Introduction

Adolescents are notorious for engaging in risky behavior (Dahl, 2004; Steinberg, 2008). Driven by high levels of sensation seeking (Steinberg et al., 2008), adolescents are more likely than adults or children to seek high-intensity rewarding experiences that have potential consequences for their health and safety. These include engaging in normative thrill-seeking behaviors such as dangerous driving, sexual intercourse without condom use and drug use, as well as foregoing more preventative behaviors that could promote health and safety, such as the use of seat belts or bicycle helmets (CDC, 2010).

Peer social context is a key factor in adolescents’ risk-taking: Risky behaviors such as reckless driving, substance use, and criminal activity are most likely to occur while adolescents are in the presence of peers (Albert et al., 2013). Adolescence is also a developmental period of substantial changes in social context and social behavior, with the emergence of romantic and sexual relationships, the development of intimate friendships and the enhanced salience of status among peers (Choukas-Bradley et al., 2015). Conceptual models of adolescent development emphasize that a central force influencing changes in behavior, affect and physiology is social reorientation, whereby the behavioral and neurobiological aspects of social-cognitive and affective processes change to prioritize peer relationships, such that friendships, sexual relationships and romantic relationships become increasingly salient (Blakemore and Robbins, 2012; Somerville, 2013). Not surprisingly, peer influence, especially for daring behaviors, becomes a more prominent motivator than parental influence or personal decision-making at this age (Liao et al., 2013).

Social reward, especially experiences marked by high positive affect and a peer context, is critical to adolescents’ risky behavior. Indeed, evidence from behavioral and neuroscience research supports adolescents’ intense sensitivity to rewarding and peer contexts. Compared with adults, adolescents take more risks during simulated driving in the presence of peers (Chein et al., 2011), are more easily distracted by rewards during cognitive control tasks (Somerville et al., 2011), and display greater response to pleasant stimuli in reward-critical regions such as ventral striatum (Galván et al., 2006). At an individual differences level, adolescents who are prone to deriving a strong sense of reward from peer relationships may respond to the unique social development experiences of adolescence with more frequent or intense engagement in risky behaviors. With increased value placed on enhancing social status, impressing peers and seeking thrills, adolescents who are more sensitive to their peers’ influence or respond to their peers’ positive affect with more enjoyment could be most liable to risky behavior.

Behaviorally, adolescents’ conversations with friends can be a context for promoting rule-breaking behavior (Dishion et al., 1996). This potentially occurs via the experience of social reward. In particular, variability in reciprocal positive affect, or the behavioral tendency to respond to another person’s positive affect with expressions of positive affect, could reveal the sensitivity to social reward that makes some adolescents engage in higher rates of risky behaviors. Because peer relationships have important value for social functioning during adolescence, close friends could provide a behavioral setting for eliciting reciprocal positive affect. Interactions with close friends could also have the potential to elicit variability in neural responses to social reward.

Neurally, individual differences in reward-circuit function are associated with adolescents’ risky behavior (Galván et al., 2007) and susceptibility to peer influence (Pfeifer et al., 2011). Thus, because of the unique developmental link between peer social reward and risky behavior that emerges in adolescence, neural sensitivity to peer reward could serve as a trait-like vulnerability factor for risky behavior during this phase of life. Indeed, recent findings have indicated that left VLPFC response to reward corresponds to traits related to risky behavior (Chase et al., 2017).

Risky behavior could be more likely for adolescents who have sensitive neural systems for processing not just reward, but social reward in particular. Adolescents’ response to social reward, increased social focus, and rates of risky behavior are putatively driven by development in a combination of reward, social and self-regulatory networks (e.g. Casey et al., 2011). The reward network includes the ventral striatum, the primary target of ventral tegmental dopamine neurons that is considered the hub of reward circuitry; the amygdala, which responds to reward receipt; and the medial prefrontal cortex (PFC), which contributes to affective experience and regulation in response to reward (Haber, 2016). The social and self-processing network includes the temporoparietal junction, which is implicated in theory of mind and responds to social reward (e.g. Eckstrand et al., 2017); the medial PFC, which processes both social and self-relevant information; and the posterior cingulate and precuneus, a combined hub of the default-mode network with a role in self-referential, autobiographical and agentic processing (Nelson et al., 2005; Northoff and Hayes, 2011; Blakemore and Mills, 2014). Other regions contribute to multiple networks involved in processing social reward. The ventromedial PFC is postulated to compute reward valuation, affect regulation and social cognition (Delgado et al., 2016); the anterior insula contributes to reward seeking and reward responding but also appears to compensate for social pain (Cristofori et al., 2015) and contribute to adolescents’ risky behavior (Smith et al., 2014); and the ventrolateral prefrontal cortex (VLPFC) is central to several aspects of affect and self-regulation (Braunstein et al., 2017), including impulsive sensation-seeking (Chase et al., 2017). This intersecting set of networks undergoes development during adolescence and coordinates the increasingly sophisticated social-affective processing and corresponding behaviors, including risky behavior, that emerge in adolescence.

This study was guided by the stance that adolescents’ neural response to peer social reward (here, positive affect with a close friend) may differ greatly across adolescents and that such individual differences (i.e. variability across people in magnitude of neural response) may predict individual differences in risky behavior. From developmental psychopathology and clinical neuroscience perspectives, these individual differences could tip the balance toward risky behavior in contexts of peer positive affect. Specifically, we hypothesized that adolescents with a combination of heightened neural response and reciprocal positive affect response to social reward will engage in a higher level of a range of risky behaviors.

To test this hypothesis, we developed a novel fMRI social reward paradigm using dynamic, personally relevant peer stimuli. This paradigm uses stimuli high in positive affect and is individualized based on video from a conversation with a close friend about a shared, high-intensity, pleasant experience. We have used similar approaches successfully in parent–child contexts to assess social-affective responding in an ecologically valid way (Whittle et al., 2009; Morgan et al., 2015). We propose that behavioral response from the conversation can capture reciprocal positive affect, while neural response can capture sensitivity to peer social reward. In all, combining rigorous behavioral observation and functional neuroimaging and using methods based on naturalistic contexts for risky behavior could lead to progress in understanding the social and affective neuroscience of adolescence.

Materials and methods

Participants and protocol

Participants were 50 typically developing community adolescents with no history of psychiatric disorder or serious health problems. Participants were ages 14–18 (M = 16.22, s.d. = 1.4), 48.6% female, and 68% European American, 27% African American and 5% mixed race. Of the original 70 participants in the study, 7 did not complete the fMRI scan, due to ineligibility (recent concussion, n = 3; claustrophobia, n = 2; mental health history, n = 2), 3 were unable to be contacted after the initial visit and 2 refused the scan. Of the 58 who completed the entire assessment, 3 did not complete the fMRI task because of technical problems, 1 had inadequate coverage of the ventral striatum (≤80%), 1 had excessive movement (>25% of volumes over 2 mm in any direction) and 3 had missing behavior data because of coding error. All 50 participants had adequate coverage in ventral brain areas and <2 mm movement in any direction. All participants identified a same-gender best friend who attended the lab visit to complete the peer interaction task (M = 15.82 years, s.d. = 1.2; race: 38% African American, 62% European American; demographic data on friends were missing for 18 participants).

Participants completed a lab visit with self-report measures and an MRI scan. Most participants (57%) were scanned within 1 month of their lab visit (median = 27.5 days, s.d. = 54.8). The University of Pittsburgh Institutional Review Board approved all research procedures, and written informed consent was obtained from each participant and a parent/guardian.

