social influence research paper example

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here .

Loading metrics

Open Access

Peer-reviewed

Research Article

The effects of information and social conformity on opinion change

Roles Conceptualization, Data curation, Formal analysis, Funding acquisition, Methodology, Project administration, Supervision, Validation, Visualization, Writing – original draft, Writing – review & editing

* E-mail: [email protected]

Affiliation School of Public Affairs, The Pennsylvania State University - Harrisburg, Middletown, Pennsylvania, United States of America

Roles Conceptualization, Data curation, Methodology, Supervision, Validation, Writing – original draft, Writing – review & editing

Affiliations Department of Biochemistry, The Pennsylvania State University, University Park, Pennsylvania, United States of America, Department of Molecular Biology, The Pennsylvania State University, University Park, Pennsylvania, United States of America, Department of Political Science, The Pennsylvania State University, University Park, Pennsylvania, United States of America

• Daniel J. Mallinson, 
• Peter K. Hatemi

• Published: May 2, 2018
• https://doi.org/10.1371/journal.pone.0196600
• Reader Comments

18 Mar 2020: Mallinson DJ, Hatemi PK (2020) Correction: The effects of information and social conformity on opinion change. PLOS ONE 15(3): e0230728. https://doi.org/10.1371/journal.pone.0230728 View correction

Extant research shows that social pressures influence acts of political participation, such as turning out to vote. However, we know less about how conformity pressures affect one’s deeply held political values and opinions. Using a discussion-based experiment, we untangle the unique and combined effects of information and social pressure on a political opinion that is highly salient, politically charged, and part of one’s identity. We find that while information plays a role in changing a person’s opinion, the social delivery of that information has the greatest effect. Thirty three percent of individuals in our treatment condition change their opinion due to the social delivery of information, while ten percent respond only to social pressure and ten percent respond only to information. Participants that change their opinion due to social pressure in our experiment are more conservative politically, conscientious, and neurotic than those that did not.

Citation: Mallinson DJ, Hatemi PK (2018) The effects of information and social conformity on opinion change. PLoS ONE 13(5): e0196600. https://doi.org/10.1371/journal.pone.0196600

Editor: Yong Deng, Southwest University, CHINA

Received: August 17, 2017; Accepted: April 16, 2018; Published: May 2, 2018

Copyright: © 2018 Mallinson, Hatemi. This is an open access article distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Data Availability: Data are available from the corresponding author’s Harvard Dataverse ( http://dx.doi.org/10.7910/DVN/YVCPDT ).

Funding: This project was supported by a $1,000 internal grant from the Penn State Department of Political Science (awarded to DJM). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Competing interests: The authors have declared that no competing interests exist.

Introduction

Information and persuasion are perhaps the most important drivers of opinion and behavioral changes. Far less attention, however, has been given to the role of social pressure in opinion change on politically-charged topics. This lacuna is important because humans have a demonstrated proclivity to conform to their peers when faced with social pressure. Be it in the boardroom or on Facebook, Solomon Asch and Muzafer Sherif’s classic studies hold true today. Individuals conform based on a desire to be liked by others, which Asch [ 1 , 2 ] called compliance (i.e., going along with the majority even if you do not accept their beliefs because you want to be accepted), or a desire to be right, which Sherif et al. [ 3 ] termed private acceptance (i.e., believing that the opinions of others may be more correct or informed than their own). These two broad schemas encompass many specific mechanisms, including, motivated reasoning, cognitive dissonance, utility maximization, conflict avoidance, and pursuit of positive relationships, among others. Information-based social influence and normative social influence (i.e., conformity pressure) both play important, albeit distinct, roles in the theories of compliance and private acceptance (see [ 4 ]). In both cases, humans exhibit conformity behavior; however only in private acceptance do they actually update their beliefs due to the social delivery of new information.

Extensions of Asch and Sherif’s path-breaking works have been widely applied across a number of behavioral domains [ 5 – 9 ], to include political participation. For example, significant attention has been focused on the import of conformity on voter turnout and participatory behaviors [ 10 ], including the effects of social pressure on the electoral behavior of ordinary citizens [ 11 – 15 ]. This body of work points to both the subtle and overt power of social influence on electoral behavior, yet little is known about the import of social conformity for politically charged topics in context-laden circumstances, particularly those that challenge one’s values and opinions.

