Participants

The study and its main analyses were preregistered and can be retrieved from the Open Science Framework at https://osf.io/mz3d2/. We have conducted this work in accordance with the preregistration and explicitly indicate throughout the manuscript if we deviate from the preregistration. The sample consisted of early to late adolescent males, ages 13–23, with (n = 51) and without (n = 62) ADHD. Data from one additional participant was missing due to technical issues. Participants were primarily recruited through previous studies that targeted both clinical and typically developing samples (Dekkers et al., 2019; Ma et al., 2020), and additionally through a recruitment website.

Participants with ADHD were included if they met the following criteria: (a) previous (lifetime) ADHD diagnosis by a licensed psychologist or psychiatrist, and (b) a current ADHD classification (all presentations) based on a parent/caretaker interview with the ADHD section of the Diagnostic Interview Schedule (DISC-IV; Shaffer et al., 2000) conducted at their home. The DISC-IV was adjusted to reflect changes from DSM-IV to DSM-5 where necessary. A total of 12 participants with ADHD recruited from a previous study comparing youth with and without ADHD (Dekkers et al., 2019), were in partial remission for a current ADHD classification based on the DISC-IV. We maintained these participants in the current analyses.² Comorbid disorders were not assessed with the DISC-IV, but reported by parents.

Comorbid disorders were allowed and reported for n = 19 participants in the ADHD group (36% of the sample). Comorbid disorders included dyslexia (18%), dyscalculia (2%), oppositional defiant disorder (2%), depression (2%), and autism spectrum disorders, either as single comorbid diagnosis or co-occurring with additional comorbid disorders (14%; this included dyslexia, dyscalculia, developmental coordination disorder, dysthymia, attachment disorder, and obsessive-compulsive disorder). Note that percentages do not add up exactly due to rounding. Dyslexia was only allowed if parents reported that their child was able to read short sentences within 5 s.

The majority of the ADHD group (n = 30; 57%) was taking stimulant medication at the time of the study (53% methylphenidate; 4% dexamphetamines). Participants using methylphenidate discontinued medication 24 h before testing to reach total wash-out (Greenhill & Ford, 2002). For participants using dexamphetamine, the required wash-out period was 48 h (Wong & Stevens, 2012). Adolescents using atomoxetine, clonidine or antipsychotic medication were excluded. Participants without ADHD were included when parents reported no current psychiatric diagnoses for their child. Dyslexia/dyscalculia was allowed, and dyslexia occurred in n = 10 participants (16%).

Descriptive characteristics of the ADHD and typically developing (TD) group are presented in Table 1, for additional information see Supporting Information S1. We further describe our sample using parent report on the Child Behavior Checklist (CBCL; Achenbach et al., 2001) to assess emotional and behavioral problems, the Disruptive Behavior Disorder Rating Scale (DBDRS; Oosterlaan et al., 2000) to examine ADHD symptoms and potential comorbid behavioral disorders (ODD and CD), and Social Responsiveness Scale-2 (SRS-2; Roeyers Thys, Druart, De Schryver, & Schittekatte, 2011) to assess for symptoms in the autism spectrum. To test for possible confounding group differences, we also acquired estimated IQ scores for intelligence (see Supporting Information S1). The estimated IQ scores fell within the average to above average range for all participants and did not differ between the ADHD group and TD group.

Measures

Peers delay discounting task

We adapted a standard delay discounting paradigm in which the amount of the immediate reward and delay duration are varied in order to obtain a discounting function (de Water et al., 2014). The discounting task involved 24 choices between a smaller reward that was available immediately, and a larger reward that was available after a delay. The amount of the immediate reward was varied based on participants' choices (Du et al., 2002), and varied between €0.15 and €9.85. The delayed reward was fixed at €10. The delay preceding the larger reward varied between 2, 14, 30 and 90 days. Six unique choices were presented for each of the four delays. For example, the first choice was always between €5 today and €10 after a delay (e.g. €5 today or €10 after 2 days). If the participant chose the immediate reward, its amount would be decreased by half on the next trial (e.g., €7.50 today or €10 in 2 days). If the delayed reward was chosen, the value of the immediate reward would be decreased by half on the next trial (e.g., €2.50 today or €10 in 2 days). The subjective value of the delayed reward for each delay was defined as the hypothetical value of the immediate reward on a seventh trial at that delay (Du et al., 2002). These subjective values were used to compute the area under the curve (AUC), which ranges between 0 and 1, in which smaller values indicate an increased preference for immediate rewards (Myerson et al., 2001).

