Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

# Preference uncertainty accounts for developmental effects on susceptibility to peer influence in adolescence

## Introduction

Although peer influences on decision-making during adolescence have been widely investigated, several important questions remain unanswered. First, claims on social susceptibility and its real-life consequences in adolescents mostly rely on cross-sectional designs and modest sample sizes. Here, we applied a longitudinal design in a large cohort of adolescents and young adults to study peer influence on a well characterised task assessing inter-temporal preferences. Inter-temporal (‘delay’) discounting is a key measure of temporal impulsivity, which has been shown repeatedly to decrease from childhood to young adulthood28,29,30. Longitudinal designs are important for addressing developmental questions as they provide a basis for disentangling cohort or sampling effects from developmental trajectories.

Third, we were interested in the co-development of brain structures that are relevant for the expression of preference uncertainty. We previously found that social susceptibility during a delay discounting task is mediated by plasticity in the medial prefrontal cortex (mPFC)36. A key mechanism contributing to plasticity itself is myelination37, a process that extends throughout adolescence and into adulthood. Recent studies in humans have observed that individual differences in the unfolding of these myelin growth trajectories are sensitive to inter-individual differences, including the expression of psychopathology38,39, environmental effects40 and social cognition41, rendering myelin a candidate marker in the co-development of brain-behaviour relationships. Likewise, animal42,43,44,45 and human41,46 work has highlighted the importance of medial prefrontal myelin plasticity for socially relevant behaviour across development. This background provided a motivation to ask, using in-vivo myelin-sensitive magnetisation transfer MRI39, whether mPFC myelin is related to developmental effects on preference uncertainty, which in turn impacts on social susceptibility.

We show that social susceptibility, in terms of preference shifts in a delay discounting task, not only decreases with age cross-sectionally, but also longitudinally. Longitudinally, we demonstrate that susceptibility to peer influence in our experimental paradigm as measured at baseline predicts the quality of peer relationships over follow-up, which might point to an adaptive role of social susceptibility in healthy adolescents. Using our computational model, we show that such preference uncertainty, here in the paradigmatic case of delay discounting, decreases both with age and over the course of our longitudinal follow-up, and longitudinal change is strongest in the youngest of our sample. Crucially, both cross-sectional and longitudinal developmental effects on social susceptibility, in terms of preference shifts in our social delay discounting task, are accounted for by developmental changes in preference uncertainty, suggesting that higher preference uncertainty in younger adolescents is a key mechanism facilitating peer influence in teenagers. On the neural level, we demonstrate that a myelin marker in medial prefrontal cortex predicts a developmental decrease in preference uncertainty over our longitudinal follow-up period.

## Results

To probe a susceptibility to social influence, we used a social version of a delay discounting task (Fig. 1). In short, this task allows us to measures a person’s temporal discounting coefficient (how much less a future reward is worth, depending on the delay of its delivery) as well as changes in their discount function pre vs. post learning about someone else’s discount preferences34,36,47 (See Fig. 1 and ‘Methods’ for details). Here, in line with previous accounts34,48, we defined susceptibility to social influence as the degree of change in one’s own discount rate towards that of a social partner, following exposure to the discounting preferences of a social partner (log kself_phase3 – log kself_phase1).

Analyses on subjects’ choice behaviour, reaction times, delay discounting preferences, as well as effects of the ability to learn from others, and the effect of the other’s delay discounting preferences are reported in the Supplementary Information.

### Cross-sectional and longitudinal developmental effects of susceptibility to social influence

At baseline, we observed a significant negative association between social susceptibility (indexed as the change in discount rate upon learning about the discount rate of another agent) and age, such that socially induced preference shifts declined with age (r = −0.10, df = 782, t = −2.94, p = 0.003, Fig. 2a).

In a longitudinal analysis, we tested whether social susceptibility also changed intra-individually, within the ~1.5 year follow-up period. Indeed, a factor time (baseline vs. follow-up) significantly predicted social susceptibility in our delay discounting task in a linear mixed effects model (F(1,566.64) = 5.11, unstandardised estimate = 0.17, p = 0.02), along with baseline age. Social susceptibility decreased intra-individually over the ~1.5 year follow-up period. The interaction of baseline age with time was not nominally significant (F(1,567.06) = 3.78, p = 0.05).

In addition, we examined a sample of participants who were also tested 6 months apart (‘short follow-up’). This comprised a sub-sample of n = 55 of the total group who came to the lab three times (baseline, 6-month ‘short’ follow-up, 1.5-years ‘long’ follow-up), in the same manner as per our main sample (see ‘Methods’). Repeating the same analysis of longitudinal effects on social susceptibility (i.e. preference shifts in our delay discounting task), in this short follow-up sample, we observed a significant effect of time point (baseline, short follow-up, long follow-up; F(2,104.93) = 6.87, p = 0.002). Post-hoc analysis showed that susceptibility decreased significantly over the 18-month period (t = 3.47, unstandardised estimate = 0.70, p < 0.001), and between 6 and 18 month of follow-up (t = 2.88, unstandardised estimate = 0.58, p = 0.005), but critically did not do so significantly over the 6-month period from baseline (t = .59, unstandardised estimate = 0.12, p = 0.56).

### Association of susceptibility to peer influence with the development of peer relationships

In previous work, we observed a positive association of susceptibility to peer influence and sociability in a smaller cross-sectional study, including young adolescents5. Here we tested for this association again, including in a longitudinal manner, in our larger dataset. We used a bivariate latent change score model49 to test for a co-development of social susceptibility, as indexed by socially induced preference shifts in our delay discounting task, and quality of peer relationships, as measured by the Cambridge Friendship Questionnaire (CFQ50,51, freely available at https://osf.io/cf59r/). In a previous study, the CFQ was shown to predict psycho-social resilience in this sample51. The model showed that the perceived quality of peer relations increased from baseline to long follow-up (significant intercept of the latent CFQ change score: z = 2.28, standardised beta = 0.09, p = 0.02, standardised beta). There was no significant covariation of social susceptibility and peer relations at baseline (z = −0.77, standardised beta  = 0.03, p = 0.44). However, the latent change score model revealed a small, but significant, positive association between social susceptibility, as measured via our delay discounting task at baseline, on rate of change in peer relation development from baseline to the first follow-up. In effect, those who showed a higher tendency to shift their preferences towards their partners’ in our experimental paradigm, also reported larger gains in the quality of peer relations from baseline to follow-up (z = 2.12, standardised beta = 0.08, p = 0.03). See Supplementary Table 1 for full output of the model.

