking when observed by their peers. This sensitivity is reflected in increased activity in brain regions associated with processing and anticipating rewards.

These findings have important implications for understanding and addressing risky behavior in adolescents. By understanding the neural mechanisms underlying peer influence on decision-making, we can develop more effective interventions to promote safe and healthy choices during this vulnerable developmental period. Future research can further investigate how different types of peer influence (e.g., pressure, modeling) and individual differences in reward sensitivity contribute to adolescent risk-taking.

Tables

Note: This table shows the brain regions where significant effects of age and social context were observed. MNI coordinates refer to the location of the peak activation within each region.

Important: This rewrite aims to simplify the original text while preserving its key points and structure. However, it is not a replacement for the original article. For a complete understanding of the research and its implications, please refer to the original publication.

Link to Article

Why Teens Act More Recklessly Around Friends

Introduction

We all know that teenagers tend to do riskier things than kids or grown-ups. This includes things like drinking too much alcohol, smoking cigarettes, having many sexual partners, getting into fights or trouble with the law, and dangerous driving (like driving too fast or while drunk) (Steinberg, Albert, Cauffman, Banich, Graham & Woolard, 2008). Experts believe these preventable actions are the biggest danger to young people in developed countries. Importantly, teenagers take risks differently from adults, both in how often and in what situations. One of the biggest differences is that teens are way more likely to take risks with friends around, which has been seen in studies on reckless driving (Simons-Morton, Lerner & Singer, 2005), drug and alcohol use (Chassin, Hussong & Beltran, 2009), and criminal behavior (Zimring, 1998). You might think teenagers take more risks around friends just because they spend more time together than adults do. However, new experiments (Gardner & Steinberg, 2005; O'Brien, Albert, Chein & Steinberg, in press) show that just having friends nearby can directly affect a teenager's choices. For example, Gardner and Steinberg (2005) had teenagers, college students, and adults play a driving simulator either alone or with two friends watching. They found that teens (and to a lesser degree, college students), but not adults, took way more risks with their friends present. Many researchers (Casey, Getz & Galvan, 2008; Luna, Padmanabhan & O'Hearn, 2010; Somerville, Jones & Casey, 2010; Steinberg, 2008; Van Leijenhorst, Moor, de Macks, Rombouts, Westenberg & Crone, 2010a; see also Ernst, Pine & Hardin, 2006) believe this difference in risk-taking is due to two systems in the brain that work together to make decisions: (i) the reward system which includes the ventral striatum (VS) and the orbitofrontal cortex (OFC), and helps us decide based on the possible good and bad outcomes; and (ii) the control system, which includes the lateral prefrontal cortex (LPFC) and helps us make thoughtful decisions by controlling impulses and weighing our options. Brain imaging studies in both adults and teenagers show these systems working together when we make choices. Impulsive or risky choices often happen when the reward system is more active (Ernst, Nelson, McClure, Monk, Munson, Eshel, Zarahn, Leibenluft, Zametkin, Towbin, Blair, Charney & Pine, 2004; Hare, Camerer & Rangel, 2009; Kuhnen & Knutson, 2005; Matthews, Simmons, Lane & Paulus, 2004; McClure, Laibson, Loewenstein & Cohen, 2004) and the control system is less active (e.g. Eshel, Nelson, Blair, Pine & Ernst, 2007; Fecteau, Knoch, Fregni, Sultani, Boggio & Pasual-Leone, 2007). Both of these brain systems change a lot during the teenage years, but at different speeds. The reward system goes through big changes in early adolescence, particularly with how dopamine receptors are spread out and how many there are (Laviola, Pascucci & Pieretti, 2001). This suggests that teenagers become more sensitive to rewards (Spear, 2009). Brain areas in this pathway (especially the VS), which helps us understand and anticipate rewards (Breiter & Rosen, 1999; Delgado, 2007; O'Doherty, 2004; Schultz, 2010), have been shown in recent studies to be much more active in teenagers when they see something rewarding or think about getting a reward (Ernst, Nelson, Jazbec, McClure, Monk, Leibenluft, Blair & Pine, 2005; Ernst, Romeo & Andersen, 2009; Galvan, Hare, Parra, Penn, Voss, Glover & Casey, 2006; Geier, Terwilliger, Teslovich, Velanova & Luna, 2010; Van Leijenhorst, Zanolie, Van Meel, Westenberg, Rombouts & Crone, 2010b). Interestingly, Galvan and his team (Galvan, Hare, Voss, Glover & Casey, 2007) found that the more active the VS was while a teenager anticipated a reward, the more they said they took risks in real life. This supports the idea that teenagers' heightened sensitivity to rewards leads them to take more risks. The parts of the brain involved in control develop more slowly and over a longer period, from pre-teen years to at least the mid-twenties. This development involves changes like less gray matter density and more myelination (Asato, Terwilliger, Woo & Luna, 2010; Giedd, 2008). These changes are thought to be linked to getting better at things like self-control (Luna), planning ahead (Luciana, Collins, Olson & Schissel, 2009), managing impulses (Steinberg), and adapting to changing rules (Crone, Donohue, Honomichl, Wendelken & Bunge, 2006). We believe that teens' tendency to take more risks around friends might be because these two brain systems are developing at different speeds. Since spending time with friends is so rewarding for teenagers (Blakemore, 2008; Spear, 2009), being around them might make the reward system even more sensitive to potential rewards of risky actions. Because teens are still developing the ability to control these reward signals, this strong desire for rewards might lead them to take more risks. In the brain, the influence of peers on teen decision-making could look like increased activity in areas related to assessing rewards. Another possibility is that having peers around directly affects control processes, which would be seen as different activity in areas related to impulse control. To test these ideas, we looked at brain activity in teenagers, young adults, and adults while they played a simulated driving game. In the game, they had to decide whether to stop at intersections or run a yellow light and risk a crash, with the goal of finishing the course quickly to earn more money. Risky decisions could lead to no delay at the intersection, but also a big crash and delay. We changed the social context by having people play the game alone and then while their friends watched them.