Testing conformity pressure in the ideological and political identity domain may explicate whether the pressure to align with an otherwise unified group is different when dealing with politically charged topics versus context-free topics such as the size of a line or the movement of a ball of light [ 2 , 16 ]. Opinions on politically charged topics are complex, value laden, aligned with cultural norms, and not easily changed [ 17 – 21 ]. It remains unknown if the effects of social conformity pressures on political opinions are conditioned by the personal nature of the locus of pressure. To be sure, social conformity is a difficult concept to measure without live interaction. An observational approach makes it difficult to untangle if or how social pressure independently affects behaviors given these variegated casual mechanisms, and whether changes in opinion that result from social interaction are due to compliance or private acceptance. Nevertheless, experiments provide one means to gain insight into how and why opinion change occurs. Here, we undertake an experiment to test the extent to which opinion change is due to persuasion through new information, social conformity pressure, or a combination of the two in a more realistic extended discussion environment.

Conformity and political behavior

Both observational and experimental research has addressed different aspects of the impact of socially-delivered information on individual behavior. Observational analyses of social networks form the backbone of much of the recent research on social influence and political behavior. Sinclair [ 22 ], for instance, demonstrates that citizen networks convey a bounded set of political information. Individuals may turn to highly informed peers [ 23 ] or aggregate information from trusted friends and family [ 24 ] in order to reduce the cost of gathering the information required to engage in political behavior (e.g., voting). In turning to their network, they are open to privately accepting this useful information. Political information, however, is not the only type of information transmitted through personal networks. Social pressure helps the network induce compliance with desired social norms [ 25 – 27 ]. In this case, members of the network provide information regarding the group’s expectations for appropriate engagement in politics. Individuals that are concerned about whether or not the group will continue to accept them therefore conform out of a desire to be liked, broadly defined. Norms are often self-enforcing, with merely the perceived threat of potential sanctions being enough to regulate behavior through compliance and self-sanctioning [ 28 , 29 ].

The debate over the practicality and reality of deliberative democracy further highlights the importance of understanding the role of political conformity in public and elite discourse. Scholars and theorists argue that political decisions are improved and legitimized under a deliberative process [ 30 – 34 ], even though deliberation does not necessarily result in consensus [ 35 ]. The crux of democratic deliberation is that participants are engaging in a rational discussion of a political topic, which provides the opportunity for each to learn from the others and thus privately update their preferences (i.e., out of a desire to be right). It results in a collectively rational enterprise that allows groups to overcome the bounded rationality of individuals that would otherwise yield suboptimal decisions [ 36 ]. This requires participants to fully engage and freely share the information that they have with the group.

Hibbing and Theiss-Morse [ 37 ], however, raise important questions about the desirability of deliberation among the public. Using focus groups, they find that citizens more often than not wish to disengage from discussion when they face opposition to their opinions. Instead, they appear averse to participation in politics and instead desire a “stealth democracy,” whereby democratic procedures exist, but are not always visible. In this view, deliberative environments do not ensure the optimal outcome, and can even result in suboptimal outcomes. In fact, the authors point directly to the issue of intra-group conformity due to compliance as a culprit for this phenomenon. The coercive influence of social pressure during deliberation has been further identified in jury deliberations [ 38 , 39 ] and other small group settings [ 40 ].

Beyond politics, there is experimental evidence of the propensity to conform out of a desire to either be liked or to be right [ 25 , 41 – 45 ]. Using a simple focus group format and pictures of lines, Asch [ 1 , 2 ] demonstrated that individuals would comply with the beliefs of their peers due to a desire to be accepted by the group, even if they disagree and even when they believe the group opinion does not match reality. To do this, Asch asked eight members of a group to evaluate two sets of lines. The lines were clearly either identical or different and group members were asked to identify whether there was a difference. Unknown to the participant, the seven other group members were confederates trained to act in concert. At a given point in the study, the confederates began choosing the wrong answer to the question of whether the lines were equal. Consequently, the participant faced social pressure from a unified group every time they selected their answer. Asch varied the behavior of the group, including the number of members and number of dissenting confederates. Participants often exhibited stress and many eventually complied with the group consensus, even though the group was objectively wrong and participants did not agree with them privately.