Risky behavior

Participants completed the standard high school version of the Youth Risk Behavior Survey [YRBS; Center for Disease Control and Prevention (CDC), 2010], an 89-item self-report instrument developed for epidemiologic research on high school students’ engagement in 6 domains of health-risk behaviors. These are (i) behaviors contributing to injury and violence, (ii) sexual behaviors related to negative outcomes, (iii) alcohol and drug use, (iv) tobacco use, (v) unhealthy dietary behaviors, and (vi) inadequate physical activity. These behaviors are considered risky based on their potential to compromise physical health (e.g. obtaining sexually transmitted infections from intercourse without condom use) or mental health (e.g. developing addiction from frequent use of illicit drugs). Thus, the YRBS does not include reward-seeking behaviors that are adaptive or that have only minor direct consequences for health and wellbeing (e.g. initiating a new romantic relationship, raising one’s hand in class).

Scores for risky behavior were computed based on a single factor created by Youssef et al. (2016) from 10 YRBS items selected to represent a broad range of risky behaviors, including substance use, seat belt use, sexual behavior and suicidality. Items in the factor include daily cigarette smoking, seat belt use while riding in a car, number of lifetime sexual partners, and having gotten into a physical fight (see Supplementary Table S1 for all items). The factor was tested with confirmatory factor analysis in a sample of 174 community adolescents and then replicated in a sample of 4135 16-year-old adolescents from the 2009 CDC Youth Risk Behavior Surveillance Survey, with good fit and a unidimensional factor structure in both samples (see Youssef et al., 2016 for factor loadings in both samples). All 10 items in the factor were included in this study, as in previous work (Eckstrand et al., 2017). Raw scores were used in analyses, as they were highly correlated with scores adjusted for factor loadings. Risky behavior scores were normally distributed, with higher scores indicating higher levels of general risky (M = 15.51, s.d. =5.57, range = 8-27).

Reciprocal positive affect behavior

Each adolescent (target) and her or his best friend were video recorded as they engaged in a 10-min conversation, with 5 min devoted to ‘the most fun you’ve ever had together’ and 5 min devoted to ‘a fun or exciting event you’d like to plan together.’ Topics were chosen based on participants’ responses to a list of pleasant events, with examples including high school graduation, visiting a local amusement park, soccer camp and weekend parties with alcohol and drug use. Similar tasks have been used successfully in families of healthy (Whittle et al., 2009) and depressed (Sheeber et al., 2007) adolescents.

Affective behavior from the conversation was coded from video for two purposes, in two ways, by separate coding teams: (i) using the LIFE coding system (Hops et al., 1995) to obtain detailed measurement of reciprocal positive affect, which was the main regressor in our models; and (ii) using ratings of participant and friend positive and neutral affect to select segments to create stimuli in the fMRI task (see below). For LIFE coding to obtain reciprocal positive affect, trained observers blind to hypotheses coded adolescents’ affect expression and verbal content from video in real time. Four constructs—aggressive, positive, dysphoric and other—were derived from the individual codes. These constructs are based on the theoretical rationale of the LIFE system, which was developed to examine the function of emotions in social contexts. The positive affect construct, which was the focus of our analyses, captures happy, caring and facilitative behavior codes.

Reciprocal positive affect—to examine the tendency to respond to a close friend’s positive affect in kind, rather than a general tendency toward positive affective behavior—was then computed from the positive affect construct of LIFE codes. This variable was computed using the Generalized Sequential Querier Program (Bakeman and Quera, 2011). Conditional lagged probabilities for positive affect were computed for participants’ behavior given friends’ behavior, by considering participants’ behaviors occurring from 1 s after the onset until 1 s after the offset of each observation of friends’ positive affect (i.e. instances in which the target participant expressed positive affect during or after the best friend’s expression of positive affect). This resulted in a 2×2 contingency table for each dyad, reflecting the probability of the consequent (participant positive affect) given the antecedent (friend positive affect), as compared to the probability of the consequent given all other peer behavior antecedents and the probability of all other participant affective behaviors given friend positive affect. From these tabled values of joint probability, adjusted residuals were computed for analyses. These values reflected associations greater than expected by chance (i.e. positive values) or lower than expected by chance (i.e. negative values) (M = 13.70, s.d. = 8.06, range = −2.95 to 33) and were distributed approximately normally, with a mean of 0 and a variance equal to 1. We note that reciprocal positive affect is intended to capture a behavioral tendency during the entire interaction and is not meaningful across smaller time segments.

Neural response to friend positive affect

Best Friend fMRI task

This novel task, personalized for each participant, contained six video clips of the participant’s best friend and six video clips of an unfamiliar, same-gender, control adolescent presented in a block design with 10-s fixation displays between blocks. Blocks were presented in a predetermined, pseudorandom order so that positive and neutral affect blocks alternated and clips with the friend or unfamiliar peer alternated (Figure 1). During the task, participants were instructed to attend to each video and to press a button at the onset of each, to ensure that they were awake and engaged.

Affective behavior from the conversation was coded from video to create task stimuli. Lab conversation videos were coded in 5-s epochs by a team of trained observers using an adaptation of the AFFEX coding system (Izard et al., 1983). This was conducted separately from the coding to compute reciprocal positive affect behavior. Specifically, videos were coded for friends’ positive and neutral affect during the lab conversation (i.e. not target participants’ affect or dyadic affect), and coding data were used to determine each participant’s stimulus segments for the friend positiveand friend neutral conditions of the fMRI task. Positive affect was coded with a score of 0–2 for its presence and intensity and neutral affect was coded with a score of 0 or 1 for its presence only.

Approximately 25% of the videotapes were additionally coded by an extensively trained master coder for reliability (mean ICC for PA = 0.93, range = 0.86–0.96; mean ICC for neutral affect = 0.92; range = 0.89–0.94). The master coder selected 20-s segments based on predominance of positive or neutral affect. As intended, the best-friend positive affect stimuli included higher levels of positive affect than neutral stimuli [(M = 1.55; s.d. = 0.37) and (M = 0.08; s.d.=0.15); t(232) = −40.24, P = 0.000], and best-friend neutral stimuli included higher levels of neutral affect than positive stimuli [(M = 0.93; s.d. = 0.16) and (M = 0.06; s.d. = 0.16); t(228) = 41.48, P = 0.000].

Stimuli included the head and shoulders of the best friend (i.e. not a view of the participant herself) and audio of both adolescents. We made efforts to ensure that video clips were equivalent in lighting, camera angle, zoom and intensity of affect. Stimuli included segments drawn from both past- and future-focused parts of the lab conversation, with positive affect clips tending to be from the past conversation (66%) and neutral clips tending to be from the future conversation (60%). Similar to methods used in Whittle et al. (2012) and Morgan et al. (2015), and to avoid participants’ inadvertent familiarity with adolescents in the control conditions, stimuli for the unfamiliar-peer positive affect and unfamiliar-peer neutral affect control conditions were drawn from video segments of dyadic interactions of adolescent actors from Eugene, Oregon. These adolescents’ training allowed them to convincingly portray a conversation with a close friend. Control stimuli were selected using the same procedures used for the best-friend stimuli. As with the best-friend stimuli, the segments selected for the unfamiliar peer positive affect stimuli had higher levels of positive than neutral affect [(M = 1.71; s.d. = 0.29) and (M = 0.04; s.d. = 0.10), respectively; t(10) = −13.19, P = 0.000], and the segments selected for the unfamiliar peer neutral affect stimuli had higher levels of neutral than positive affect [(M = 0.75; s.d. = 0.50) and (M = 0.06; s.d. = 0.16), respectively; t(6) = 3.00, P = 0.024]. Control stimuli were presented to each participant to match the participant’s gender and approximate age.