Previous findings4,5 showed that developmental effects on social susceptibility might be particularly pronounced in younger teenagers. Thus, we repeated the path analysis separately on age specific subsamples of our main sample, namely a younger (participants who were <18 at both baseline and long follow-up, n = 153) and an older (participants who were adults, i.e. ≥18 years old, at both measurement time points, n = 320) subsample. Comparing this model to a model where the path of interest (social susceptibility at baseline → quality change in peer relations at long follow-up) was constrained to be equal between the younger and older subsample, revealed a significant advantage of fitting age-dependent, sub-group-specific parameters (Log likelihood Ratio Test, Δχ2 = 4.47, Δdf = 1, p = 0.03). This indicates differences in a younger vs. older subgroup regarding the degree to which social susceptibility, as measured with our task at baseline, influences development of peer relations. Analysing the path of interest (baseline task measure of social susceptibility on rate of change in peer relation) separately for the younger (<18 years) and older (≥18 years) subgroup, revealed a significant effect of social susceptibility on social development in the adolescent (<18 years) group alone (z = 2.17, standardised beta  = 0.20, p = 0.03, see Fig. 3), whereas there was no significant coupling of social susceptibility on social development in the young adult (<18 years) group (z = 0.44, standardised beta = 0.02, p = 0.68). Results in the younger age group remained significant when controlling for age or sex (all betas > 0.21, all ps < 0.04). This suggests that greater susceptibility to social influence earlier in adolescence might be an important factor affecting development of integrative social relationships as we grow up.

A maladaptive trait, which has been extensively discussed in the context of peer influence in the literature16,17,18,19,52, is substance consumption. Thus, we repeated the structural equation model to analyse an association with the development of substance consumption (alcohol, cigarettes, cannabis). In this generally healthy sample, we did not find any significant association of social susceptibility with substance consumption (all zs ≤ |1.23|, all standardised betas ≤ |0.10|, all ps > 0.22).

### Computational modelling: developmental effects on preference uncertainty

One reason for conformity is informational influence, whereby humans use observational information to reduce uncertainty about what to like, even if, as in our task, this does not produce immediate material benefits31. We tested the hypothesis that the observed developmental reductions on social susceptibility, indexed by preference shifts in delay discounting towards the direction of the experimental partner, occur as a consequence of developmentally decreasing uncertainty about own preferences (‘preference uncertainty’). Thus, in a next step, we used a previously validated formal computational model34 to estimate an individual ‘preference uncertainty’ parameter (See Supplementary Information for effects of other parameters of the model). This allowed us to test for developmental effects on preference uncertainty and whether this accounts for the observed susceptibility to social influence.

Preference uncertainty was significantly associated with social shift at both measurement time points, consistent with an informational account of conformity (all rs > 0.50, all ts > 15.57, all ps < 0.001, see Supplementary Fig. 5).

At baseline, preference uncertainty negatively correlated with age (r = −0.15, t = −4.36, df = 780, p < 0.001, Fig. 4a).

A longitudinal linear mixed effects model revealed a significant effect of baseline age (F(1,564.53) = 14.22, p < 0.001), time (F(1,564.86) = 5.69, unstandardised estimate = 0.11, p = 0.02), and an interaction of baseline age and time (F(1,565.41) = 9.29, unstandardised estimate = 0.05, p = 0.002) on preference uncertainty. To visualise the latter interaction, we plot longitudinal (intra-individual) change as a function of age at baseline in Fig. 4b. Figure 4b suggests that relevant longitudinal changes in preference uncertainty were strongest in those that were ≤17 years of age at baseline.

In addition, we again analysed the reduced sample of participants who were also tested 6 months apart (‘short follow-up’). In line with findings in our main study sample, we observed a significant effect of time on preference uncertainty (F(2,105.29) = 3.57, p = 0.03) in the ‘short follow-up’ subsample. Post-hoc inspection revealed that preference uncertainty decreased within-person over the three measurement time points in the reduced subsample. Post-hoc tests showed that a contrast of first and last (1.5 year follow-up) time points was significant (t = 2.67, unstandardised estimate = 0.33, p = 0.009), whereas neither the change between baseline measurement and 6-month follow-up was significant (t = 1.23, unstandardised estimate = 0.149, p = 0.22), nor the change between 6-month and 1.5 year follow-up (t = 1.44, unstandardised estimate = 0.18, p = 0.15). This pattern of findings—more pronounced effects after a greater period of time had elapsed—seems inconsistent with changes being due to a mere training effect, as in this case we would expect a stronger change after 6 compared to 18 months.

### Mediation analyses: preference uncertainty accounts for developmental effects of social susceptibility

Our computational model of social shifts posits a specific mechanism by which social influence arises, namely a reduction in one’s own preference uncertainty by learning how someone else performs a task31,34. As preference uncertainty declines with age, and decreases longitudinally within person, this raises a possibility that age-related changes in social susceptibility (as found here in the paradigmatic case of a delay discounting task, as well as in previous studies using different behavioural paradigms) is driven by the age-related change in preference uncertainty. To test this, we first set up a model where we tested a possible mediation of the cross-sectional age effects on preference shift by preference uncertainty on baseline data. We found that an effect of age on preference shift in delay discounting was accounted for by the mediating effect of preference uncertainty, corresponding to a significant full mediation (significant proportion of mediation53z = 3.03, standardised beta= 0.723, p = 0.002, Fig. 5A). That is, the significant age effects on social susceptibility, as found here and in many previous studies, are, in this study, explained by age effects on preference uncertainty.

It is important to note valid concerns regarding cross-sectional mediation analyses and the degree of causal inference that can be drawn from them54. Here, we cautiously apply these analyses building on the temporal order of our task, as well as theoretical assumptions on informational influence including recent experimental studies demonstrating a role for uncertainty as a driving influence55.

In a next step, we examined the covariation of longitudinal change in preference uncertainty with longitudinal change in social susceptibility, using latent change score modelling. To do so, we changed the autoregressive and coupling effects to co-variances, to display and model the unconditional change scores. We observed a significant covariation of rates of change in both parameters (z = 3.82, standardised beta = 0.205, p < 0.001, Supplementary Table 2), in line with our assumption that development of social susceptibility is accounted for by development of preference uncertainty.

### Co-development of brain structural correlates and preference uncertainty

To establish whether there is a structural neural correlate for developmental effects on preference uncertainty, we used bivariate latent change score modelling testing for the co-development of cognitive and brain structural development in a subsample of subjects who underwent both our experimental task sessions and structural MRI (at baseline and after ~1.5 years, n = 186, see ‘Methods’). In a hypothesis-driven approach, we focussed on myelination, given its pivotal role in adolescent brain development and plasticity, and building on recent findings on its association with trait measures in adolescent development39. Medial prefrontal cortex (mPFC) myelin development has been demonstrated to be associated with socially relevant behaviour in several previous rodent42,43,44,45 and human41,46 studies. In a similar discounting task, utilising repetition-suppression to assess neural plasticity during fMRI, we previously showed that mPFC is the principal region expressing within-task plasticity as preferences shift36. We, therefore, focused on mPFC as a key region of interest, testing whether a myelin marker in this region relates to developmental effects on preference uncertainty. See Supplementary Information methods for pre-processing of the MRI data.