Using a much more complex and context-laden format—a youth summer camp with real campers—Sherif et al. [ 3 ] demonstrated private acceptance whereby humans internalize and conform to group norms because consensus suggests that they may have converged on a right answer. In this case, the boys in the camp quickly coalesced into competing factions and initial outliers in the groups conformed out of a desire to win competitions (i.e., be right). While the groundbreaking Robbers Cave experiments revealed a great deal about group behavior well beyond conformity, we focus specifically on this particular aspect of the findings, which have stood the test of time in numerous replications and extensions across a wide variety of social domains [ 46 – 52 ].

Replication of Asch’s experimental work, in particular, has met varying degrees success. Lalancette and Standing [ 53 ]found that Asch’s results were mixed when using a prompt more ambiguous than unequal lines. Further, Hock [ 54 ] critiques the Asch design for not replicating a real life situation. Focusing on divorce attitudes, Kenneth Hardy provided an early application of Asch’s public compliance and Sherif’s private acceptance theories to political opinions using a similar small-group format with six confederates and one participant. Confederates offered not only their opinions, but also reasons for their opinions, which provided a methodological innovation by introducing more information than just the confederates’ votes. Hardy’s work provided an important starting point for identifying the process of conformity in the political realm, but it remains limited. He only utilized men in his study and did not allow for repeated discussion to assess how long participants hold up to conformity pressure. In a more recent study, Levitan and Verhulst [ 55 ]found persistence in political attitude change after interaction with a unanimously-opposing group, but they did not incorporate any discussion.

Our experiment builds on these works by examining the micro-process underlying opinion change for a politically charged topic discussed in a real context. We bridge between studies that allow for no discussion with those that study day-long deliberations in order to determine if group influence has a stronger effect, even when the discussion centers on an attitude closely tied with social identity. Our interaction of about an hour simulates a likely real-world example of dialogue. More importantly, our design allows us to speak to the debate over social influence by pulling apart the desires to be right (private acceptance) and liked (compliance). Our main goal is not to completely predict the general public’s behavior, but rather to identify the independent causal role of social pressure on opinion change, given the known import of information effects. We expect conformity pressure and information to have joint and independent effects on opinion change.

Variation in conformity behavior

While our primary interest is in identifying the average effects of information and conformity pressure on opinion change, we nevertheless recognize that there is variation in humans’ responses to social pressure, depending on observed and unobserved individual characteristics. Thus the average treatment effect recovered can mask substantively important heterogeneity [ 56 , 57 ]. For instance, not all of Asch’s or Hardy’s subjects complied with group opinion and there was a great deal of variation in how willing Sherif et al.’s campers coalesced into cohesive and functioning groups. In order to address this possibility we test three factors that have been previously identified as covarying with the average propensity to conform: personality traits, self-esteem, and ideology. The most consistent evidence points towards those who change their opinions as being generally more agreeable, neurotic, and having lower self-esteem [ 58 ].

Generating hypotheses regarding the import of other personality and ideological dispositions on opinion change for political, moral and identity-laden topics is more complicated. Extant research indicates support for both stability and change for these traits and differs in the source of that change, i.e., whether it is informational or social. For example, on the one hand we might expect those who are more politically conservative to be more likely to conform to the group overtly, given extant studies showing conservatives think less negatively toward conformity and comply more often to group pressure and norms [ 59 – 61 ]. In addition, conservatives are also higher on the Conscientiousness personality trait, and this trait both reflects and is related to more conformist behavior [ 62 – 64 ].

On the other hand, conservatism, by definition, advocates the status quo and is related to resistance to change and greater refusal to privately accept new information, specifically if that information contradicts one’s values [ 65 , 66 ], leading to a greater likelihood of internal stability. In a similar manner, those high in openness and more politically liberal, while more likely to take in new information, and thus possibly more likely to privately accept it, are also less prone to restrictive conformity, and thus possibly less likely to conform publicly [ 67 ]. We treat these propositions as secondary hypotheses, and explore their import in a limited manner given restrictions in the data.

Materials and methods

In order to explicate the independent and joint effects of compliance and private acceptance, we designed an experiment, conducted at the Pennsylvania State University from May to December of 2013, which placed participants in a deliberative environment where they faced unified opposition to their expressed opinion on a political topic that is relevant to their local community. We assessed participants’ privately-held opinions, absent the group, before and after the treatment in order to determine whether those who expressed a change in opinion during the discussion only did so verbally in order to comply with the group and gain acceptance or if they privately accepted the group’s opinion and truly updated their own values. The group discussed the topic openly, for approximately 30–45 minutes, also allowing us to assess participant behavior throughout the discussion. We discuss the specifics in more detail below.