Method

Participants

We studied 40 people (14 teenagers – eight girls, ages 14–18 years, average = 15.7, SD = 1.5; 14 young adults – seven girls, ages 19–22 years, average = 20.6, SD = 0.9; and 12 adults – six girls, ages 24–29 years, average = 25.6, SD = 1.9). Everyone agreed to participate and understood the study, which was approved by Princeton and Temple Universities, and they were paid for their time.

Procedure

Task Design

The Stoplight task (Figure 1) is a simple driving game where you steer a car down a straight road from the driver's seat. You play four rounds, two in each social situation. Each round has a track with 20 intersections (each one is a separate trial), and takes less than 6 minutes to finish (depending on your choices and luck). At each intersection, you decide (by pressing a button) whether to brake for a changing traffic light (green to yellow to red). As you approach the intersection, the light turns yellow, and you choose to risk a crash (GO) or brake and wait for the green light (STOP). Importantly, the timing of the lights and the chance of a crash were random, so you couldn't predict them. We encouraged risk-taking by offering money for finishing quickly. If you didn't brake and successfully crossed an intersection, you saved time, but braking and waiting for the green light took longer. However, crashing took even longer than stopping. We collected data from the scanner and matched it to the fMRI scans using E-Prime (Psychology Software Tools, Pittsburgh, PA), displayed on an LCD screen with headphones and a button box. You can find more details about the task online.

The Stoplight Driving Game.

We asked everyone to bring two friends of the same gender and around the same age (within 2 years). We changed the social situation during the study, making sure the order was balanced. In the ALONE condition, you played without anyone watching. In the PEER condition, we told you your friends were watching from a monitor in the next room. This was a surprise. During the break before the PEER condition, and after each scan in this condition, your friends could talk to you through the intercom. To make it realistic, they could talk naturally, but we asked them to say specific things, like that they were there, could see you play, and made guesses about how you'd do. We told them to avoid saying anything that might make you change your behavior on purpose.

Questionnaires

After the scan, we asked you to answer some questionnaires about how impulsive you are (using a shorter version of the Barratt Impulsiveness Scale; Patton, Stanford & Barratt, 1995), how much you like new and exciting experiences (using part of the Zuckerman Sensation Seeking Scale; Zuckerman, Eysenck & Eysenck, 1978), and how well you resist doing what your friends do (using the RPI scale; Steinberg & Monahan, 2007). Not everyone answered all the questions or finished every questionnaire.

Brain Imaging

We used a special brain scanner called a 3 Tesla Siemens Allegra magnet at Princeton University to take pictures of your brain. We used two types of scans: a structural scan (MPRAGE) to get a detailed 3D picture of your brain, and a functional scan (EPI) to see your brain activity while you played the Stoplight game.

fMRI Data Analysis

We used a special computer program called AFNI (Cox, 1996) to analyze the brain scans. First, we had to clean up the images and align them with a standard brain template. Then, we used a statistical technique called a general linear model (GLM) to find areas of the brain that were more active when making decisions in the Stoplight game, both alone and with friends watching. We also compared the brain activity between different age groups (teens, young adults, adults) to see if there were any differences. To be sure our findings were reliable, we used strict statistical thresholds (p < .005, FWE < .05) and only looked at brain areas that showed consistent activity across multiple participants.