Using an anatomically defined mPFC mask (based on the Harvard-Oxford atlas, see Supplementary Information), we extracted estimates of a myelin-sensitive marker, Magnetisation Transfer saturation (MT). This allowed us to investigate a cross-domain coupling that captures the extent to which change in one domain is a function of the starting level in the other using latent change score modelling (model fit indices: pChi2 = 0.74, CFI = 1, RMSEA = 0.00, SRMR = 0.02). Sex, IQ and scanning site were added to the imaging analysis a priori as they were previously found to be associated with myelin-related measures39,40,56,57. We observed that baseline intra-cortical mPFC MT was predictive of the longitudinal change in preference uncertainty (z = −2.14, standardised beta = −0.13, p = 0.03, Fig. 6), i.e. more intra-cortical mPFC myelin (as approximated via MT) lead to a more pronounced reduction in preference uncertainty over the ~1.5 year follow-up.

Preference uncertainty did not significantly predict longitudinal brain development, and there was no covariance in rates of change in both domains (all standardised betas ≤ 0.10 all ps ≥ 0.13, see Supplementary Table 3 for full output of the latent change score model). We performed a control analysis using an ROI chosen on account of its high developmental relevance, based on our previous findings that in this cohort age-related and longitudinal changes in MT show a peak in angular gyrus39. Repeating the same latent change score model in a control region (based on a Harvard-Oxford atlas defined ROI of angular gyrus, model fit indices: pChi2 = 0.83, CFI = 1, RMSEA = 0.00, SRMR = 0.02), we did not observe any significant brain-behaviour coupling (all standardised betas ≤0.10, all ps ≥ 0.12).

## Discussion

In adolescence, a greater susceptibility to social influence is considered a driver of maladaptive real-life behaviour (e.g. drinking, reckless driving, delinquency, self-harming behaviours)12,13,14,15,18,19. Here, using a longitudinal design involving a large cohort of 14–24 year-olds, combined with a social delay discounting paradigm as well as quantitative brain imaging, we replicate a finding of increased social susceptibility in adolescents and characterise its computational and neuro-developmental correlates.

We replicate previous developmental effects on social susceptibility in the paradigmatic case of a delay discounting task, showing that social susceptibility decreases with age from adolescence to adulthood. We extend on this previous finding in a number of ways. First, we provide longitudinal evidence for a developmental decrease in susceptibility to social influences. Second, we show that our task-based measure of social susceptibility in young healthy adolescents predicts longitudinal improvements in peer relationships. Third, we outline cognitive and computational correlates for these effects, providing evidence in line with the idea that higher social susceptibility in younger adolescents may be explained by this population being more uncertain about their delay discounting preferences, thereby rendering them more prone to adopt others’ preferences. Over the course of development this ‘preference uncertainty’ decreases, which may in turn attenuate a need, and the potential impact of, social influences on one’s own behaviour. Last, we demonstrate a candidate neuro-developmental correlate of this effect by showing that a myelin-sensitive marker within mPFC, a key region mediating social preference shifts, predicts longitudinal change in preference uncertainty.

We replicate our previous finding, from a young adult sample, showing delay discounting preferences can be systematically changed by learning about another’s delay discounting preferences36,47. Extending these findings, we now show that the degree of a preference shift is, in this sample of 14–24 year-olds, most pronounced during younger adolescence, a developmental period characterised by significant social-affective transformations1. Importantly, our longitudinal design allowed us to demonstrate a within-person developmental decrease in this social susceptibility. A separate analysis in a retest sample showed that these observations are not likely to be due to a training effect.

In the context of our computational modelling analysis, a limitation is that model fit correlated with age, such that younger participants were fit less well by our computational model. In some sense, this is inherent in our research question, as the key variable we examine, preference uncertainty, drives choice variability. Our finding of both decreasing preference uncertainty and increasing model fit with age reflect that behaviour is less variable in older subjects within our sample. Varying goodness-of-fit of a specific model is a challenge for developmental modelling studies61,62 and may be indicative of either varying stochasticity or different cognitive processes underlying task behaviour at different ages. Potential solutions are now emerging in the literature, including novel data analysis strategies (e.g. data-driven approaches63) or experimental design specifically tailored to disentangle random from directed choices across development64.

In addition to preference uncertainty, our study also showed that features of the observed other influence how much young people shift towards this other’s preference. Thus, social susceptibility was stronger towards the other person if they were more patient than oneself. This fits with findings in the domain of risk preferences, which highlight that in an experimental set-up where choices were not observed, those in late adolescence are more influenced by safe than by risky information6,24.

Further limitations of this study relate to ecological validity, in that a social influence was implemented by a computerised agent rather than a real social interaction partner. The study did not include a non-social control condition. However, the experimental design followed closely our previous study in young adults of an age range overlapping with the older half of our sample36. That study included additional control experiments that provided evidence that social influence effects in this task cannot be explained by simple stimulus- or action-based reinforcement, but are depended on mentalizing36. On this basis we consider our conclusions are likely to remain valid, even if participants are influenced by anthropomorphic digital agents. Indeed, influence by digital agents seems particularly relevant in the context of adolescent development, where evidence points to this age group being susceptible to influence from digitalised social input via (social) media (including bots, computer algorithms and fictive characters in video games and series). These influences have recently been shown to exert a deleterious impact on adolescents’ and young adults’ mental health22,65,66 notwithstanding potential positive influences and impact on prevention67.