Because our focus was positive affect within a familiar peer context (e.g. rather than positive affect from a familiar vs unfamiliar peer), the contrast generated for analyses was friend positive affect > friend neutral affect. This contrast allowed assessment of neural response to a type of social reward that is relevant to adolescents’ risky behavior. The contrast was defined across all three 20-s blocks of friend positive affect stimuli and all three 20-s blocks of friend neutral affect stimuli. The onset and duration of each condition was based on all 20-s segments of best friends’ behavior selected for inclusion as stimuli based on AFFEX coding. Thus, the two conditions had an identical duration of data analyzed.

fMRI acquisition and preprocessing

Each participant was scanned using a Siemens 3-T TIM Trio scanner. Structural images were acquired using MPRAGE 160 axial slices, 1.2-mm thick (TR/TE = 2300/2.98 ms, FOV = 256×240 cm², matrix = 256 × 240, flip angle = 9°). BOLD functional images for the friend task were acquired in a single run, with a gradient echo planar imaging sequence and covered 39 axial slices, 3.1-mm thick, beginning at the cerebral vertex and encompassing the entire cerebrum (TR/TE = 2000/28 ms, FOV= 20×20 cm², matrix = 64 × 64, flip angle = 90°).

Preprocessing and analysis of fMRI data were completed using SPM8 (http://www.fil.ion.ucl.ac.uk/spm). Structural images for each participant were segmented to focus on gray matter. For each functional scan, data were realigned to correct for head motion. Volumes with excess motion (>3 s.d. from the subject’s mean, >0.5 mm scan-to-scan translation, or >0.01 degrees of scan-to-scan rotation) were identified using Artifact Detection Toolbox (ART; http://www.nitrc.org/projects/artifact_detect/) software. Preprocessed data were inspected prior to first-level analysis to ensure that all participants had fewer than 25% of volumes with excessive movement detected by ART. Images were then spatially normalized into standard stereotactic space (Montreal Neurological Institute template) using a 12-parameter affine model and smoothed with a 6 mm full-width at half-maximum Gaussian filter. Voxel-wise signal intensities were ratio normalized to the whole-brain global mean. Voxels were resampled during preprocessing to 2 mm³.

Data analyses

First-level tests for effect of task within each participant were calculated at each voxel using paired t-tests for friend positive affect > friend neutral affect. In addition, exploratory analyses to address the specificity of effects examined the contrasts unfamiliar-peer positive affect > unfamiliar-peer neutral affect and (friend positive affect > friend neutral affect) > (unfamiliar-peer positive affect > unfamiliar-peer neutral affect) (see below).

Second-level analysis was conducted using a within-sample t-test masked for regions that respond to fMRI paradigms focusing on social stimuli. The mask was obtained from the Neurosynth platform (neurosynth.org; Yarkoni et al., 2011), which provides meta-analytic findings across fMRI studies measuring specific constructs. Given our focus on social reward, we selected the search term social, which, at the time of analysis, yielded a map of results from exactly 1000 studies of social stimuli. The map included the following regions: dorsomedial, ventromedial and ventrolateral PFC; precuneus; temporoparietal junction (including right superior temporal gyrus); temporal pole; amygdala; and ventral striatum (Supplementary Figure S1). Type I error for the within-sample t test was controlled by applying a voxel-wise height threshold of P < 0.0001 and family-wise error correction at cluster level of P < 0.05, which is consistent with current recommendations for rigorous adjustment for multiple comparisons in fMRI research.

Mean BOLD response for a sphere of 2 mm around the peak voxel of each cluster resulting from the second-level analysis above was then extracted for moderation analyses. One moderation model was computed for each cluster. Additional analyses were performed extracting (i) the BOLD response within each of the significant clusters and (ii) the BOLD response within a priori anatomical masks of social reward regions significantly activated by the task using a small volume correction. Age and gender were included as covariates in t-tests to adjust for their potential role, even though participants’ risky behavior, affective behavior, and neural response to reward did not vary with gender, race or age (all Ps > 0.07).

Moderation analyses tested the hypothesis that behavioral×neural response to peer social reward predicts risky behavior. Analyses were conducted using the PROCESS macro for SPSS (Hayes, 2012). In these analyses, which do not require significant association between independent variable and dependent variable to test moderation, reciprocal positive affect was entered as the independent variable, extracted BOLD variable for each region was entered as the moderating variable, and risky behavior was entered as the dependent variable. Type I error in moderation models was controlled using the sliding linear model (SLIM; Wang et al., 2011), a method designed for data sets with dependence structure, as is the case for fMRI variables extracted from second-level model described above. Age and gender were not covaried in moderation analyses given that they were previously controlled for in the neuroimaging analyses. However, supplementary analyses revealed that the addition of age and gender as covariates did not affect the significance of the results.

Results

Neural response to the Best Friend task

Participants exhibited neural response to best friend positive affect relative to best friend neutral affect in nine regions that have been reported, across studies, to respond to social stimuli: VLPFC (bilateral), dorsomedial PFC, superior temporal gyrus (bilateral), middle temporal gyrus (bilateral), anterior insula, and fusiform gyrus (Table 1 and Figure 2). Whole-brain analyses confirmed the response of these regions (see Supplementary Table S2 and Supplementary Figure S2). Thus, the task effectively engaged the circuitry of interest.

Regression analyses in SPM indicated that neural response to best friends’ positive affect was unrelated to reciprocal positive affect behavior or risky behavior. Also, bivariate correlations revealed that reciprocal positive affect and risky behavior were unrelated (r = 0.07, P = 0.63).

Interaction of neural response and behavioral response to friend positive affect as a predictor of risky behavior

The moderation model for the left VLPFC cluster was significant [R² = 0.23, F(3, 45) = 4.43, P = 0.008, SLIM P = 0.01]. Based on outlier analyses, one case was excluded from this analysis based on having a studentized residual >2. Furthermore, moderation analyses indicated that the combination of left VLPFC response and contingent positive affect behavior predicted risky behavior [ΔR² = 0.19, F(1, 45) = 11.21, P = 0.002, SLIM P = 0.03]. That is, adolescents’ left VLPFC response to their best friends’ positive affect moderated the association between their shared positive affect during an interaction with that friend and their behavior across multiple health-risk domains.

Moderation models with the other clusters activated by the task were nonsignificant (F = 0.03 for right temporoparietal junction, 1.78 for right middle temporal gyrus, 1.41 for left anterior insula, 1.09 for right VLPFC, 2.46 for right superior temporal gyrus, 0.86 for dorsomedial PFC, 0.01 for left middle temporal gyrus, 0.01 for right fusiform gyrus; Ps = 0.12–0.94).

To further examine moderation findings, we applied the Johnson–Neyman technique, which explicates an interaction effect involving a continuous moderator variable by indicating regions of significance, or the values of a continuous moderator variable above or below which there is a conditional association between the independent and dependent variables (Preacher et al., 2006). This technique indicated that left VLPFC response moderated the association between reciprocal positive affect behavior and risky behavior, with significant interaction effects evident at both low and high levels of neural response [conditional effects at VLPFC response <−0.07 (12% of cases; n = 8) or >0.80 (59%; n = 20); Figure 3]. That is, adolescents who reported more behaviors such as using illicit substances, getting in fights or having multiple sexual partners were those with either higher left VLPFC response and higher reciprocal positive affect (the predicted combination of effects) or lower left VLPFC response and lower reciprocal positive affect (an unexpected combination of effects). Other regions that exhibited response to a best friend’s positive affect—such as dorsomedial PFC and temporoparietal junction—did not significantly moderate the association between reciprocal positive affect and risky behavior.

Exploratory analyses

We conducted two sets of additional analyses to examine the results further. First, to test the specificity of findings to the best-friend context, we used a within-sample ttest of response to unfamiliar-peer positive affect > unfamiliar-peer neutral affect and a paired t test of response to (friend positive affect > friend neutral affect) vs (unfamiliar-peer positive affect > unfamiliar-peer neutral affect). No significant clusters emerged for response to unfamiliar peer positive affect or for unfamiliar peer > best friend positive affect. However, adolescents exhibited greater response to best friends’ positive affect than to unfamiliar peers’ positive affect in a cluster including the VLPFC, superior temporal gyrus, and inferior frontal gyrus [119 voxels, t = 5.78, P < 0.001, cluster pFWE = 0.01, (−54, 10, 2)]. Second, to test the potential contribution of ventral striatum, which did not emerge at our specified statistical threshold but which increases in adolescence and is related to reward sensitivity (e.g. Braams et al., 2015), we probed results using a lower statistical threshold of P < 0.005. No significant clusters emerged in the ventral striatum, and the only striatal area showing response at this more liberal threshold was the caudate tail. We also used an anatomical ventral striatum mask for the direct comparison of best friend and unfamiliar peer positive affect described above, and no significant clusters emerged.