In designing the experiment, we leveraged a unique time in Penn State’s history, the aftermath of the Jerry Sandusky child abuse scandal and the firing of longtime Head Coach Joe Paterno. The firing provided an ideal topic of discussion and a hard test of conformity pressure given the fact that it exhibited high salience on campus, was politically charged, and connected to the participants’ identities as Penn State students. The question posed to our participants was whether or not they felt that Coach Paterno should have been fired by the University’s Board of Trustees in November 2011. Previous research demonstrates that undergraduates may not have as clearly defined political attitudes as older adults on many topics and thus may be more susceptible to conformity pressure from peers due to non-attitudes [ 68 ]. This informed our choice of the discussion topic, as Paterno’s role in the abuse was not only highly salient on the Penn State campus, but typically invoked strong and diametrically opposed opinions in the undergraduate population and the general Penn state community. We begin by providing some background on this issue and its connection to identity and politics.

Firing of Penn State football Head Coach Joe Paterno

The first week of November 2011 was a whirlwind for students at Penn State. Police arrested former defensive coach Jerry Sandusky on charges of child sexual abuse following the release of a grand jury report by the Pennsylvania Office of the Attorney General. In the midst of a national media firestorm and with evidence mounting that the University President, Athletic Director and Head football Coach had been aware of Sandusky’s activities, Penn State President Graham Spanier resigned and the Board of Trustees relieved Paterno of his duties. They also placed the Athletic Director, Tim Curley, and Vice President, Gary Schultz, on administrative leave after being indicted for perjury regarding their testimony about their knowledge of Sandusky’s sexual assaults of young boys. Immediately after the firings and suspensions, students poured into campus and downtown State College, causing damage and flipping a news van [ 69 ]. Various student protests persisted for weeks. The following summer brought Sandusky’s conviction, but controversy has not subsided, especially in Pennsylvania. The firing is continually alive at Penn State, as lawsuits against the university and the trials of Spanier, Curley, and Shultz continue to progress as Paterno’s family and supporters seek to restore his legacy.

While the real-life context of our design adds to its external validity, the discussion topic’s high salience and likelihood of evoking a strong opinion also improves the internal validity of the experiment. Paterno was more than an employee; he was the image of Penn State, “an extension of [the students’ and alumni’s] collective self” ([ 70 ], 154), and thus tied to students’ identities as members of the community [ 71 ]. As reported at the time of the scandal:

“More than any other man, Mr. Paterno is Penn State–the man who brought the institution national recognition… Paterno is at the core of the university’s sense of identity.” [ 72 ].

Given the emotion surrounding this issue, it is not unlike morality policies that evoke strong responses from individuals [ 73 ], thereby providing a hard test of conformity pressure on value- and identity-laden opinions. There is no better example of this than the ongoing pursuit of justice by the children subjected to abuse by Catholic priests and the mounting evidence of systematic concealment and enablement of such abuse by the Catholic Church. The similarities between Penn State and the Church persist on nearly every level, including the scandals threatening an important aspect of its members’ identities. In this way, the experience of students following the child abuse scandal at Penn State generalizes to politically relevant circumstances where organizational power and personal identities are challenged.

In addition to being a highly salient and identity-laden topic of discussion, the Paterno firing is a social and political issue. It weighed heavily on the 2012 Board of Trustees elections, when many candidates campaigned on their support for Paterno. Furthermore, Pennsylvania Governor Tom Corbett was a de facto member of the Board and originally launched the Sandusky investigation while serving as the state Attorney General. As a board member, Corbett advocated for Paterno’s firing and faced both praise and criticism across the Commonwealth. As a result of the scandal, Pennsylvania passed legislation that clarifies responsibilities for reporting child abuse and heightens penalties for failures to report. The abuse received national recognition. When asked for his reaction to the firing, President Obama called on Americans to search their souls and to take responsibility for protecting children [ 74 ]. Thus, there is recognition by elites, the public, the media, and the academy that Paterno’s firing is an inherently political issue. Furthermore, the topic has personal importance to the participants, is identity laden, and relevant at the local, state, and national-levels. Having described the context of the topic of discussion, we now turn to describing the experimental protocol.