In our structural brain imaging analysis, we uncovered a neural correlate of preference uncertainty. MPFC myelin development has repeatedly been shown to be associated with social cognition and socially relevant behaviour42,43,44,45,41,46. Using the same social delay discounting task, during functional imaging, we have shown that mPFC expresses neuronal plasticity that predicts preference malleability36. Myelin maturation unfolds throughout adolescence and into young adulthood and is a key mechanism underlying neuronal plasticity37,68. This motivated a hypothesis that a marker of myelin in the mPFC would relate to developmental effects on preference uncertainty. Whilst we were primarily interested in the co-development of preference uncertainty and mPFC maturation, especially the unfolding of myelination based on an a priori hypothesis, we acknowledge that mPFC myelination is likely to be associated with other cognitive or socio-emotional functions, as for example shown previously in rodents69,70. Indeed, we found that baseline MT in this region was predictive of a greater reduction in preference uncertainty over time. In contrast, baseline preference uncertainty did not predict changes in MT, and there was no association between the rates of change in both measures. This suggests that the observed longitudinal reduction in preference uncertainty over time is accelerated when myelin in the mPFC is at a more mature absolute level. This finding suggests that structural differences in mPFC myelin precede cognitive development of preference uncertainty, a brain-behaviour dynamic that is also referred to as ‘structural scaffolding’ in developmental cognitive neuroscience71. It is compatible with an hypothesis that the current state of the brain (here, MT-related myelination) provides, in some sense, preconditions that facilitate cognitive growth71. The absence of a cross-sectional brain-behaviour correlation highlights that the level of mPFC myelin is not reflective of absolute behavioural maturity, but rather reflects a potential for subsequent growth, consistent with myelin’s role in plasticity. Furthermore, this effect is independent of baseline age, pointing towards inter-individual differences in the timing of this maturational process. These findings underscore the importance of brain structural maturation in cognitive development during adolescence.

Notably, the effect size of the developmental effects on susceptibility to social influence, as well as their association with peer relations or neuro-developmental markers, was lower than might be expected from theoretical accounts and lower sample size studies. This is consistent with reports on large-scale replication efforts of original findings in psychological science in bigger sample sizes, which reported only effect sizes on average of 1/2 the originally reported effects72. On the one hand, it highlights the need for larger sample sizes and replication in order to estimate meaningful effect sizes in the field of developmental psychology. As this is not a direct replication study of a previously reported investigation, it may, however, also be dependent on precise methodological details and the specific demographics of our sample, e.g. on the fact that our sample did not include very young adolescents for whom strong susceptibility effects have been reported previously4,5 (but also note73). Our findings stress the importance of longitudinal designs for developmental psychology. Indeed, the association of peer susceptibility with the quality of real-life peer relations, as well as neuro-developmental markers, were, albeit with moderate effect sizes, only observed within-person, but not across-participants, potentially due to the higher power of within-subject designs as compared to between-subject designs.

In sum, our study showcases the role of computational modelling and large-scale, longitudinal developmentally sensitive studies6,74, identifying the psychological mechanisms and neuro-developmental processes which underpin the phenomenon of susceptibility to social influences over adolescent to young adult development.

## Methods

### Main sample

The experimental task was delivered as part of a task battery administered to a sample of community dwellers between the ages of 14 and 24 in Cambridgeshire and London, as part of the Neuroscience in Psychiatry Network (NSPN) project60. Participants 16 years of age and above provided written informed consent. Participants below this age provided assent, and their parent (or legal guardian) provided written informed consent. The Cambridge Central Research Ethics Committee approved the study (12/EE/0250). Data for this task was available from n = 784 (401 female) participants for baseline. N = 738 of this baseline data has informed a previously published computational model validation paper34. Participants were 14.10–24.99 years old (mean = 19.05, sd = 2.96) at baseline. N = 569 (284 female) participants returned for a second assessment ~1.5 years later. Mean age at follow-up was 20.29 years (range: 15.11–26.48 years, sd = 2.97) while mean time between first and second assessment was 1.48 years (range: 0.98–2.62 years, sd = 0.30). Structural imaging and task data were available (and passed quality assessment) for n = 186 participants for both measurement time points (91 females).

### Retest ‘short follow-up’ subsample

A subsample of n = 55 participants completed the task three times, with an additional interim session after a ~6-month follow-up period. Considering all three measurement time points in this subsample allows us to approximate whether observed longitudinal differences are merely due to retest effects (i.e., practice effects, familiarity with the task). Even though changes after 6 months could also reflect development (during rapid periods of adolescent maturation)75, here we hypothesised a pattern of results consistent with a developmental (rather than a retest) account, that would show a stronger effect over the long-term (after ~1.5 years) than over the short-term (after ~6 months).

We used a social version of a delay discounting task (Fig. 1,34,36). The task was programmed in MATLAB 2012a using the Cogent graphic toolbox (http://www.vislab.ucl.ac.uk/Cogent/). The task consisted of three phases. In phase 1, participants played 60 trials of a temporal discounting task where they had to decide whether to choose between a smaller amount of money paid out immediately or a larger amount paid out at an indicated time in the future. Phase 1 decisions were used to determine their initial value kphase1 in a standard hyperbolic discounting model76:

$${V}_{D}=\frac{{R}_{D}}{1+{kD}}$$
(1)

where $${V}_{D}$$ is the delay-discounted value of a reward, R is the reward, D is the delay, and k is the hyperbolic discounting parameter. Based on empirical observations that the population follows approximately normal distributions in log (k) space, all reported analyses are based on log (k)34,36,77.

The 60 trials of phase 1 included 30 offer pairs from a standard set covering a wide range of values of k. Half of the trials were an interleaved set of 30 from an adaptive algorithm which calculated a probability distribution over possible values of k and then selected a pair of options likely to reduce the entropy of that distribution as much as possible (see34 for details). Participants were instructed to respond according to their own true preferences.

In phase 2, a second player was introduced. Participants were instructed to make choices in the same delay discounting task for the other player so as to learn the discounting preferences of the other. This preference of the other person was based on the baseline preference of our participant. In a between-subjects manner, the observee’s delay discounting preferences was manipulated such that the other was chosen to be either more or less patient than the participant himself. More specifically, in 2/3 of the cases, the observee was chosen to have kother shifted from kself_phase1 by one standard deviation towards the mean of the population distribution, and in 1/3 of cases away from it (in log space). This procedure was repeated for the longitudinal administration of the task, such that the direction of the preference of the other (more vs. less patient than our participant) could be either the same or different for baseline vs. follow-up. Participants received feedback as to whether their choice on behalf of the other was correct in terms of the other’s discounting preference. where correct choices were defined using a simulation of the other’s choice based on their discounting preference. In case the participant’s response matched the simulation’s prediction, the choice was coded as correct. These ‘learning about the other’ trials were presented until either the participant got eight correct answers out of the most recent 10 (minimum: 20 trials), or until 60 learning trials were completed. See Supplementary Results for additional analyses of learning performance.

In phase 3, we interleaved mini-blocks of 10 trials ‘choose for self’, which were as in phase 1, and 10 trials ‘choose for other’, in order to keep the other’s discounting preference. This allowed us to estimate social shift scores (log kself_phase3– log kself_phase1) that evaluate a change in delay discounting preference pre- vs. post learning about the other, and thus inter-individual differences in susceptibility to social influence. Note that age remained significantly associated with social susceptibility, also when using a relative measure of preference shift (log kself_phase3 – log kself_phase1) / log kself_phase1.