Discussion

Adolescents’ combined neural response and behavioral response to their best friends’ PA—but, tellingly, neither type of response alone—was associated with their engagement in a range of real-world risky behaviors. Surprisingly, greater engagement in risky behaviors was associated with the combination of higher neural and higher behavioral response and the combination of lower neural and lower behavioral response to best friends’ positive affect. The neural response moderating the association between positive affect behavior and risky behavior was evident in the left VLPFC, a region associated with trait-like tendencies toward impulsive sensation-seeking.

Both higher and lower reciprocal positive affect were related to higher levels of risky behavior. Moreover, this was specifically the case for those with higher and lower VLPFC response, respectively. This seeming u-shaped association between neural–behavioral response to friend positive affect and risky behavior suggests two possible pathways to adolescents’ risky behavior: high susceptibility to social reward and intense positive affect, or relative indifference to social reward. Perhaps strong reactivity to social reward promotes thrill-seeking attempts to enhance or maintain positive affect, whereas weak reactivity promotes engagement in risky activities for other reasons. Adolescents in the former pathway could find themselves with poor health outcomes, but they might also derive some benefits from their tendencies: in the context of typical development, sensitivity to social rewards could predispose adolescents to obtain social status or to engage in adaptive, pro-social behaviors (see Telzer, 2016). Adolescents in the latter pathway might be those with low baseline levels of reward responding or high susceptibility to boredom, for whom risky behaviors serve to compensate for a tendency toward blunted responding (Zuckerman, 1996).

The pathway involving less sensitivity to social reward occurred less frequently in our sample than the pathway involving greater sensitivity. This pathway could be more evident in clinical populations, such as adolescents with serious conduct problems and callous–unemotional traits (Blair et al., 2014); depression, a disorder accompanied by low response to reward in the striatum (e.g. Forbes et al., 2009); or suicidality, a class of risky behavior included in our outcome variable and associated with other risky behaviors (Stewart et al., 2017). Alternatively, we might not have sampled that extreme range of responding. In all, there could be a sweet spot for sensitivity to social reward, whereby extremely low or high intensity of sensitivity could lead to problem-level risky behaviors while moderate intensity leads to levels appropriate for promoting affiliation with peers, adaptive status-seeking and individuation from parents.

Neural and behavioral responses were assessed during an ecologically valid context: a conversation about an intensely positive, shared experience. This allowed us to extend the investigation of adolescents’ risky behavior beyond traditional models of susceptibility to peer influence and measures of isolated, individual responses elicited by standardized, static stimuli. In contrast to most studies of adolescent social processing, which have focused on the mere presence of peers, evaluations by virtual peers, or cognition about peers (e.g. Jarcho et al., 2015; Bolling et al., 2016; Will et al., 2016; see Pfeifer and Blakemore, 2012 for a review), we focused on what is potentially the most powerful, salient context for adolescents’ risky behavior: a dynamic interaction with a close friend involving heightened positive affect. This was also our first investigation employing the innovative Best Friend fMRI task. Building on other recent work using stimuli from family relationships (Whittle et al., 2012; Morgan et al., 2015), the task engaged regions commonly related to social processing, including dorsomedial PFC, ventrolateral PFC, anterior insula and temporoparietal junction. Notably, neural response to positive affect was stronger in the context of close friendship than in the context of a generic peer: a set of regions involved in social and affective processing exhibited more response to the best friend, whereas no clusters exhibited greater response to the unfamiliar peer.

In addition, while extant studies have rarely focused on individual differences, our task examined variability in neural response as a statistical predictor of risky behavior. The stimuli convey the social context, autobiographical history and meaningful pleasant experiences that accompany real-world peer influence on adolescents’ risky behavior. Thus, it appears that dynamic positive affect experienced with a close friend can powerfully engage neural social-affective circuitry and, in combination with positive affect behavior, reveal individual differences in the potentially problematic reward-driven behavior that peaks at adolescence (Dahl, 2004).

Reciprocal positive affect with a close friend predicted risky behavior only in combination with left VLPFC response to the friend’s affect, indicating that positive affect might be especially meaningful in the context of sensitivity in associated neural systems. The left VLPFC, while not a direct focus of our hypotheses, is an intriguing player in affective, reward, and self-relevant processing. Recent work has linked individual differences in the function of left VLPFC—in a subregion similar to that identified in this study—to trait impulsive sensation seeking (Chase et al., 2017) and adolescents’ rule breaking behaviors (Bebko et al., 2014). Left VLPFC is also a putative biomarker of bipolar disorder, a form of mental illness notable for excessive reward-driven behavior (Phillips and Swartz, 2014). This region also plays a role in risky choices (Eshel et al., 2007) and effortful affect regulation (Braunstein et al., 2017; Phillips et al., 2008). Also, while interpretations of BOLD response as trait-like should be undertaken with caution given limited test-retest reliability, the region of left VLPFC in which we observed results is emerging across studies as an indicator of stable tendencies toward sensation-seeking. Left VLPFC involvement could perhaps contribute to adolescents’ risky behavior through the necessity of integrating reward, social, and self-relevant information when modulating the pursuit of rewarding goals.

Contrary to previous findings on the neural correlates of adolescents’ risky behavior (e.g. Braams et al., 2015), we did not observe ventral striatum response. This was the case even when we applied a more liberal statistical threshold. While the ventral striatum plays a central role in basic reward responding and reward learning (Haber, 2016), it might not be as critically involved in responding to complex reward stimuli that also require social and self-processing. In addition, the ventral striatum appears to play an important role in learning (or prediction error) that involves social reward contingencies (Lockwood et al., 2016; Will et al., 2017). Given that our task was designed to assess neural response to social feedback without creating contingencies that might be violated by such feedback, it likely did not engage the ventral striatum in this way. Our exploratory analyses also found that ventral striatum response did not differ between the best-friend positive affect condition and the unfamiliar-peer positive affect condition, suggesting that familiarity might not be sufficient to elicit response.

Similarly, our fMRI task did not elicit response in some other regions putatively involved in social reward (e.g. VMPFC). Response in some expected regions also did not moderate the association between positive affect and risky behavior (e.g. temporoparietal junction). Perhaps these regions contribute to risky behavior but do not serve as mechanisms of the association between sensitivity to peer social reward and engagement in activities such as drug use, physical fights or inconsistent condom use.

Several methodological issues are worth noting. First, our outcome variable was a cross-domain composite of risky behavior (Youssef et al., 2016) measured through self-report. Including real-time, objective measures of risky behavior in natural environments will be valuable in future studies. Second, our coding system yielded a general construct for positive affect, but facets of positive affect such as affection, happiness, excitement and contentment may function differentially in adolescents’ peer social interactions. Third, we did not assess friendship quality and thus were not able to incorporate it into analyses of brain–behavior associations. Fourth, positive fMRI stimuli tended to be segments from the past-focused conversation, whereas neutral stimuli tended to be segments from the future-focused conversation. This was unintended but could reflect the power of real experiences over imagined experiences to elicit positive affect. Fifth, our fMRI paradigm allows a potential role for memory in the best-friend conditions but not the unfamiliar-peer conditions, as participants only took part in the best-friend conversation and actors performed in the unfamiliar-peer conversation. Our analyses are likely not influenced by this difference as they focused on best-friend conditions, but future versions of this paradigm could include an alternative control condition with interactions between the participant and an unfamiliar peer. Finally, risky behavior was defined by potential harm to health or safety, rather than as general impulsivity or adaptive reward-seeking, either of which could have a different pattern of association with neural and behavioral response to best friends’ affect.