Results

Behavioral Results

We compared how people made decisions in the PEER and ALONE conditions. Similar to Gardner and Steinberg (2005), we saw that teenagers and adults acted similarly when alone, but teenagers took more risks when their friends were watching (Figure 2). Though we didn't have enough people to make it statistically significant [F(2, 38) = 2.66, p = .084), only teenagers took significantly more risks with friends watching, making more GO decisions [t(13) = 2.16, p = .025, one-tailed] and crashing more often [t(13) = 4.06, p < .001, one-tailed]. You can find more details online.

Stoplight Task Performance.

To be sure our driving game measured what it was supposed to, we looked at how your answers on the questionnaires related to how you played the game. How you play could be because of how much you like excitement (making you take risks) or how good you are at controlling yourself (helping you stop). However, a bigger study showed that how people played this game was linked to how much they enjoyed excitement, but not how impulsive they were (Steinberg). We found the same thing. Like the bigger study, we used a statistical test (regression analysis) and found that how people played was linked to how much they liked excitement (s = .438, t = 2.40, p = .024), but not how impulsive they were (s = −.009, t = .049, ns). This shows that our driving game measures what it should, and that how much someone enjoys rewards or excitement affects how they play.

fMRI Results

We found significant differences in several brain regions (Table 1), but we'll focus on the ones that showed an effect of age or an interaction between age and social context, as these are most relevant to our theory. When comparing age groups, we found that for all the regions that showed an age effect, adults had significantly more activity than teens (adolescents vs. young adults, adolescents vs. adults). This was especially noticeable in the LPFC (Figure 3a), with young adults falling somewhere in between (Figure 3c, left).

Figure 3 Regions showing a main effect of age and an age by social condition interaction.

Table 1

Regions showing significant (FWE < .05) main and interactive effects of age and social condition in association with Stoplight Task decision-making

As we predicted, we found a significant interaction between age and social context in the VS and OFC – areas involved in predicting and understanding rewards (Figures 3b and 3c). Looking closer, we saw that teenagers, but not adults or young adults, had much greater activity in the VS and OFC when their friends were watching. Furthermore, teenagers had greater activity in these reward-related areas than adults when their friends were watching, but not when they were alone. Another analysis looking for brain areas where the difference in activity between the PEER and ALONE conditions was correlated with age confirmed these findings, showing significant negative correlations [r(38) < −.40, p < .01] in both the VS and OFC. This means that the younger someone was, the bigger the difference in brain activity between being observed by peers and being alone. We also looked at whether the level of brain activity during a trial could predict whether the person would make a risky or safe decision (GO vs. STOP). We focused on the LPFC, VS, and OFC regions and compared activity during GO and STOP trials (across both social contexts). In teenagers, greater activity in both the VS (Figure 4a) and OFC was linked to risky choices, meaning these regions were more active when they decided to run the yellow light. Older groups didn't show this difference, and the LPFC activity didn't differ between GO and STOP trials for any age group (more details online).

Activity in the Brain

To further understand these findings, we looked at whether individual differences in self-reported traits were related to brain activity. While self-reported sensation seeking and impulsivity did not correlate with activity in any of the ROIs, self-reported resistance to peer influence (RPI) was significantly correlated with the magnitude of the neural peer effect (Peer vs. Alone) in the VS [r(28) = −.54, p < .01; Figure 4b]. This correlation remained significant even after controlling for age [r(28) = −.51, p < .01]. This means that teenagers who said they were more easily influenced by their peers also showed greater differences in VS activity between the PEER and ALONE conditions, and this relationship was not simply due to their age.