We informed participants that one of the ‘choose for self’ trials from the entire task would be chosen at random and the choice they made paid out for real at the appropriate delay. Participants were instructed that there was no financial incentive to make correct choices in the ‘choose for other’ trials. Note that this study was based on a previous study in our lab36 which included two carefully designed control conditions, providing evidence that participants in this task mentally simulate the other agent. In this previous study, the first control condition tested an alternative hypothesis that behavioural shifts simply reflect stimulus/action-based reinforcement or priming effects. This control condition consisted of the same stimuli and actions, but a framing of another agent with inter-temporal preferences was removed (see ‘Methods’ and Supplementary Results in36 for details). There, we demonstrated that no behavioural shifts occur in response to the other in this control condition, despite exposure to mathematically equivalent choices and similar outcomes. On the contrary, when participants played with an anthropomorphic computer agent that showed inter-temporal preferences, we found no significant difference in preference shift rates compared to a human partner. This constitutes evidence that, in our context, the ‘active ingredient’ of social influence most likely lies in ‘mentalising’ or ‘Theory of Mind’78,79 in relation to the other agent, rather than their physical nature.

### Computational modelling

Our model was first introduced and validated on the majority of baseline datasets of this study in34.

#### Rationale

The focus of our study here was to investigate the neuro-developmental evolution of social susceptibility. Here we could exploit the analytical tool derived from our previous work, well suited to test neuro-developmental hypotheses regarding delay discounting, preference uncertainty relevance of the other34. In short, we adopt a Bayesian approach to model a change in belief in one’s own delay discounting preferences as a function of (i) ‘preference uncertainty’ (as reflected in a participants’ choice variability, in simple terms this captures how uncertain a person is, in the delay discounting task, about their own preferences prior to any social exposure in the task); and (ii) the relevance of the social influence source.

We describe participants as holding uncertain beliefs $$p\left(k\right)$$, modelled by a Gaussian distribution, over their log-discounting coefficient. Thus, they are uncertain about their discounting preferences, sampling a new value from their belief distribution at each trial. The width of this belief distribution captures a subject-specific degree of preference uncertainty. The model further assumes that both the subject and the social partner come from the same reference distribution $$N\left(k,{\sigma }^{2}\right)$$, the width of which describes the ‘relevance of the other’ and is a fitted parameter of the model. By observing the preference choice data of the other, dO, subjects can update their own preference belief distribution $$p\left(k\right)$$ in light of what they learn about the other. In this Bayesian formulation, the more uncertain subjects are about their preferences (the more choice variability they show in the absence of social influence), the more they shift after learning about others. Thus, this model formalises the notion of informational conformity, namely conforming with others to reduce one’s own uncertainty. Please refer to our previous model validation study34 for the algorithmic implementation. There we found, based on model selection, that decision variability during the task was best described as reflecting uncertainty about discounting preference, as opposed to decision noise added after evaluation (as, for example, in the softmax rule). See Supplementary Information for parameter recovery results.

### Structural MRI

Of the participants who had completed the task twice, n = 186 (after quality assessment) were also scanned twice (baseline and 1.5 years follow-up) on identical Siemens Magnetom TIM Trio whole-body 3 T MRI scanners in Cambridge and London as per the quantitative multi-parameter mapping (MPM) protocol. This included a whole-brain multi-echo FLASH magnetisation transfer weighted contrast at 1 mm isotropic resolution (TR: 23.7, α = 6°, 176 sagittal slices, FOV = 256 mm × 240 mm, matrix = 256 × 240 × 176). Quantitative magnetisation transfer saturation (MT) maps were derived using biophysical models with the hMRI toolbox (www.hmri.info) for SPM (Wellcome Centre for Human Neuroimaging, London, UK, http://www.fil.ion.ucl.ac.uk/spm). These maps have been shown to correlate highly with histological measures of myelin80,81. Based on our a priori hypothesis, we created an anatomically defined mask of the mPFC based on the probabilistic Harvard-Oxford cortical structural atlas (thresholded at 30%). Mean MT values from within this mask region were extracted from each map using FSL (www.fmrib.ox.ac.uk/fsl/) as a proxy for intra-cortical mPFC myelination. See Supplementary Information for more details with regards to MRI pre-processing, quality assessment, and control analyses.

### Statistical analysis

All data were analysed using R 3.4.382 with R Studio Version 1.1.38383. Mixed models for longitudinal analyses included a categorical within-subject factor ‘measurement time point’ (baseline vs. follow-up) and a random intercept per subject. They were fit and p values were calculated based on a Kenward-Roger approximation for degrees-of-freedom using the R package afex84. In all mixed models, discounting preference of the other (more vs. less patient than the participant) was included as a covariate. When analysing age and developmental effects on social susceptibility, i.e. social shift towards the other, own discounting preference was included as a covariate. All continuous predictors were centred on zero. Post-hoc contrasts were computed using the R package emmeans85. We additionally fit general additive models using a thin plate regression spline as implemented in the R package mgcv86 to consider non-linear relationships between age and our outcome variables of interest. As in both these models effective degrees-of freedom ~1 indicated that the models reduced the effect of age to a linear term, we report linear mixed model results. Latent change score and mediation models were fit using the package lavaan87 with R code provided in49, freely available at https://osf.io/4bpmq/files/. In all models, we used a robust estimation procedure (‘mlr’ implemented in lavaan) to account for non-normality in the data. Plots were generated using ggplot288.

### Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.

## Data availability

All behavioural, self-report and MRI-ROI data analysed during the current study are available via the Open Science Framework (https://osf.io/jpks2/). Source data are provided with this paper.

## Code availability

Code for computational modelling as well as the statistical analysis is available at https://osf.io/jpks2/.

## References

1. Blakemore, S.-J. & Mills, K. L. Is adolescence a sensitive period for sociocultural processing? Annu. Rev. Psychol. 65, 187–207 (2014).

2. Foulkes, L. & Blakemore, S.-J. Is there heightened sensitivity to social reward in adolescence? Curr. Opin. Neurobiol. 40, 81–85 (2016).

3. Knoll, L. J., Leung, J. T., Foulkes, L. & Blakemore, S.-J. Age-related differences in social influence on risk perception depend on the direction of influence. J. Adolesc. 60, 53–63 (2017).

4. Knoll, L. J., Magis-Weinberg, L., Speekenbrink, M. & Blakemore, S.-J. Social influence on risk perception during adolescence. Psychol. Sci. https://doi.org/10.1177/0956797615569578 (2015).

5. Reiter, A. M. F., Suzuki, S., O’Doherty, J. P., Li, S. C. & Eppinger, B. Risk contagion by peers affects learning and decision-making in adolescents. J. Exp. Psychol. Gen. Advance online publicationhttps://doi.org/10.1037/xge0000512 (2019).