The task can be designated as potentially real, because participants were informed that one round would be selected as a reward at the end of the task, and each choice was therefore assumed to have been made as if it had real-life consequences (Scheres, de Water et al., 2013). To make the rewards more explicit, participants were presented with real money on the table during the task: one €0.50 coin, one €1 coin, one €2 coin, one €5, and one €10 bill. The €10 bill was related to the largest amount of money that was included in the choices in the task.

The delay discounting task consisted of three fixed-order within-subject conditions: alone, with alleged peer feedback and alone again (cf. Van Hoorn et al., 2016). We utilized a between-subjects design in which we compared two types of feedback: impulsive peer feedback and non-impulsive peer feedback (Figure 1A). Peer feedback consisted of images of thumbs up or ‘likes’ of five fictitious peers. The thumbs up were presented on the screen (in fixed order, the same for each participant) after the participants made their decision. In the impulsive peer feedback condition, the majority of peers liked the impulsive option (i.e., 3, 4, or 5 fictitious peers gave a thumbs up, averaging to 75% of the five thumbs up across trials, for the smaller sooner reward). The task was programmed such that following the majority of impulsive peers' likes corresponded with an AUC of 0.018 (impulsive). In the non-impulsive peer feedback condition, the majority of peers liked the non-impulsive option, that is, the larger, delayed reward (see Figure 1B). Following the majority of peers' likes in the non-impulsive condition corresponded with an AUC of 0.993 (patient).

Procedure

This task was administered as part of a larger study that was approved by the medical ethical committee of the Leiden University Medical and the ethical committee of the Institute of Psychology at Leiden University. Parents signed an informed consent form prior to completing the online parent questionnaires, and a separate informed consent to participate in the DISC-interview (ADHD group only). When at the lab, all participants and their parents (in case of minors) signed an informed consent form prior to the start of the study. All participants were tested one-on-one by trained experimenters, who could provide help when necessary. Following the task instructions, including six practice trials, participants were informed that the computer would randomly pick one trial that would be selected for pay-out. In fact, all participants received €2.50 as bonus compensation for this task, €40 for participating in the larger study, as well as a goodie bag that contained small gifts. Parents received €10 to complete a set of online questionnaires, and €10 for participating in the DISC-interview (the latter for the ADHD group only). The task was presented on a laptop using E-prime (version 3), had self-paced timing and lasted about 10 min on average.

Statistical analyses

The dependent variable was the AUC in each of the three within-subjects conditions: alone, with alleged peer feedback and alone again (cf. Van Hoorn, Van Dijk, et al., 2016). Our main preregistered analysis used a mixed ANOVA testing the between-subjects effects of Feedback Type (Impulsive/Non-impulsive) and Group (ADHD/TD) on the change in individual's AUC towards peer feedback and when playing alone again (Alone 1/with feedback/Alone 2), with repeated measures of the last factor and age as a covariate. Since previous work also emphasizes a more dimensional approach to ADHD (Coghill & Sonuga-Barke, 2012; Nikolas & Burt, 2010), we exploratively test a more continuous relationship between inattention and hyperactivity–impulsivity symptoms and susceptibility to social influence. We conducted exploratory regression analyses with the difference score AUC in Alone 2—AUC in Alone 1 as the dependent variable, separately for each Feedback condition (impulsive/non-impulsive). Age and scores on DBDRS Inattention and Hyperactivity–Impulsivity (mean-centered) were included as independent variables in these analyses.³