Participant recruitment

The experiment was advertised as a study on political discussion in upper- and lower-level social science courses, as well as through campus fliers and a university research website. As an incentive, participants were entered into a raffle for one of eight $25 gift cards to Amazon. The first participants completed the study in May 2013 and data collection closed in December 2013. There were no major developments in the Sandusky scandal during our data collection phase, thus we believe that no outside events threaten the validity of the study. The firing of the four university officials, Joe Paterno’s death, Jerry Sandusky’s conviction, issuance of the Freeh Report, and the National Collegiate Athletic Association’s sanctions all occurred prior to the start of data collection. This study was approved by the Pennsylvania State University Office for Research Protections Institutional Review Board (Study# 41536) on February 20, 2013. All participants in the treatment group signed a written informed consent form prior to participating in the study. Participants in the control group supplied implied consent by completing the online survey after reading an informed consent document on the first web page of the survey. Penn State’s IRB approved both methods of consent. Consent materials can be found with other study reproduction materials at the corresponding author’s dataverse ( http://dx.doi.org/10.7910/DVN/YVCPDT ). Thus, all participants provided informed consent and all procedures contributing to this work complied with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975.

A total of 58 students participated in either the treatment or control groups. Compared to observational studies, this may appear a small number, but it comports with current research norms that require high participant involvement and a substantial amount of their time [ 75 , 76 ] and is consistent with the sample sizes for the foundational work in this area [ 2 , 6 ]. The pre- and post-test, discussion session and debriefing required at least 1.5 hours of each participant’s time. Researchers spent, on average, at least eight hours per participant recruiting, coordinating, and scheduling discussion groups, running discussion sessions, and coding behavioral data. The study generally targeted current undergraduates, but three graduate students and one recent graduate also participated. Upon volunteering to take part in the study, participants were randomly assigned to either the treatment (n = 34) or control (n = 24) group using a coin flip. The total sample includes an un-randomized 16 person pilot of the experimental protocol. See S3 File for additional information on this pilot group, its characteristics, and analyses showing their inclusion does not affect the main findings.

Pre-test survey

Fig 1 presents the study design including information provided to the treatment and control groups (in black) and the points at which we measured their opinion regarding Paterno’s firing (in red). Both groups were administrated a pre-test survey using Qualtrics. The treated group completed this survey before attending a discussion session. In addition to basic demographic characteristics, we collected a number of psychological and behavioral traits for every participant. Ideology was measured by an attitudinal measurement of ideology, a Liberalism-Conservatism scale [ 77 ] widely used to prevent measurement error that arises from the difficulty in accurately collapsing a complex view of politics into a single dimension. This measure of ideology is well validated (e.g., Bouchard et al. 2003) and serves as the basis for modern definitions of ideology across disciplines [ 78 , 79 ]. The measure relies on respondents simply agreeing or disagreeing with a broad range of political and social topics, from evolution to taxes. In this case, we used 48 different topics, which generate an additive scale of conservatism ranging from 0 (very low) to 48 (very high). In addition to measuring our participant’s political ideology, we assessed their self-esteem using Rosenberg’s [ 80 ] scale and personality using McCrae and John’s [ 81 ] 44-question Big 5 dimensions of personality: openness to experience, conscientiousness, extraversion, agreeableness, and neuroticism.

• PPT PowerPoint slide
• PNG larger image
• TIFF original image

This figure presents each phase of the study, including information provided to treated and control groups (in black) and the points at which we measured their opinion of the Paterno firing (in red).

https://doi.org/10.1371/journal.pone.0196600.g001

Finally, all participants were asked their opinion on five policies that affect undergraduates at Penn State: alcohol possession on campus; government oversight of academic performance; the firing of Paterno; prevention of State Patty’s Day celebrations; and use of the student activities fee. Participants recorded their opinion using a five-point Likert scale from “strongly agree” to “strongly disagree.” We included five different topics on the survey so that treatment group participants would be unsure as to which topic they would be discussing.

Discussion group

After completion of the online survey, participants in the treatment group were scheduled individually for a discussion session. Each discussion group was comprised of a single participant and two to four trained confederates (we compare differences in the number of experimenters and find no effects; for more information see S4 File ). A total of five unique confederates, three females and two males, were used across the length of the study. Among them were four political science Ph.D. candidates of varying experience and one recent graduate who majored in political science. The confederates looked young and dressed informally, and were not distinguishable from our undergraduate students. In terms of training, the confederates were not strictly scripted so that the discussion would not appear forced or scripted. Instead, the experimenter and other volunteers took part in pre-experiment tests as mock participants so that the confederates could argue both sides of the Paterno firing and develop the consistent points they used for the duration of the study (see S2 File ). Fig 2 shows a typical discussion session. Discussion sessions were held in a conference room with all of the group members sitting around a table. There was no fixed seating arrangement.