6. Ciranka, S. & van den Bos, W. Social influence in adolescent decision-making: a formal framework. Front. Psychol. 10, 1915, https://doi.org/10.3389/fpsyg.2019.01915 (2019). ARTN 1915.

7. Buritica, J. M. R., Heekeren, H. R. & van den Bos, W. The computational basis of following advice in adolescents. J. Exp. Child. Psychol. 180, 39–54, https://doi.org/10.1016/j.jecp.2018.11.019 (2019).

8. Chein, J., Albert, D., O’Brien, L., Uckert, K. & Steinberg, L. Peers increase adolescent risk taking by enhancing activity in the brain’s reward circuitry. Dev. Sci. 14, F1–F10 (2011).

9. Smith, A. R., Chein, J. & Steinberg, L. Peers increase adolescent risk taking even when the probabilities of negative outcomes are known. Dev. Psychol. 50, 1564 (2014).

10. Somerville, L. H. et al. Dissecting “Peer Presence” and “Decisions” to Deepen Understanding of Peer Influence on Adolescent Risky Choice. Child Dev. 90, 2086–2103 (2018).

11. Len-Rios, M. E., Hughes, H. E., McKee, L. G. & Young, H. N. Early adolescents as publics: a national survey of teens with social media accounts, their media use preferences, parental mediation, and perceived Internet literacy. Public Relat. Rev. 42, 101–108, https://doi.org/10.1016/j.pubrev.2015.10.003 (2016).

12. Mckay, H. D. Social influence on adolescent behavior. JAMA 182, 643-&, https://doi.org/10.1001/jama.1962.03050450043010 (1962).

13. Reyna, V. F. & Farley, F. Risk and rationality in adolescent decision making: implications for theory, practice, and public policy. Psychol. Sci. Public Interest 7, 1–44, https://doi.org/10.1111/j.1529-1006.2006.00026.x (2006).

14. Albert, D. & Steinberg, L. Inhibitory control and drug abuse prevention 211–226 (Springer, 2011).

15. Loke, A. Y. & Mak, Y.-w Family process and peer influences on substance use by adolescents. Int. J. Environ. Res. Public Health 10, 3868–3885 (2013).

16. Weitzman, E. R., Nelson, T. F. & Wechsler, H. Taking up binge drinking in college: the influences of person, social group, and environment. J. Adolesc. Health 32, 26–35 (2003).

17. Johnston, K. L. & White, K. M. Binge-drinking: a test of the role of group norms in the theory of planned behaviour. Psychol. Health 18, 63–77, https://doi.org/10.1080/0887044021000037835 (2003).

18. Lundborg, P. Having the wrong friends? Peer effects in adolescent substance use. J. Health Econ. 25, 214–233 (2006).

19. Tsakpinoglou, F. & Poulin, F. Best friends’ interactions and substance use: the role of friend pressure and unsupervised co-deviancy. J. Adolesc. 60, 74–82 (2017).

20. Sawyer, S. M. et al. Adolescence: a foundation for future health. Lancet 379, 1630–1640, https://doi.org/10.1016/S0140-6736(12)60072-5 (2012).

21. Romer, D. et al. Social influences on the sexual-behavior of youth at risk for hiv exposure. Am. J. Public Health 84, 977–985, https://doi.org/10.2105/Ajph.84.6.977 (1994).

22. Niederkrotenthaler, T. et al. Association of increased youth suicides in the United States with the release of 13 Reasons Why. JAMA Psychiatry Published online May 29, 2019. 10.1001/jamapsychiatry.2019.0922, https://doi.org/10.1001/jamapsychiatry.2019.0922 (2019).

23. Telzer, E. H., Van Hoorn, J., Rogers, C. R. & Do, K. T. Advances in child development and behavior Vol. 54 215–258 (Elsevier, 2018).

24. Braams, B. R., Davidow, J. Y. & Somerville, L. H. Developmental patterns of change in the influence of safe and risky peer choices on risky decision-making. Dev. Sci. 22, e12717, https://doi.org/10.1111/desc.12717 (2019).

25. van Hoorn, J., van Dijk, E., Meuwese, R., Rieffe, C. & Crone, E. A. Peer influence on prosocial behavior in adolescence. 26, 90–100 (2016).

26. Choukas-Bradley, S., Giletta, M., Cohen, G. L. & Prinstein, M. J. Peer influence, peer status, and prosocial behavior: an experimental investigation of peer socialization of adolescents’ intentions to volunteer. J. Youth Adolesc. 44, 2197–2210 (2015).

27. Foulkes, L., Leung, J. T., Fuhrmann, D., Knoll, L. J. & Blakemore, S. J. Age differences in the prosocial influence effect. Dev. Sci. 21, e12666, https://doi.org/10.1111/desc.12666 (2018).

28. van den Bos, W., Rodriguez, C. A., Schweitzer, J. B. & McClure, S. M. Adolescent impatience decreases with increased frontostriatal connectivity. Proc. Natl Acad. Sci. U.S.A. 112, E3765–E3774, https://doi.org/10.1073/pnas.1423095112 (2015).

29. Olson, E. A., Hooper, C. J., Collins, P. & Luciana, M. J. P. Adolescents’ performance on delay and probability discounting tasks: contributions of age, intelligence, executive functioning, and self-reported externalizing behavior. Pers. Individ. Dif. 43, 1886–1897 (2007).

30. Laube, C., Suleiman, A. B., Johnson, M., Dahl, R. E. & van den Bos, W. Dissociable effects of age and testosterone on adolescent impatience. Psychoneuroendocrinology 80, 162–169, https://doi.org/10.1016/j.psyneuen.2017.03.012 (2017).

31. Toelch, U. & Dolan, R. J. Informational and normative influences in conformity from a neurocomputational perspective. Trends Cogn. Sci. 19, 579–589, https://doi.org/10.1016/j.tics.2015.07.007 (2015).

32. Deutsch, M. & Gerard, H. B. A study of normative and informational social influences upon individual judgment. J. Abnorm. Soc. Psychol. 51, 629 (1955).

33. Somerville, L. H. The teenage brain sensitivity to social evaluation. Curr. Dir. Psychol. Sci. 22, 121–127 (2013).

34. Moutoussis, M., Dolan, R. J. & Dayan, P. How people use social information to find out what to want in the paradigmatic case of inter-temporal preferences. PloS Comput. Biol. 12, e1004965 (2016). ARTN e1004965 10.1371/journal.pcbi.1004965.

35. Revelt, D. & Train, K. Customer-specific taste parameters and mixed logit: Households’ choice of electricity supplier, No E00-274, Economics Working Papers, University of California at Berkeley, https://EconPapers.repec.org/RePEc:ucb:calbwp:e00-274 (UCB, 2000).