Discussion

As expected, we found that teenagers, but not adults, took more risks when their friends were watching. This backs up what Gardner and Steinberg (2005) found, even though we had to change how we set up the experiment for the brain scanner. Our findings are important because they show that even when friends are in another room and can't talk to you directly, teenagers still take more risks, meaning it's not just about being pressured by friends to do something risky. We thought this might be because teenagers' brains are still developing and the parts responsible for making decisions and processing rewards aren't quite in sync yet. So, we looked at whether having friends around changed activity in the brain areas responsible for controlling impulses or processing rewards. We saw that teenagers and adults used their brains differently when thinking about rewards, which other studies have also found. What's important is that this difference was linked to whether their friends were watching. Compared to adults, teenagers' reward systems (VS and OFC) were more active when they made decisions, but only when their friends were watching. Adults, on the other hand, didn't show any difference in these areas whether they were alone or being observed. Actually, for adults, these reward-related areas were less active when friends were watching than when they were alone. This could mean that adults didn't think the tasks were rewarding, even though they could earn money. Or, it could be that adults don't find being watched by their peers rewarding and are better at using their LPFC to ignore the reward system when making decisions. This idea is supported by the fact that adults used their LPFC (the control center) more than teenagers when making decisions in the game, regardless of whether their friends were watching. This might mean that adults rely more on thinking things through when making choices. The fact that teenagers' VS and OFC were more active when they made risky choices suggests that being around friends makes teenagers more sensitive to potential rewards, leading to riskier behavior. Furthermore, the connection between how sensitive the VS was to the social situation and how much teenagers said they were influenced by peers suggests that our findings are relevant to real-life risky decisions. It's important to note that the influence of peers on decision-making wasn't just a constant effect, but rather happened in short bursts, specifically when the teenagers were making decisions about risk. This means it's not just about how teenagers process social information in general, but how this processing interacts with their reward system when making choices. That's why we didn't see differences in other brain areas related to social processing. Around puberty, there are changes in the brain, such as the number and distribution of receptors for dopamine and oxytocin, especially in areas related to processing rewards (Spear, 2009). These changes happen at the same time that teenagers start to care more about what their peers think (Csikszentmihalyi, Larson & Prescott, 1977). Because teenagers find being with friends so rewarding (Spear, 2009), knowing their friends are watching might make their reward systems even more sensitive to potential rewards, especially when making risky decisions. Our study supports this idea and helps us understand how teenagers' brains work when they're making risky choices. It's important to remember that our game measured both decision-making and anticipation of rewards at the same time. So, although we saw the biggest differences in brain areas related to rewards, it's possible that other things were at play, like how teenagers handle uncertainty or distractions, or even just how many times they had to press buttons. However, we didn't see differences in other brain areas that would support those explanations. These findings help us understand why teenagers do risky things, especially around their friends. It seems that teenagers' brains are wired to be more sensitive to rewards in social situations, which can lead them to make impulsive decisions. This information is crucial for developing ways to prevent risky behavior in teenagers, such as teaching them how to make better decisions and resist peer pressure. By understanding how the teenage brain works, we can help them navigate the challenges of adolescence and make healthier choices.

Link to Article

Why Are Teenagers More Likely to Take Risks When Their Friends Are Watching?

Introduction

We all know that teenagers do more risky things than kids or adults. They're more likely to drink too much, smoke, have many dating partners, get into trouble with the law, and have bad car accidents, usually because of reckless driving or driving drunk (Steinberg, Albert, Cauffman, Banich, Graham & Woolard, 2008). Experts believe that these dangerous behaviors are the biggest danger to young people in developed countries, and we need to find ways to stop them.

What's interesting is that teenagers take more risks when they're with their friends than when they're alone. We see this in how they drive (Simons-Morton, Lerner & Singer, 2005), use drugs and alcohol (Chassin, Hussong & Beltran, 2009), and even in criminal behavior (Zimring, 1998). Of course, teenagers spend a lot of time with their friends, so you might think that's why they take more risks when they're together. But studies show that just being around friends actually makes teenagers make riskier choices (Gardner & Steinberg, 2005; O'Brien, Albert, Chein & Steinberg, in press)!

One study (Gardner & Steinberg, 2005) had teenagers, college students, and adults play a driving game, either alone or with friends watching. They found that teenagers (and, to a lesser degree, college students) took way more risks when their friends were watching!

Scientists believe that two parts of the brain work together to affect decision-making, and these parts develop differently in teenagers, leading to more risks:

1. The Reward System: This part of the brain (including areas like the ventral striatum and orbitofrontal cortex) is like a thrill-seeker, always looking for exciting rewards and trying to avoid punishment.

2. The Control System: This part of the brain (especially the lateral prefrontal cortex) acts like a brake, keeping impulses in check and helping us think carefully about our choices.

Brain scans show that these systems talk to each other, and when people make risky choices, the reward system is usually more active while the control system is less active (Ernst, Nelson, McClure, Monk, Munson, Eshel, Zarahn, Leibenluft, Zametkin, Towbin, Blair, Charney & Pine, 2004; Hare, Camerer & Rangel, 2009; Kuhnen & Knutson, 2005; Matthews, Simmons, Lane & Paulus, 2004; McClure, Laibson, Loewenstein & Cohen, 2004; Eshel, Nelson, Blair, Pine & Ernst, 2007; Fecteau, Knoch, Fregni, Sultani, Boggio & Pasual-Leone, 2007).