Clockwise from bottom left: Experimenter, confederate, confederate, participant, and confederate. Note the participant’s seemingly disengaged body language. This participant ultimately changed their opinion.

https://doi.org/10.1371/journal.pone.0196600.g002

At the beginning of each discussion session, the experimenter reminded the group that the general purpose of the experiment is to understand political decision-making and how individuals form political opinions. They were told that a topic was randomly selected for each discussion group from the five included in the pre-test survey, with their topic being the firing of Paterno. Prior to the start of open discussion, group members were provided a sheet of excerpts from the Freeh Report [ 82 ] regarding Paterno’s involvement in the Sandusky scandal at Penn State (see S1 File ). They were told that the information was drawn from independent investigations and was meant to refresh their memories, given that two years had passed since the firing.

After providing time to read the information sheet, the group was polled verbally regarding whether or not they believed Paterno should have been fired (yes or no). The participant was always asked to answer first. This allowed the confederates to subsequently express the opposite opinion throughout the discussion. Though very little time passed between completion of the pre-test surveys and participation in the discussion groups, we did not rely on the opinions expressed in the pre-test surveys as the basis of our confederates’ opinion. We recorded and used the verbal response as the respondents’ opinion. This also ensures that our confederates were responding to the precise opinion held by the participant at the start of the discussion session. This way we could track the effect of conformity pressure on their opinion throughout the session.

The group was then provided 30 minutes for open discussion; however, discussion was allowed to go beyond 30 minutes in order allow participants to finish any thoughts and reflect a more natural interaction. During this discussion, up to four confederates argued the opposition position to greater or lesser degrees depending on the confederate, including responding to and interacting with the participant and even agreeing with the participant on certain points. At the conclusion of the discussion time, group members were told that researchers wished to understand their true opinion at that moment and that we would be aggregating the individual opinions from our groups in order to gain a sense of overall student opinion on each of the five topics. Thus, they were instructed to complete an anonymous ballot with their final opinion. The anonymous ballot allowed us to measure whether their opinion had actually changed during the discussion, conforming to other people’s behavior due to private acceptance that what they are saying is right, or were only publicly complying with other people’s behavior, without necessarily believing in what they are doing or saying.

Each discussion session was video recorded for the purposes of coding both verbal and non-verbal indications of their opinion. Two coders were hired to review each discussion session video and record a series of behavioral characteristics of the participants (not reported in this paper) as well as their impression regarding whether the participants verbally changed their opinion during the course of the discussion (a binary yes/no). The principal investigators also coded each video. We used the modal code from all four coders, with the principal investigators re-reviewing the videos to break six ties. Fleiss’s Kappa [ 83 ] indicates moderate agreement among raters on the verbally expressed opinion (0.54, p < 0.001).

The combination of anonymous balloting and video recording for verbal cues is an important aspect of the study design that allows us to pull apart whether participants conformed out of a desire to be right, liked, or a combination of the two. Finally, we debriefed each participant to explain the full purpose of the study, including any and all possible points of deception, and to gather information about their personal feelings on being in the minority during the discussion.

Control group

We utilized a control group in order to identify the independent effect of social pressure on opinion change. Their behavior established a baseline expectation for the amount of opinion change we could expect with just the introduction of new information and no interpersonal interaction. This baseline then allows us to compare the two groups, social influence treatment and control, in order to tease apart the independent and joint effects of social conformity pressure and information on opinion change.

To this end, the control group took the same pre-test survey as the treatment group. However, after completion of the survey, instead of being in a deliberative session, control group participants read additional information on a topic that was “randomly” selected from the five opinion questions. Based on their opinion regarding the firing of Paterno, we presented them with the same sheet of information provided to the treated as well as a summary of the same pro- and counter-arguments used by the actual confederates during the discussion group sessions (see S1 and S2 Files). After reading these, control group participants were asked whether they believe Paterno should have been fired (yes or no) and the strength of that opinion (very strongly, somewhat strongly, neutral). If they changed their opinion at this juncture, we consider they did so only because of the introduction of new information, as there was an absence of social pressure. Thus, our design allows us to parse out the effect of the discussion group and the social pressure emerging from an unanimity of opinion opposite to the participants.