Non-impulsive peer feedback decreases impulsivity

Results from the mixed ANOVA indicated no main effects of Feedback type (p = .663) or Condition (p = .109), but did reveal an interaction of Feedback type by Condition, (F(2,216) = 14.610, p < .001, partial η² = 0.119). There was no main effect nor an interaction effect of Group (ps > .2), indicating that the effects of peer feedback on impulsive choice were similar for adolescents with and without ADHD. Finally, there was no effect of the covariate Age (F(1,108) = 1.812, p = .181). Means for the area under the curve (AUC) as a measure for impulsive choice in each condition are displayed in Figure 2.

Bonferroni-adjusted post-hoc tests were used to unpack the interaction between Feedback type and Condition. In the impulsive peer feedback condition, participants' decision-making was similar when they were alone, during peer feedback, and alone again (ps > .2), suggesting no effects of impulsive peer feedback. In contrast, results yield a significant decrease in impulsive choice in the non-impulsive peer feedback condition, such that the AUC in Alone 1 < Peer feedback (p < .001), Peer Feedback < Alone 2 (p = .002), and Alone 1 < Alone 2 (p < .001; see Figure 2). Hence, the effects of peer feedback remain and appear to become even stronger when adolescents go back to making decisions alone.

Inattention and hyperactivity–impulsivity relate to peer feedback susceptibility

Given the differential effects on peer influence across feedback condition, exploratory regression analyses were run separately for the Impulsive and Non-impulsive feedback conditions. In the Impulsive feedback condition, susceptibility to peer feedback was not predicted by individual differences in hyperactivity or inattentive symptoms (p = .893). In the Non-impulsive feedback condition, the regression model was significant (F(3,52) = 5.925, R²_Adj = 0.221, p = .002). Inattention symptoms were a positive predictor, (β = .453, t = 2.50, p = .016), while hyperactivity–impulsivity (β = −.712, t = −3.80, p < .001) and age (β = −.365, t = −2.83, p = .007) were negative predictors. This demonstrates that in the total sample, more inattention symptoms related to greater susceptibility to non-impulsive peer feedback (i.e., becoming less impulsive), whereas being older and having more hyperactivity-impulsivity is associated with reduced susceptibility (see Figure 3).

Peer feedback decreases impulsive choice in adolescents with and without ADHD

A large body of research indicates that the peer context plays a major role in shaping adolescent decision-making (Andrews et al., 2021). More specifically for impulsivity, peer influence studies using the delay discounting task have revealed that presence or observation of impulsive peers increase impulsive choice in adolescents (Gilman et al., 2014; O’Brien et al., 2011; Weigard et al., 2014). Here, we observed that non-impulsive peer feedback decreased impulsive choice, although impulsive peer-feedback did not increase impulsive choice. While the direction of our effects appears in contrast with earlier work, this inconsistency can be explained using social learning theory (Akers, 2011; Bandura, 2001). An important element in this theory is that peer influence works through internalizing social norms from the peer context. Yet, the peer influence process is active; adolescents do not just passively absorb the social norms from peers around them (Brechwald & Prinstein, 2011). This is apparent in our data since adolescents did not follow their peers' social norms in the impulsive peer feedback condition. Our participants' own norms (i.e., baseline delay discounting decisions) were quite different from the social norms relayed in the impulsive peer feedback condition, but relatively similar to peers' non-impulsive social norms. In sum, our findings align with the idea that adolescents appear to be influenced by peers when social norms are relatively closely aligned with their own norms (Ciranka & van den Bos, 2019), and they do not blindly follow their peers' feedback.

We also observed a transfer of non-impulsive peer feedback to subsequent individual decisions. Similar to previous results in the domain of prosocial behavior (Van Hoorn, Van Dijk, Meuwese et al., 2016), this carry-over effect suggests that the social norms provided by the peer group are to some extent maintained over time. Interestingly, adolescents preferred larger delayed rewards even more in the subsequent alone round than during peer feedback. Speculatively, this may suggest that the social norms were not only learned but perhaps also internalized (Staub, 1972), due to confirmation in their preference for larger delays. Future studies could include a follow-up after a few months to assess whether peer feedback continues to guide impulsive decisions over time, or employ a longitudinal design to get a better understanding of causality.