Here's the catch: these brain systems grow and change a lot during teenage years, but at different speeds.

The reward system is like a teenager itself, going through a lot of changes early on, especially in how sensitive it is to dopamine, a chemical that makes us feel good (Laviola, Pascucci & Pieretti, 2001). These changes make teenagers much more sensitive to rewards (Spear, 2009).

Meanwhile, the control system is more like a slow and steady adult. It takes its time to mature, becoming better at things like controlling impulses, planning, and adapting to rules (Asato, Terwilliger, Woo & Luna, 2010; Giedd, 2008; Luna; Luciana, Collins, Olson & Schissel, 2009; Steinberg; Crone, Donohue, Honomichl, Wendelken & Bunge, 2006).

So, imagine this: teenagers' brains are wired to find rewards extra exciting, especially when they're with friends. But the part of their brain that helps them stop and think isn't quite ready to handle all that excitement. This mismatch might explain why they take more risks when their friends are around.

To test this, scientists studied the brain activity of teenagers, young adults, and adults while they played a driving game. The game let them decide whether to stop at a yellow light or speed through and risk a crash. The goal was to finish the game as fast as possible to earn money. Sometimes, taking a risk meant getting ahead, but if they crashed, they lost even more time.

Here's the twist: Participants played the game alone and while their friends watched.

Method

Participants

The study included 40 people: 14 teenagers (ages 14–18), 14 young adults (ages 19–22), and 12 adults (ages 24–29). Everyone volunteered and got paid for their time.

Procedure

Task Design

Imagine a simple driving game on a straight track with traffic lights. That's the "Stoplight task" (Figure 1). Participants controlled a car and had to decide whether to stop at a yellow light or risk a crash by speeding through. They completed four rounds of the game, two alone and two with friends watching. The tricky part? The timing of the lights and the chance of a crash were unpredictable. Players were encouraged to take risks because finishing faster meant earning more money. Stopping at a yellow light cost a little time, but crashing cost even more! Figure 1 The Stoplight driving game. In the Stoplight driving game, players aimed to reach the end of a straight track as quickly as possible. The track had 20 intersections with traffic lights, and each intersection was a separate challenge. As the car approached the intersection, the light turned yellow, and the player had to choose between stopping (STOP) or taking a risk by going through (GO). Stopping caused a short delay, but successfully running the light meant no delay. However, crashing resulted in a much longer delay. Subjects played four rounds of the game (two alone and two with a peer audience).

Manipulation of Social Context

Participants were asked to bring two friends of the same age and gender. During the "ALONE" condition, they played the game without anyone watching. But during the "PEER" condition, they were told their friends were watching their every move on a screen in another room! This switch was a surprise.

To make it feel real, the friends could talk to the participant through an intercom during breaks, but they couldn't say anything to intentionally influence their choices.

Self-Report Questionnaires

After the game, participants answered questionnaires about how impulsive they are, how much they enjoy thrills, and how easily they are influenced by their friends.

fMRI Data Acquisition

Participants' brain activity was scanned while they played the game using a special machine called an fMRI.

fMRI Data Analysis

Scientists used computer programs to analyze the brain scans and see which parts of the brain were more active during the game, especially during the decision-making moments.

Results

Behavioral Results

The study found that teenagers, but not adults, took more risks when their friends were watching (Figure 2). This means that teenagers were more influenced by their friends' presence, even though their friends weren't telling them what to do.

fMRI Results

Brain scans showed more activity in the reward system (ventral striatum and orbitofrontal cortex) in teenagers when they made decisions with friends watching compared to when they were alone. This suggests that seeing their friends made teenagers more sensitive to the excitement of potentially winning.

Meanwhile, adults showed similar activity in the control system (lateral prefrontal cortex) whether they were alone or with friends watching. This suggests that adults were better at using their control system to regulate their impulses, even when their friends were watching.

Discussion

The study showed that teenagers take more risks when their friends are around because their brains are wired differently than adults'. Seeing their friends seemed to make the reward system more active, making risky choices more appealing, while their control system struggled to keep up.

This research helps us understand why teenagers sometimes make risky decisions. It's not just because they're being pressured by their friends; it's because their brains are still developing, and they're more sensitive to the excitement of rewards when their friends are around.

By understanding these brain differences, we can develop better ways to help teenagers make safer choices. For example, we can teach them how to recognize risky situations and how to resist peer pressure. We can also create environments that are less likely to encourage risky behavior.

Link to Article