In all, this study points to the value of examining brain–behavior interactions when investigating behaviors relevant to adolescents’ mental and physical health. Next steps will include extending this research to larger samples, clinical samples, peer groups as well as dyads and real-time risky behavior. A key, related developmental question is whether associations between peer reward and risky behavior are specific to adolescence. Our findings also represent a methodological advance in assessing adolescents’ affective processing and underscore the importance of ecologically valid, personally relevant paradigms for studying this developmental period.

Link to Article

Introduction

Adolescents are well-known for engaging in risky behavior, driven by a desire for intense, rewarding experiences (Dahl, 2004; Steinberg, 2008). This includes activities like reckless driving, unsafe sex, and drug use, alongside neglecting preventative measures like seatbelts or helmets (CDC, 2010).

The influence of peers is crucial in this context, as adolescents are more likely to engage in risky behavior when with peers (Albert et al., 2013). This coincides with significant social development during adolescence, including the formation of romantic relationships and close friendships (Choukas-Bradley et al., 2015). This period is marked by social reorientation, where peer relationships take precedence (Blakemore and Robbins, 2012; Somerville, 2013), making peer influence a stronger motivator than parental influence or individual decision-making (Liao et al., 2013).

Social reward, particularly positive experiences shared with peers, is central to this increased risk-taking. Adolescents demonstrate heightened sensitivity to reward and peer contexts compared to adults, taking more risks in peer settings, being more susceptible to reward-related distractions, and displaying stronger neural responses to pleasurable stimuli (Chein et al., 2011; Somerville et al., 2011; Galván et al., 2006). This suggests individual differences in sensitivity to peer social reward could contribute to varying levels of risky behavior.

Behaviorally, conversations with friends can encourage rule-breaking, potentially through social rewards (Dishion et al., 1996). Reciprocal positive affect, the tendency to respond to another's positive emotions with their own, may be a key factor. Close friendships provide a setting for such interactions and could offer insight into neural responses to social reward.

Neurally, individual differences in reward circuitry are linked to both risky behavior (Galván et al., 2007) and susceptibility to peer influence (Pfeifer et al., 2011). This suggests neural sensitivity to peer reward could be a vulnerability factor for risky behavior. Studies have shown a connection between left VLPFC response to reward and traits associated with risky behavior (Chase et al., 2017).

Adolescents' increased focus on social interaction and risk-taking are likely driven by the development of reward, social, and self-regulatory networks (Casey et al., 2011). This includes the ventral striatum, amygdala, medial prefrontal cortex (PFC), temporoparietal junction, posterior cingulate, precuneus, ventromedial PFC, anterior insula, and ventrolateral prefrontal cortex (VLPFC). These interconnected networks undergo significant development during adolescence, influencing social-affective processing and contributing to behaviors like risk-taking.

This study hypothesizes that individual differences in adolescents' neural responses to peer social reward, specifically positive affect shared with a close friend, can predict their engagement in risky behaviors. We posit that adolescents exhibiting both heightened neural responses and reciprocal positive affect in response to social reward are more likely to engage in risky behaviors.

To test this, we developed a novel fMRI social reward paradigm using personalized video stimuli of participants interacting with close friends. This paradigm aims to capture both behavioral (reciprocal positive affect) and neural responses to peer social reward, offering a more ecologically valid approach to understanding the social and affective neuroscience of adolescence.

Materials and methods

Participants and protocol

Fifty typically developing adolescents (ages 14-18; M = 16.22, SD = 1.4; 48.6% female; 68% European American, 27% African American, 5% mixed race) participated in the study, with no history of psychiatric or serious health issues. Participants with incomplete data (fMRI, behavioral, etc.) were excluded. All participants identified a same-gender best friend who also participated in the study.

Participants completed self-report measures and underwent an MRI scan. The majority (57%) were scanned within one month of their lab visit. Ethical approval was obtained, and informed consent was acquired from all participants and their parents/guardians.

Risky behavior

The Youth Risk Behavior Survey (YRBS; CDC, 2010) was used to assess risky behavior across six domains: injury and violence, sexual behaviors, alcohol and drug use, tobacco use, dietary behaviors, and physical activity. A single factor representing a range of risky behaviors was derived from 10 YRBS items (Youssef et al., 2016), including substance use, seatbelt use, sexual behavior, and suicidality.

Reciprocal positive affect behavior

Adolescents and their best friends engaged in a 10-minute video-recorded conversation, split into two 5-minute segments focusing on shared positive experiences. Affective behavior was coded from the videos using the LIFE coding system (Hops et al., 1995) to measure reciprocal positive affect.

Reciprocal positive affect was calculated using the Generalized Sequential Querier Program (Bakeman and Quera, 2011), determining the probability of a participant's positive affect contingent upon their friend's positive affect. Adjusted residuals reflecting associations greater or lower than chance were used in analyses.

Neural response to friend positive affect

Best Friend fMRI task

A personalized fMRI task was developed, utilizing six video clips of the participant's best friend and six of an unfamiliar, same-gender adolescent in a block design. Blocks alternated between positive and neutral affect and friend/unfamiliar peer. Participants indicated their engagement by pressing a button at the start of each clip.

Stimuli were selected based on positive and neutral affect coding of the conversation videos using an adaptation of the AFFEX coding system (Izard et al., 1983). The primary contrast of interest was friend positive affect > friend neutral affect.

fMRI acquisition and preprocessing

Scanning was conducted using a Siemens 3-T TIM Trio scanner. Structural and functional images were acquired using standard procedures. Preprocessing and analysis were done using SPM8, incorporating realignment, normalization, smoothing, and artifact detection.

Data analyses

First-level analyses used paired t-tests to examine the contrast of interest within each participant. Second-level analysis employed a within-sample t-test masked for regions responsive to social stimuli, identified through Neurosynth (Yarkoni et al., 2011).

Moderation analyses used the PROCESS macro for SPSS (Hayes, 2012) to investigate the hypothesis that the interaction between behavioral and neural responses to peer social reward predicts risky behavior. The sliding linear model (SLIM; Wang et al., 2011) controlled for Type I error. Exploratory analyses examined the specificity of effects to the best-friend context and investigated potential ventral striatum involvement.

Results

Neural response to the Best Friend task

Significant neural response to best friend positive affect relative to neutral affect was observed in nine regions associated with social processing: bilateral VLPFC, dorsomedial PFC, bilateral superior temporal gyrus, bilateral middle temporal gyrus, anterior insula, and fusiform gyrus (Table 1, Figure 2). Whole-brain analyses confirmed these findings.

Regression analyses showed no direct relationship between neural response to best friends' positive affect and reciprocal positive affect behavior or risky behavior. Similarly, no direct relationship was found between reciprocal positive affect and risky behavior.

Interaction of neural response and behavioral response to friend positive affect as a predictor of risky behavior

The moderation model for left VLPFC was significant (R2 = 0.23, F(3, 45) = 4.43, P = 0.008, SLIM P = 0.01). Further analysis revealed that the interaction between left VLPFC response and contingent positive affect behavior predicted risky behavior (ΔR2 = 0.19, F(1, 45) = 11.21, P = 0.002, SLIM P = 0.03). This suggests that left VLPFC response moderates the relationship between shared positive affect and engagement in risky behavior.

The Johnson–Neyman technique (Preacher et al., 2006) revealed significant interaction effects at both low and high levels of left VLPFC response. This indicates that risky behavior was associated with both higher neural and behavioral responses, and lower neural and behavioral responses to best friends' positive affect. No other regions significantly moderated this relationship.

Exploratory analyses

Analyses investigating the specificity of findings to the best-friend context showed no significant response to unfamiliar peer positive affect. However, a stronger response was observed for best friends' positive affect compared to unfamiliar peers' positive affect in a cluster including the VLPFC, superior temporal gyrus, and inferior frontal gyrus.