On a group level, we found that adolescents with and without ADHD were equally susceptible to peer influence from a group of five peers. Consistent with social learning theory (Akers, 2011; Bandura, 2001) and accounts of differential reward sensitivity (Sonuga-Barke, 2002, 2003) we had expected to find that adolescents with ADHD would be more susceptible to peer influence. Alternatively, one might hypothesize that ADHD-related executive function impairments, as well as positive illusory biases would yield less sensitivity to peer feedback. That is, adolescents with ADHD would be less likely to inhibit pre-potent responses and consequently respond before considering contingencies and feedback from peers; are less likely to evaluate and consider possible contingencies and feedback from peers due to impaired working-memory (Kofler et al., 2011; Patros et al., 2015), and more likely to overestimate their performance and feedback from peers due to positive illusory biases (Owens et al., 2007). While neither of these hypotheses were supported by the data, our findings resonate with a related study that also found no differences in susceptibility related to ADHD status when a single peer encouraged risk-taking behavior (Dekkers et al., 2020). Taken together, the findings suggest that in principle adolescents with ADHD appear equally susceptible to peer influence, either from a single peer or a small group of peers. Note however that adolescents with ADHD affiliate with deviant peers more often (Nijmeijer et al., 2008), and therefore it has been suggested that they are less likely than typically developing peers to experience positive peer influence that could help them make less impulsive choices (Dekkers et al., 2020).

Inattention and hyperactivity–impulsivity symptoms associated with susceptibility to peer feedback

While our main analysis did not show any group-level differences related to an ADHD diagnosis, our exploratory findings revealed that variability in inattention, hyperactivity–impulsivity (Coghill & Sonuga-Barke, 2012; Nikolas & Burt, 2010) and age were associated with adolescents' susceptibility to non-impulsive peer feedback. Specifically, higher inattention was associated with greater susceptibility, whereas more hyperactivity–impulsivity symptoms and being older was related to reduced susceptibility. The age effect resonates with previous work that illustrates how the impact of social influence slowly decreases across age (Andrews et al., 2021; Knoll et al., 2015). To our knowledge, the current work is the first to examine how inattention and hyperactivity–impulsivity symptoms are associated with susceptibility to peer influence, especially related to the opportunities of non-impulsive peer influence (but see Scheres et al., 2010 for a similar analytical approach with a delay discounting task).

Some studies have examined social difficulties associated with the behavioral phenotypes of each ADHD presentation (i.e., predominantly inattentive, hyperactive-impulsive, or combined). The nature of social difficulties tends to vary between the different ADHD presentations, but it is clear that all presentations experience impairments in social decision-making (Humphreys et al., 2016; Maedgen & Carlson, 2000), leading to difficulties with peer relations (Bagwell et al., 2001). On a symptom-level, our results show that more inattention is related to greater susceptibility to peer influence. Consistent with previous work, this may fit with a greater lack of assertiveness (Solanto et al., 2009) or as a means to be accepted in a group and gain status (Brechwald & Prinstein, 2011) in individuals with the predominantly inattentive presentation of ADHD. On the other hand, greater hyperactivity-impulsivity symptoms are related to lower susceptibility to non-impulsive peer influence, perhaps due to lower (social) self-control or working memory (Kofler et al., 2011; Patros et al., 2015; Solanto et al., 2009), which would leave less room for peer feedback to affect decision-making.