Results and discussion

The core finding of this study revolves around the question to what extent will people conform to an opposing opinion on a topic that is salient, politically charged, and informs some aspect of their identity? Furthermore, can we evoke deviation rates similar to the foundational studies that relied on less complex aspects of one’s psychology [ 1 ]? And most important, what type of change is occurring? For those participants who changed their opinions, was it due to new information (i.e., private acceptance), social pressure (i.e., public compliance), or some combination of the two? To answer these questions, we first examined the degree of opinion change in both the treatment and control groups. For the control group, we compared their initial opinion from the pre-test survey with the opinion they provided after reading the information sheet and counter-arguments. Fig 3 displays the percentage of each group that did and did not change their opinion. Within the control group, which received the same information as the discussion group, but had no social interaction, only 8 percent of the participants changed their opinion. The information-based change we observed is consistent with extant research [ 84 , 85 ]. In addition, though a large proportion of the control group did not change their opinion, some did moderate it (i.e., strengthened or weakened) based on the receipt of new information alone. See S5 File for a further breakdown of these changes.

https://doi.org/10.1371/journal.pone.0196600.g003

Turning to the treatment group, 38 percent of our treated participants changed their opinion between the initial vote (after receiving information and prior to the discussion) and the final secret ballot. Our complex, identity, and value-laden topic returned findings that comport remarkably close to the deviation rates of Asch [ 2 ] and those that follow (for a meta-analysis, see [ 6 ]). If we consider all other things equal, the 30 percent increase in opinion change is dependent on the treatment of participating in the group discussion (χ 2 = 5.094, p < 0.05). This finding remains unchanged if the 16 non-randomized members of the pilot study are removed from the treatment group (though the p-value of the chi-square declines to 0.10, due to the smaller n, see S3 File ). As further evidence, Table 1 presents logistic regression results demonstrating the treatment effect. Namely, being in the treatment condition increases the odds of opinion change by 581 percent. Meaning, social pressure and/or the personal delivery of information, as opposed to simple exposure to new information, had a profound influence on either true opinion change through private acceptance or conformity through public compliance. Due to the small sample size, we are hesitant to include additional covariates in this model, but instead use t-tests below to examine differences in the characteristics of participants who changed their opinion and those who did not.

https://doi.org/10.1371/journal.pone.0196600.t001

Sources of change

Moving to our secondary analyses, the research design also allowed us to parse out the specific sources of change within the treatment group. Recall we accounted for both true opinion change (i.e., the anonymous ballot at the end of discussion) and verbal opinion change (i.e., declared opinion change during group discussion captured in video and coded by independent raters) for those in the treatment condition. Therefore, we divided those in the treatment group into four subgroups in order to better understand why they changed their opinion. Table 2 shows the percentages of participants in the treatment group who changed their opinion overtly, covertly, or not at all. In sum, 47 percent did not change their opinion between the start and end of the discussion session. A total of 33 percent changed both overtly and covertly, meaning they verbally expressed an opinion change and wrote a changed opinion on their secret ballot. We argue that this group responded to a combination of the desires to be right and liked. Of the remaining participants, 10 percent changed due to a desire to be liked (overtly, but not covertly) and 10 percent due to a desire to be right (covertly, but not overtly). Though only anecdotal, one of the participants in the desire to be right category went so far as to tell the experimenter that he agreed with the group but adamantly refused to agree openly. Such participants were swayed by the introduction of new information out of a strong desire to be right, but apparently did not want to look like they were changing their opinion. Thus, our first set of analyses confirms that information plays an important role in opinion change, but social pressure also has a substantive and, at least in this context, a larger effect. For even a topic so important to one’s identity, participants changed their previously held opinions.

https://doi.org/10.1371/journal.pone.0196600.t002

Psychological differences

Having established the main findings of our study and the relative import of the two causal mechanisms for why participants changed their opinion, we now turn to examining how underlying traits, including ideology, personality, age and sex, differ between those that changed their opinion and those that did not. Demographic differences are included for descriptive purposes. First, we assessed differences between pro- and anti-firing participants. Second, we examined the relationship between direction of opinion change and trait differences between participants that changed their opinion and those that held firm. Due the nature of the experiment and specific focus on the question of causality, these tests are secondary to the main findings in the paper. For the following analyses, the sample sizes are small and in some cases and the findings only speculative.