36. Garvert, M. M., Moutoussis, M., Kurth-Nelson, Z., Behrens, T. E. & Dolan, R. J. Learning-induced plasticity in medial prefrontal cortex predicts preference malleability. Neuron 85, 418–428 (2015).

37. Fields, R. D. A new mechanism of nervous system plasticity: activity-dependent myelination. Nat. Rev. Neurosci. 16, 756–767, https://doi.org/10.1038/nrn4023 (2015).

38. Vanes, L. D. et al. White matter tract myelin maturation and its association with general psychopathology in adolescence and early adulthood. Hum. Brain Mapp. 41, 827–839 (2020).

39. Ziegler, G. et al. Compulsivity and impulsivity traits linked to attenuated developmental fronto-striatal myelination trajectories. Nat. Neurosci. 22, 992 (2019).

40. Ziegler, G. et al. Childhood socio-economic disadvantage predicts reduced myelin growth across adolescence and young adulthood. Hum. Brain Mapp. 41, 3392–3402 (2019).

41. Ereira, S. et al. Social training reconfigures prediction errors to shape Self-Other boundaries. (under review).

42. Liu, J. et al. Impaired adult myelination in the prefrontal cortex of socially isolated mice. Nat. Neurosci. 15, 1621 (2012).

43. Bonnefil, V. et al. Region-specific myelin differences define behavioral consequences of chronic social defeat stress in mice. eLife 8, e40855 (2019).

44. Makinodan, M., Rosen, K. M., Ito, S. & Corfas, G. J. S. A critical period for social experience–dependent oligodendrocyte maturation and myelination. Science 337, 1357–1360 (2012).

45. Makinodan, M. et al. Effects of the mode of re-socialization after juvenile social isolation on medial prefrontal cortex myelination and function. Sci. Rep. 7, 1–9 (2017).

46. Wiesmann, C. G., Schreiber, J., Singer, T., Steinbeis, N. & Friederici, A. D. J. N. c. White matter maturation is associated with the emergence of Theory of Mind in early childhood. Nat. Commun. 8, 14692 (2017).

47. Nicolle, A. et al. An agent independent axis for executed and modeled choice in medial prefrontal cortex. Neuron 75, 1114–1121, https://doi.org/10.1016/j.neuron.2012.07.023 (2012).

48. Suzuki, S., Jensen, E. L., Bossaerts, P. & O’Doherty, J. P. Behavioral contagion during learning about another agent’s risk-preferences acts on the neural representation of decision-risk. Proc. Natl Acad. Sci. 113, 3755–3760 (2016).

49. Kievit, R. A. et al. Developmental cognitive neuroscience using latent change score models: a tutorial and applications. Dev. Cogn. Neurosci. 33, 99–117, https://doi.org/10.1016/j.dcn.2017.11.007 (2018).

50. van Harmelen, A. L. et al. Friendships and family support reduce subsequent depressive symptoms in at-risk adolescents. PLoS ONE 11, e0153715, https://doi.org/10.1371/journal.pone.0153715 (2016).

51. van Harmelen, A. L. et al. Adolescent friendships predict later resilient functioning across psychosocial domains in a healthy community cohort. Psychol. Med. 47, 2312–2322, https://doi.org/10.1017/S0033291717000836 (2017).

52. Prinstein, M. J., Boergers, J. & Spirito, A. Adolescents’ and their friends’ health-risk behavior: factors that alter or add to peer influence. J. Pediatr. Psychol. 26, 287–298, https://doi.org/10.1093/jpepsy/26.5.287 (2001).

53. Iacobucci, D., Saldanha, N. & Deng, X. Y. A meditation on mediation: evidence that structural equations models perform better than regressions. J. Consum. Psychol. 17, 139–153, https://doi.org/10.1016/S1057-7408(07)70020-7 (2007).

54. Maxwell, S. E. & Cole, D. A. Bias in cross-sectional analyses of longitudinal mediation. Psychol. Methods 12, 23 (2007).

55. Ciranka, S. K., & van den Bos, W. A Bayesian Model of Social Influence under Risk and Uncertainty. PsyArXiv. https://doi.org/10.31234/osf.io/mujek (2020).

56. Schmitt, J. E., Raznahan, A., Liu, S. & Neale, M. C. The genetics of cortical myelination in young adults and its relationships to cerebral surface area, cortical thickness, and intelligence: A magnetic resonance imaging study of twins and families. NeuroImage 206, 116319 (2020).

57. Grydeland, H., Walhovd, K. B., Tamnes, C. K., Westlye, L. T. & Fjell, A. M. Intracortical myelin links with performance variability across the human lifespan: results from T1-and T2-weighted MRI myelin mapping and diffusion tensor imaging. J. Neurosci. 33, 18618–18630 (2013).

58. Perkins, H. W., Craig, D. W. & Perkins, J. M. Using social norms to reduce bullying: a research intervention among adolescents in five middle schools. Group Process Interg. 14, 703–722, https://doi.org/10.1177/1368430210398004 (2011).

59. Paluck, E. L., Shepherd, H. & Aronow, P. M. Changing climates of conflict: a social network experiment in 56 schools. Proc. Natl Acad. Sci. U.S.A. 113, 566–571, https://doi.org/10.1073/pnas.1514483113 (2016).

60. Kiddle, B. et al. Cohort profile: the NSPN 2400 Cohort: a developmental sample supporting the Wellcome Trust NeuroScience in Psychiatry Network. Int. J. Epidemiol. 47, 18–19g (2017).

61. Master, S. L. et al. Distentangling the systems contributing to changes in learning during adolescence. Dev. Cogn. Neurosci. 41, 100732 (2020).

62. Crawley, D. et al. Modeling flexible behavior in childhood to adulthood shows age-dependent learning mechanisms and less optimal learning in autism in each age group. PLoS Biol. 18, e3000908 (2020).

63. Dezfouli, A., Griffiths, K., Ramos, F., Dayan, P. & Balleine, B. W. Models that learn how humans learn: the case of decision-making and its disorders. PLoS Comput. Biol. 15, e1006903 (2019).

64. Somerville, L. H. et al. Charting the expansion of strategic exploratory behavior during adolescence. J. Exp. Psychol.: General 146, 155 (2017).

65. Nesi, J., Rothenberg, W. A., Hussong, A. M. & Jackson, K. M. Friends’ alcohol-related social networking site activity predicts escalations in adolescent drinking: mediation by peer norms. J. Adolesc. Health 60, 641–647, https://doi.org/10.1016/j.jadohealth.2017.01.009 (2017).

66. Ayers, J. W., Althouse, B. M., Leas, E. C., Dredze, M. & Allem, J.-P. J. J. Internet searches for suicide following the release of 13 Reasons Why. JAMA Intern. Med. 177, 1527–1529 (2017).