Together, these exploratory findings may provide an explanation for the absence of group-level (interaction) effects of ADHD in our main analyses on the effects of non-impulsive feedback, given that inattention and hyperactivity–impulsivity symptoms had effects in the opposite direction. Moreover, ADHD is a very heterogeneous disorder, such that many different combinations of inattention and hyperactivity–impulsivity symptoms can lead to the diagnosis in all its presentations (Faraone et al., 2021). A continuous approach combined with disentangling contributions of different symptoms may be informative to investigate complex social processes such as peer influence. It is important to emphasize that the differential relations of ADHD symptoms with peer influence were only observed when adolescents were presented with non-impulsive peer norms. Future research should further examine whether impulsive peer-influence relates to inattention and hyperactivity–impulsivity symptoms. Potentially, when impulsive peer-influence aligns more closely to an individual's own level of impulsivity, peer influence may be more symmetrical, swaying others towards more impulsive choice behavior (Ciranka & van den Bos, 2019). Moreover, it is an open question how our findings translate to other relevant domains of behavior in adolescents with ADHD such as risk-taking and (pro)social behavior. Gaining more insights into susceptibility to peer influence in a wide range of domains may help identifying those adolescents who may particularly profit from positive peer influence, or those who are at risk for negative peer influence.

Limitations and future directions

A few limitations of this study should be noted. First, we included a peer manipulation with optimal experimental control (i.e., peer feedback from five unknown peers). While the ages of these peers were not revealed to the participants, the peers' ages were not always perfectly matched, especially for our older participants. It is also unclear whether the effects would be similar if feedback came from adolescents' actual friends. Social identity theory (Islam, 2014) predicts that the effects would likely be stronger, given that adolescents more strongly identify with their friends than unknown peers. In that light, the current study represents a conservative test of the power of peers, which already showed promising findings.

Second, we did not examine the effects of our peer manipulation on (impulsive) behavior outside the experimental task. However, previous work has already shown that delay discounting task manipulations are related to health-risk behaviors (e.g., smoking, consuming unhealthy foods; Amlung et al., 2016; Audrain-McGovern et al., 2009; Jackson & MacKillop, 2016). It would be a valuable addition for future research to link the current experimental findings to real-world impulsive or risk behaviors.

The third limitation that can be noted is that we only included males in the current study. Although ADHD is more prevalent in males (6%) than females (3%) before the age of 18 (Dalsgaard et al., 2020), future research could include females and examine whether they similarly benefit from non-impulsive peer feedback. Given that girls generally show lower levels of hyperactivity–impulsivity (Gershon & Gershon, 2002), one hypothesis based on the current work could be that they are more susceptible to non-impulsive peer influences. Future work employing a between-subjects design should include a larger sample size for optimal power, including boys and girls, to replicate the current findings.

Finally, steep discounting of delayed rewards in delay discounting tasks has been described as a trans-diagnostic mechanism (Amlung et al., 2019), as it is not only characteristic of adolescents with ADHD (Jackson & MacKillop, 2016; Marx et al., 2021; Patros et al., 2016), but also frequently observed in adolescents with disorders that are highly comorbid with ADHD, such as substance abuse (Reynolds & Fields, 2012) and conduct disorder (White et al., 2014). While examining the effects of comorbidity was outside of the scope of the current study, a broad clinical assessment targeting comorbidity would be an interesting avenue for future research, as interventions targeting delay discounting could potentially reduce not only ADHD symptoms, but also symptoms of comorbid disorders.

Conclusions and implications for clinical practice

In conclusion, our study provides empirical support for the opportunities of positive peer feedback to decrease impulsive choice in adolescents with and without ADHD. Moreover, we exploratively showed that across our sample of adolescents there is substantial variability in susceptibility to peer feedback as a function of both hyperactivity-impulsivity and inattention. Our findings suggest that it would be relevant to include a social component in an intervention to reduce impulsive choice in youth with ADHD (e.g., peer role models or peer feedback on behavior). This is in line with a recent review which indicates that behavioral peer interventions for ADHD are well-established psycho-social interventions (Evans et al., 2018). Interestingly, these programs mostly rely on trained staff or parents to facilitate successful behavior. Our findings suggest that involving (virtual) peer processes could be utilized in a similar way to decrease impulsive behaviors in adolescents with ADHD.