Further exploration of ventral striatum involvement, despite its lack of significance at the predetermined threshold, revealed no significant clusters even at a lower threshold. Analyses using an anatomical mask also showed no significant findings.

Discussion

This study highlights the importance of examining both neural and behavioral responses to peer social reward in understanding adolescents' risky behavior. The findings demonstrate that the interaction between left VLPFC response and reciprocal positive affect during interaction with a close friend predicts engagement in various risky behaviors. This suggests two potential pathways: one characterized by high susceptibility to social reward and intense positive affect, and the other by relative indifference to social reward.

The former pathway, more prevalent in this sample, suggests a possible link between heightened sensitivity to social reward and a desire to seek out and maintain positive affect through risky behavior. While potentially detrimental to health, this sensitivity may also contribute to positive social development (Telzer, 2016). Conversely, the latter pathway, observed less frequently, could reflect a tendency towards blunted reward responding and a need to compensate through risky behavior (Zuckerman, 1996). This pathway might be more pronounced in clinical populations.

The use of an ecologically valid, dynamic paradigm focusing on real-life interactions with close friends is a significant strength of this study. This approach captures the complexities of social reward processing beyond traditional paradigms.

The finding that left VLPFC response moderates the relationship between positive affect and risky behavior aligns with its role in impulse control, affect regulation, and reward processing (Chase et al., 2017; Bebko et al., 2014; Phillips and Swartz, 2014; Eshel et al., 2007; Braunstein et al., 2017; Phillips et al., 2008). This region might be crucial for integrating reward, social, and self-relevant information when navigating reward-seeking behavior.

The absence of significant ventral striatum findings, even with relaxed thresholds, suggests that while crucial for basic reward processing (Haber, 2016), it might not be centrally involved in complex social reward processing without explicit reward contingencies (Lockwood et al., 2016; Will et al., 2017).

Future research should address limitations such as the use of self-reported risky behavior, a general measure of positive affect, and the lack of assessment of friendship quality. Further investigation into specific facets of positive affect, objective measures of risky behavior, and the role of friendship quality could enhance our understanding of this complex relationship.

This study emphasizes the significance of brain-behavior interactions in adolescent risk-taking. Utilizing ecologically valid paradigms and investigating individual differences in neural and behavioral responses to social reward are crucial for advancing our knowledge of the social and affective neuroscience of adolescence. Future research expanding to larger, diverse samples and incorporating real-time behavioral measures will be critical in furthering this understanding.

Link to Article

How Positive Feelings with Friends are Related to Risky Behavior in Teenagers

Introduction

We know that teenagers are more likely to take risks than adults or children (Dahl, 2004; Steinberg, 2008). This is often because teenagers are more excited by new and intense experiences (Steinberg et al., 2008). This can lead them to do risky things, such as driving dangerously, having sex without condoms, and using drugs. They are also less likely to do things that could keep them safe, like wearing seatbelts or helmets (CDC, 2010).

Teenagers are especially likely to take risks when they are with their friends (Albert et al., 2013). This makes sense because adolescence is a time when teenagers' relationships are changing a lot. They start to have romantic relationships and develop strong friendships. Their status among their peers also becomes more important (Choukas-Bradley et al., 2015). Researchers believe that these changes are driven by a process called social reorientation, where the teenage brain and behavior change to prioritize peer relationships (Blakemore and Robbins, 2012; Somerville, 2013). This means that teenagers are more likely to be influenced by their friends than by their parents or their own judgment, especially when it comes to risky behavior (Liao et al., 2013).

One of the main reasons why teenagers take risks with their friends is because of something called social reward. Social rewards are experiences that make us feel good, especially when they happen with people we like. Studies have shown that teenagers are very sensitive to social rewards. For example, compared to adults, teenagers take more risks while driving in a driving simulator if they are with their friends (Chein et al., 2011). They are also more easily distracted by rewards when they are trying to focus on a task (Somerville et al., 2011). Brain imaging studies have found that teenagers' brains show a stronger response to rewarding stimuli, especially in areas of the brain that are important for reward processing, such as the ventral striatum (Galván et al., 2006).

Previous research suggests that the way teenagers talk to their friends can actually encourage risky behavior (Dishion et al., 1996). This could be because of social rewards. For example, teenagers who are good at responding to their friends' positive emotions with their own positive emotions might be more likely to engage in risky behavior.

Brain imaging studies have shown that individual differences in the way teenagers' brains respond to rewards are linked to their likelihood of taking risks (Galván et al., 2007) and being influenced by their peers (Pfeifer et al., 2011). Specifically, the way the ventrolateral prefrontal cortex (VLPFC) responds to rewards has been linked to personality traits related to risk-taking (Chase et al., 2017).

The goal of this study was to see if teenagers' brain responses to positive emotions from their friends could predict how likely they were to engage in risky behavior. We thought that teenagers who showed a strong brain response to their friends' positive emotions, and who were also good at responding to those emotions with their own positive emotions, would be the most likely to engage in risky behavior.

To test this, we created a new brain imaging task called the Best Friend fMRI task. In this task, teenagers watched videos of their best friend and videos of unfamiliar teenagers. The videos were designed to make the teenagers feel either positive or neutral emotions. We also measured teenagers' risky behavior using a survey.

Materials and methods

Participants and protocol

The study included 50 teenagers between the ages of 14 and 18. They were all healthy and had no history of mental health problems. The teenagers all had a best friend who also participated in the study.

The teenagers came to the lab and filled out some surveys. They also had their brain scanned while they completed the Best Friend fMRI task.

Risky behavior

Teenagers' risky behavior was measured using a survey called the Youth Risk Behavior Survey (YRBS; Center for Disease Control and Prevention (CDC), 2010). The YRBS asks about a variety of risky behaviors, such as substance use, sexual behavior, and fighting.

Reciprocal positive affect behavior

The teenagers and their best friends were videotaped while they had a 10-minute conversation. The teenagers were asked to talk about the most fun they had ever had together and to plan a fun event together.

The videos were coded for how often the teenagers responded to their friends' positive emotions with their own positive emotions. This was called reciprocal positive affect.

Neural response to friend positive affect

Best Friend fMRI task

This task was designed to measure how teenagers' brains responded to seeing their friends express positive emotions.

The teenagers watched videos of their best friend and videos of unfamiliar teenagers. The videos were designed to make the teenagers feel either positive or neutral emotions.

The teenagers' brain activity was measured using fMRI while they watched the videos.

fMRI acquisition and preprocessing

The teenagers' brains were scanned using a Siemens 3-T TIM Trio scanner.

Data analyses

First, we looked at which areas of the brain were more active when the teenagers were watching videos of their best friend expressing positive emotions compared to when they were watching videos of their best friend expressing neutral emotions.

Next, we tested whether teenagers' brain responses to their friends' positive emotions interacted with their reciprocal positive affect behavior to predict their risky behavior.

Results

Neural response to the Best Friend task

The brain imaging results showed that several areas of the brain were more active when the teenagers were watching videos of their best friend expressing positive emotions compared to when they were watching videos of their best friend expressing neutral emotions. These areas included the VLPFC, dorsomedial PFC, superior temporal gyrus, middle temporal gyrus, anterior insula, and fusiform gyrus (Table 1 and Figure 2). These areas of the brain are all involved in processing social information and rewards.

Interaction of neural response and behavioral response to friend positive affect as a predictor of risky behavior

We found that teenagers' brain responses to their friends' positive emotions, in combination with their reciprocal positive affect behavior, predicted their risky behavior. Specifically, teenagers who showed a strong brain response to their friends' positive emotions and who were also good at responding to those emotions with their own positive emotions were the most likely to engage in risky behavior. However, we also found that teenagers who showed a weak brain response to their friends' positive emotions and who were not good at responding to those emotions with their own positive emotions were also more likely to engage in risky behavior.