67. Zhang, J. et al. Leveraging social influence to address overweight and obesity using agent-based models: the role of adolescent social networks. Soc. Sci. Med. 125, 203–213, https://doi.org/10.1016/j.socscimed.2014.05.049 (2015).

68. Fields, R. D. Myelination: an overlooked mechanism of synaptic plasticity? Neuroscientist 11, 528–531, https://doi.org/10.1177/1073858405282304 (2005).

69. Yang, Y. et al. Neonatal maternal separation impairs prefrontal cortical myelination and cognitive functions in rats through activation of Wnt signaling. Cereb. Cortex 27, 2871–2884 (2017).

70. Maas, D. A. et al. Interneuron hypomyelination is associated with cognitive inflexibility in a rat model of schizophrenia. Nat. Commun. 11, 1–16 (2020).

71. Kievit, R. & Simpson-Kent, I. L. It’s About Time: Towards a Longitudinal Cognitive Neuroscience of Intelligence. PsyArXiv. https://doi.org/10.31234/osf.io/n2yg7 (2019).

72. Collaboration, O. S. Estimating the reproducibility of psychological science. Science 349, aac4716 (2015).

73. Defoe, I. N., Dubas, J. S., Dalmaijer, E. S. & van Aken, M. A. Is the peer presence effect on heightened adolescent risky decision-making only present in males? J. Youth Adolesc. 49, 693–705 (2020).

74. Bolenz, F., Reiter, A. M. & Eppinger, B. Developmental changes in learning: computational mechanisms and social influences. Front. Psychol. 8, 2048 (2017).

75. Cole, D. A. et al. The development of multiple domains of child and adolescent self‐concept: a cohort sequential longitudinal design. Child Dev. 72, 1723–1746 (2001).

76. McClure, S. M., Ericson, K. M., Laibson, D. I., Loewenstein, G. & Cohen, J. D. Time discounting for primary rewards. J. Neurosci. 27, 5796–5804, https://doi.org/10.1523/JNEUROSCI4246-062007 (2007).

77. Pooseh, S., Bernhardt, N., Guevara, A., Huys, Q. J. & Smolka, M. N. Value-based decision-making battery: a Bayesian adaptive approach to assess impulsive and risky behavior. Behav. Res. Methods 50, 236–249 (2018).

78. Frith, C. & Frith, U. Theory of mind. Curr. Biol. 15, R644–R645 (2005).

79. Suzuki, S. et al. Learning to simulate others’ decisions. Neuron 74, 1125–1137 (2012).

80. Schmierer, K., Scaravilli, F., Altmann, D. R., Barker, G. J. & Miller, D. H. Magnetization transfer ratio and myelin in postmortem multiple sclerosis brain. Ann. Neurol. 56, 407–415, https://doi.org/10.1002/ana.20202 (2004).

81. Turati, L. et al. In vivo quantitative magnetization transfer imaging correlates with histology during de- and remyelination in cuprizone-treated mice. NMR Biomed. 28, 327–337, https://doi.org/10.1002/nbm.3253 (2015).

82. R Development CORE TEAM, R. (R foundation for statistical computing Vienna, Austria 2008).

83. R Core Team. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org/ (2020).

84. Singmann, H., Bolker, B., Westfall, J., Aust, F. & Ben-Shachar, M. S. afex: Analysis of Factorial Experiments. R package version 0.28-1 (2015).

85. Lenth, R. V. emmeans: Estimated Marginal Means, aka Least-Squares Means. R package version 1.6.0. https://CRAN.R-project.org/package=emmeans (2021).

86. Wood, S. & Wood, M. S. Package ‘mgcv’. R package version 1, 29 (2015).

87. Rosseel, Y. lavaan: An R Package for Structural Equation Modeling. Journal of Statistical Software, 48, 1–36 http://www.jstatsoft.org/v48/i02/ (2012).

88. Wickham, H. Ggplot2: elegant graphics for data analysis. (Springer, 2016).

## Acknowledgements

The Wellcome Trust funded the ‘Neuroscience in Psychiatry Project’ (NSPN). All NSPN members (Supplementary Table 4) are supported by a Wellcome Strategic Award (ref 095844/7/11/Z).  R. Dolan is supported by a Wellcome Investigator Award (ref 098362/Z/12/Z). The Max Planck – UCL Centre for Computational Psychiatry and Ageing is a joint initiative of the Max Planck Society and UCL. A.M.F. Reiter acknowledges support from the German Research Foundation (Deutsche  Forschungsgemeinschaft e, DFG RE 4449/1-1, SFB 940-3 project B7) and by a 2020 NARSAD Young Investigator Grant from the Brain & Behaviour Foundation. M. Moutoussis receives support from the NIHR UCLH Biomedical Research Centre.  R.A. Kievit is supported by a Sir Henry Wellcome Trust Grant 107392/Z/15/Z and MRC Programme Grant SUAG/014 RG91365. E. Bullmore is in receipt of an NIHR Senior Investigator Award. P. Fonagy is in receipt of a National Institute for Health Research (NIHR) Senior Investigator Award (NF-SI-0514-10157). P. Fonagy was in part supported by the NIHR Collaboration for Leadership in Applied Health Research and Care (CLAHRC) North Thames at Barts Health NHS Trust. The views expressed are those of the authors and not necessarily those of the NHS, the NIHR or the Department of Health. We are grateful to the NSPN management and research assistant teams and to all participants for supporting our research. We thank design@ukw.de for help with the design of the figures.

## Author information

Authors

### Contributions

A.M.F.R., M.M., L.V., R.K., E.T.B., I.M.G., P.F., P.B.J., R.J.D. and. NSPN consortium designed research, all authors performed research, A.M.F.R. and M.M. analysed the data, R.K. and L.V. contributed analytic tools, A.M.F.R. wrote the paper, M.M., L.V., R.K., E.T.B., I.M.G., P.F., P.B.J., R.J.D. provided significant revisions.

### Corresponding author

Correspondence to Andrea M. F. Reiter.

## Ethics declarations

### Competing interests

The authors declare no competing interests.

Peer review information Nature Communications thanks Martin Schulte-Ruether and the other, anonymous, reviewers for their contribution to the peer review of this work.

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

## Rights and permissions

Reprints and Permissions

Reiter, A.M.F., Moutoussis, M., Vanes, L. et al. Preference uncertainty accounts for developmental effects on susceptibility to peer influence in adolescence. Nat Commun 12, 3823 (2021). https://doi.org/10.1038/s41467-021-23671-2

• Accepted:

• Published:

• DOI: https://doi.org/10.1038/s41467-021-23671-2