This suggests that there might be two different pathways to risky behavior in adolescence. Some teenagers might engage in risky behavior because they are very sensitive to social rewards and get a lot of pleasure out of positive social interactions. These teenagers might take risks because they are looking for exciting and rewarding experiences. Other teenagers might engage in risky behavior because they are not very sensitive to social rewards and have a hard time experiencing pleasure from positive social interactions. These teenagers might take risks because they are bored or because they are trying to feel something, even if it is negative.

Exploratory analyses

We also did some additional analyses to see if our findings were specific to the best-friend context. We found that teenagers' brains responded more strongly to positive emotions from their best friends than to positive emotions from unfamiliar teenagers. This suggests that the relationship between social rewards and risky behavior might be especially strong in the context of close friendships.

Discussion

This study found that teenagers' brain responses to their friends' positive emotions, in combination with their reciprocal positive affect behavior, predicted their engagement in a variety of risky behaviors. This suggests that teenagers who are very sensitive to social rewards, and who are also good at responding to their friends' positive emotions with their own positive emotions, might be the most likely to engage in risky behavior. However, we also found that teenagers who were not very sensitive to social rewards, and who were not good at responding to their friends' positive emotions, were also more likely to engage in risky behavior. This suggests that there might be two different pathways to risky behavior in adolescence.

These findings have important implications for understanding and preventing risky behavior in teenagers. It is important to identify teenagers who might be at risk for engaging in risky behavior, and to develop interventions that can help them make healthier choices.

Link to Article

Introduction

You know how teenagers get a bad rap for doing risky stuff? It's true, they often look for big thrills that could be bad for their health and safety (Dahl, 2004; Steinberg, 2008). This comes from a need for excitement, which is super strong in teenagers (Steinberg et al., 2008). Think driving too fast, having sex without protection, or using drugs. They also might not do things that keep them safe, like wearing seat belts or helmets (CDC, 2010).

Here's the thing: friends have a HUGE influence on risky behavior. Stuff like reckless driving, substance use, and getting in trouble with the law is way more likely when teenagers are with their friends (Albert et al., 2013). This makes sense because the teenage years are all about big changes in friendships and relationships. Romantic relationships start, close friendships grow, and fitting in with friends becomes super important (Choukas-Bradley et al., 2015). Basically, teenagers' brains and behaviors start prioritizing friends over everything else (Blakemore and Robbins, 2012; Somerville, 2013). So, yeah, friends have more sway than parents or even personal choices, especially when it comes to risky stuff (Liao et al., 2013).

Think about it: having fun with friends, especially when everyone's in a good mood, is a big deal for teenagers, maybe even a reason they take risks. Studies back this up, showing that teenagers are more sensitive to rewards and being around friends. Compared to adults, teenagers take more risks while driving with friends (Chein et al., 2011), get distracted more easily by rewards (Somerville et al., 2011), and their brains show a bigger response to good things (Galván et al., 2006). Some teenagers get a bigger sense of reward from friendships, so they might do more risky things to impress their friends or keep the good times rolling.

Here’s how it can play out: hanging out with friends, especially close friends, can lead to more rule-breaking (Dishion et al., 1996). How? When someone's happy, it can rub off on the people around them. This back-and-forth happiness is called reciprocal positive affect. The more sensitive a teenager is to this, the more likely they are to take risks.

Our brains have reward circuits, and how these circuits work can be linked to risky behavior (Galván et al., 2007) and how much friends influence us (Pfeifer et al., 2011). So, if someone's brain is wired to find social rewards super rewarding, it could make them more likely to take risks during their teenage years. In fact, there's evidence that a part of the brain called the left VLPFC, which lights up when we experience rewards, might be connected to risky behavior (Chase et al., 2017).

This study focused on the idea that teenagers' brains react differently to their friends' good vibes. We figured that the stronger their brain reacted to a friend's happiness, the more likely they were to do risky things.

To test this, we created a special brain scan task where teenagers watched videos of their best friend being happy. We then compared their brain activity to how often they did risky things in real life.

Materials and methods

Participants and protocol

We studied 50 normal teenagers between 14 and 18 years old (average age = 16.22 years) who had no history of mental health problems. They were mostly female (48.6%) and mostly white (68%). Each teenager brought their best friend of the same gender (whose average age was 15.82 years).

During the study, teenagers filled out questionnaires and had their brains scanned. The brain scans happened pretty soon after the questionnaires, usually within a month. We got permission from everyone involved (teenagers and parents) before starting.

Risky behavior

Teenagers completed a questionnaire called the Youth Risk Behavior Survey [YRBS; Center for Disease Control and Prevention (CDC), 2010]. This survey, which is given to high schoolers, asks about six types of risky behavior: things that could cause injuries, sexual behaviors, alcohol and drug use, smoking, unhealthy eating, and not getting enough exercise. These behaviors are considered risky because they could hurt their physical health (like getting an STI) or mental health (like becoming addicted to drugs).

We focused on ten questions from the YRBS that covered a range of risky behaviors like smoking cigarettes every day, wearing seatbelts, how many sexual partners they've had, and getting into physical fights (You can see all the questions in Supplementary Table S1). We used these questions to calculate a risky behavior score for each teenager. The higher the score, the more risks they took.

Reciprocal positive affect behavior

Each teenager and their best friend were videotaped having a 10-minute conversation. They talked about the most fun they'd ever had together and planned a fun event for the future. We picked these topics because they usually make people happy.

Experts watched the videos and coded how much positive emotion each person expressed, like being happy or caring. We then looked at reciprocal positive affect - basically, how much one person's good mood rubbed off on the other person. This helped us see if someone was more likely to be happy when their friend was happy.

Neural response to friend positive affect

Best Friend fMRI task

We created personalized videos for each teenager to watch while we scanned their brains. The videos showed six clips of their best friend and six clips of another teenager they didn't know (but who was the same gender) The clips were shown in a random order, with some showing positive emotions and some showing neutral emotions. While they watched, teenagers pressed a button at the start of each clip to make sure they were paying attention.

Remember the video of the teenagers talking with their friends? Experts watched those videos and picked out 20-second clips where the friend was showing either strong positive emotions (like laughing or smiling) or neutral emotions (like having a neutral facial expression). We then showed these clips to the teenagers while they were in the brain scanner.

We also showed them clips of other teenagers showing positive and neutral emotions. These clips came from actors who were trained to act like they were best friends having a conversation. This was important so that the teenagers in the scanner wouldn't recognize anyone in the videos except for their actual best friend.

We wanted to see how their brains reacted to their friend's positive emotions. So, we compared their brain activity when they were watching their friend being happy to when they were watching their friend being neutral.

fMRI acquisition and preprocessing

We used a special machine called an fMRI scanner to take pictures of the teenagers' brains while they were watching the videos. The fMRI scanner uses magnets to measure brain activity. We used a specific type of scan called a "BOLD" scan, which detects changes in blood flow in the brain. When a part of the brain is more active, it needs more blood, and the BOLD scan can pick that up.

We also took regular pictures of their brains for reference. Before we analyzed the brain scans, we had to do some processing to make sure the images were clear and lined up correctly.

Data analyses

First, we looked at the brain data for each teenager to see which parts of their brain were more active when they watched their friend being happy compared to neutral.

Then, we zoomed in on brain regions that are usually involved in social stuff, like recognizing faces, understanding emotions, and feeling rewarded.

Finally, we used fancy statistics to see if there was a link between (1) how much their brain reacted to their friend's happy clips, (2) how much positive emotion they showed when interacting with their friend, and (3) how many risks they took in real life.

Results

Neural response to the Best Friend task

When teenagers saw their best friend being happy (compared to neutral), nine areas of their brains lit up. These areas are known to be involved in social stuff, like the VLPFC, which is linked to self-control, and the dorsomedial PFC, which helps us understand other people's thoughts and feelings. (Table 1 and Figure 2).

Interaction of neural response and behavioral response to friend positive affect as a predictor of risky behavior

Here's the interesting part: we found a link between teenagers' brain activity, their behavior, and their risky behavior, specifically in a part of the brain called the left VLPFC.

Remember how we measured how much someone's happiness rubbed off on their friend (reciprocal positive affect)? We found that this "emotional contagion" was related to risky behavior, BUT only when we factored in how active their left VLPFC was.

It turned out that teenagers who were both highly responsive to their friend's happiness (high reciprocal positive affect) and had a strong brain response in the left VLPFC were more likely to engage in risky behavior. This suggests that they might be more easily influenced by their friends' emotions and more likely to seek out thrills.

But here's the twist: teenagers who were less responsive to their friend's happiness and had weaker left VLPFC activity were also more likely to engage in risky behavior! This was unexpected. It could be that these teenagers are less sensitive to rewards in general and are looking for intense experiences to feel something. Or, they might be more likely to take risks for reasons that have nothing to do with their friends.

Exploratory analyses

We wanted to be sure that these findings were specific to best friends and not just any friend. So, we looked at whether teenagers' brains reacted differently to the positive emotions of their best friend versus a random teenager they didn't know. It turned out that they did! Their brains showed a stronger response to their best friend's happiness. This shows that the type of relationship really matters!

Discussion

This study showed that when teenagers see their close friends being happy, it can light up the reward centers in their brains, especially the left VLPFC. And, this brain activity, combined with their own tendency to mirror their friend's happiness, could predict how likely they were to do risky things.

Here's what we think might be going on:

• High reward sensitivity: Teenagers who are super responsive to their friend's happiness might be more likely to seek out exciting experiences, even if they're risky. It's like they get a rush from seeing their friends happy and want to keep that feeling going.

• Low reward sensitivity: Teenagers who are less responsive to their friend's happiness might need more intense experiences to feel the same level of excitement. They might be more likely to take risks because they're bored or because they don't find everyday activities rewarding enough.

It's important to remember that this study doesn't mean that all teenagers who are highly responsive to their friends are destined to make unhealthy choices. Some teenagers might be really good at managing their impulses, while others might need extra support.

This study also highlighted the power of friendship during the teenage years. Friends can have a big influence on our emotions and behavior. But by understanding how our brains react to social rewards, we can learn to make healthier choices.

Link to Article

Introduction Teenagers are known for doing risky things, like driving too fast or trying drugs. Teenagers are more interested in exciting experiences than adults or children, even if those experiences could be dangerous. This means they might drive dangerously, have sex without protection, or use drugs. They also might not do things that keep them safe, like wearing seat belts or helmets.

Friends are a big influence on teenagers' risky behavior. When teenagers are with their friends, they are more likely to do risky things, like driving recklessly, using drugs, or getting in trouble with the law. As teenagers grow up, their friendships become more important. They start to have boyfriends and girlfriends, develop close friendships, and care more about what their peers think of them. Experts believe that teenagers go through a "social reorientation" during this time. This means that their brains and behaviors change, and peer relationships become super important to them. It's not surprising then that teenagers are more likely to listen to their friends than their parents when it comes to risky behavior.

Having fun with friends, especially when everyone is happy and excited, is a big reason why teenagers take risks. Studies have shown that teenagers are more sensitive to rewards and the presence of peers than adults. For example, teenagers take more risks when driving in a simulator if their friends are watching. They also have a harder time focusing on tasks if there are rewards involved. This is because the reward centers of their brains, like the ventral striatum, are more active than adults'. Teenagers who really enjoy spending time with their friends and care about their friends' opinions may be more likely to engage in risky behavior. This is because they might feel pressured to impress their friends or do what their friends are doing, even if it's risky.

When teenagers talk to their friends, they might encourage each other to break rules. This could be because of something called "reciprocal positive affect," which is when someone responds to another person's positive emotions with their own positive emotions. For example, if a teenager laughs at their friend's joke, that's reciprocal positive affect. Teenagers who are very sensitive to social rewards, like having their friends laugh at their jokes, might be more likely to engage in risky behavior. Close friends provide a good opportunity for teenagers to experience reciprocal positive affect, which is why they might be more likely to take risks when they're together.

The way a teenager's brain responds to rewards can also play a role in risky behavior and how easily they are influenced by their peers. Since teenagers are so focused on their peers and rewards, how their brains respond to those things could make them more likely to engage in risky behavior. One study found that teenagers who showed a lot of activity in a certain part of their brain (the left VLPFC) when they received a reward were also more likely to engage in risky behavior.

Risky behavior might be more likely in teenagers who are very sensitive to social rewards. This could be because of how their brains develop during adolescence. Different parts of the brain are involved in processing rewards, social information, and self-regulation. These parts of the brain work together to help teenagers understand and respond to social situations.

This study looked at how teenagers' brains responded to their close friends' positive emotions. The researchers wanted to see if these brain responses could predict risky behavior. They thought that teenagers who showed a strong response in their brains to their friends' positive emotions, and who also showed a lot of positive emotions themselves, would be more likely to engage in risky behavior.

Materials and methods Participants and protocol The study included 50 teenagers between the ages of 14 and 18 who had never had a mental health disorder or serious health problem. The teenagers were asked to bring a close friend of the same gender to the lab to participate in the study.

Risky behavior The teenagers filled out a survey called the Youth Risk Behavior Survey (YRBS) to measure their risky behavior. The survey asks questions about things like smoking, drinking, drug use, sexual activity, and fighting.

Reciprocal positive affect behavior The teenagers and their friends were videotaped having a 10-minute conversation about fun experiences they had shared or wanted to share in the future. Researchers watched the videos and coded the teenagers' and friends' emotions. This helped them measure "reciprocal positive affect," or how much teenagers responded to their friends' positive emotions with their own positive emotions.

Neural response to friend positive affect Best Friend fMRI task The teenagers then went into an fMRI machine, which is a big machine that takes pictures of the brain. While they were in the fMRI machine, the teenagers watched short videos of their friends and other teenagers. Some of the videos showed the teenagers' friends expressing positive emotions, while others showed their friends with neutral expressions. The researchers wanted to see how the teenagers' brains responded to seeing their friends happy.

fMRI acquisition and preprocessing The researchers used a special computer program to analyze the fMRI data. This program helped them see which parts of the teenagers' brains were most active when they were watching the videos of their friends.

Data analyses The researchers used statistical analyses to see if there was a relationship between the teenagers' brain responses, their behavior during the conversation with their friends, and their scores on the risky behavior survey.

Results Neural response to the Best Friend task The researchers found that certain parts of the teenagers' brains were more active when they were watching videos of their close friends expressing positive emotions. These areas of the brain are known to be involved in social processing. This means that these areas of the brain help us understand and respond to other people.

Interaction of neural response and behavioral response to friend positive affect as a predictor of risky behavior The researchers found that teenagers who showed a strong brain response to their friends' positive emotions and who also showed a lot of positive emotions themselves were more likely to engage in risky behavior. This was especially true for teenagers who showed a strong response in a particular part of their brain called the left VLPFC.

Exploratory analyses The researchers also did some additional analyses to see if their findings were specific to close friends or if they would also see similar results with unfamiliar peers. They found that teenagers' brains responded more strongly to positive emotions from their close friends than from unfamiliar peers.

Discussion This study found that teenagers who are very sensitive to social rewards, like seeing their close friends happy, might be more likely to engage in risky behavior. This is especially true for teenagers whose brains show a strong response to their friends' positive emotions. The study also found that teenagers' brains respond more strongly to positive emotions from close friends than from unfamiliar peers.

This study is important because it helps us understand why some teenagers are more likely to engage in risky behavior. It also shows that our brains and our behavior are connected, and that the people we spend time with can influence our behavior.

The researchers suggest that future studies should look at these findings in larger groups of teenagers and include teenagers with mental health problems. It would also be interesting to see if these findings are specific to adolescence, or if they also apply to adults.

Link to Article