Across both the treatment and control groups, the pre-test survey showed almost two-to-one support for Paterno keeping his job (i.e., against the firing). As mentioned earlier, “JoePa” was not only a symbol of Penn State, but also an icon to its students, and to some degree seen as a reflection of them. Table 3 displays the average demographic and psychological measures for those for and against the firing, based on the pre-test survey. The only statistically significant difference between the groups is their political ideology. The group opposed to Paterno’s firing is, on average, more conservative in their attitude positions than those that called for his firing. It is important to note that these are college students, and thus the overall distribution of ideology exhibits a liberal skew. However, Fig 4 demonstrates that the pro-firing group is not only less conservative, on average, but is also more ideologically narrow, whereas those that did not support the firing are more conservative, but also drawn from a wider ideological span. This finding suggests that ideology is a substantial factor for individuals that supported the firing. Whereas support for Paterno may have a less pronounced ideological dimension, those supporting his firing may focus more narrowly on the issue of child abuse and the responsibility of those in leadership to protect vulnerable citizens. Given that ideology is the only difference we could identify among participants’ opinions prior to the start of the experiment, we next examined whether there were differences between participants who changed their opinion and those that did not in both the treatment and control conditions.

https://doi.org/10.1371/journal.pone.0196600.t003

https://doi.org/10.1371/journal.pone.0196600.g004

Tables 4 and 5 provide a sense of how demographic and psychological characteristics differ between participants who changed their opinion and those who did not. Table 4 includes both treatment and control participants, whereas Table 5 focuses solely on the treatment group. We found evidence both supporting and refuting our hypotheses presented above. There were consistent significant differences ( p < 0.05) in conservatism and conscientiousness. Namely, participants who changed their opinion are less conservative and less conscientious. Given the reported relationships between these two traits, this finding makes sense. Additionally, when all subjects are pooled ( Table 4 ), there is also a significant difference in neuroticism, with opinion changers registering higher on this scale. Both suggest that political and psychological traits may play a role in the mean shift demonstrated above. There were no differences based on the number of confederates. Meaning, participants were no more or less effected by social pressures from greater (4) or fewer (2) opponents in the discussion environment. These results demonstrate that individual differences exist across individuals that change their opinion and those that do not. Additional research will be required to both confirm and expand upon these findings. What we do find, however, is in line with expectations derived from past research and points to useful areas of future inquiry.

https://doi.org/10.1371/journal.pone.0196600.t004

https://doi.org/10.1371/journal.pone.0196600.t005

All participants were debriefed upon completion of the discussion and informed to all aspects of the study. Participants were asked during the debriefing how they felt about being the only dissenting voice. Forty-seven percent of the treatment group participants freely offered that they felt pressured or intimidated. Twenty-nine percent also freely said that they felt like they had to dig in and defend their position during the discussion. This included six people that ultimately changed their minds. One said, “I’m not getting any support in this room. Alright I’ll defend my own position.” Another said, “I feel extra pressure to explain myself.” For some, their defensiveness continued into the debriefing. In particular, some students that did not change their opinion continued defending themselves when talking one-on-one with the experimenter, even after it was explained no matter which position they took, they would face opposition. This demonstrates that some participants are put on the defensive when faced with a unified opposition. Of those that expressed feeling defensive, some dug-in deeply and did not budge at all, while others opened up to the influence of their peers as the discussion progressed. This behavior comports the foundational work of Asch [ 1 , 2 ] and Milgram [ 86 ] and strongly suggests that our participants indeed experienced social pressure in the treatment condition, but differs in that it highlights the variance in how individual’s react to such pressure.

Limitations

We wish to call attention to two specific limitations of this study that are discussed above and in the supplementary materials, but warrant further mention. The first limitation is the inclusion of a meaningful, relative to the overall sample size, non-randomized pilot of the treatment condition. While this had no substantive effect on the results, it is important to recognize and we discuss this in more detail in the S3 File . Second, Fig 1 makes apparent that we use two similar, but slightly different scales for opinion throughout the study. Namely, pre-test opinion is measured on a five-point Likert scale and the remaining opinion measures are dichotomous (yes/no), with an additional strength question for the control group. Our primary analyses, however, rely on the comparison of the two yes/no answers in the treatment group; the verbal designation of yes/no at the beginning of the discussion section and the yes/no in the post discussion ballot. We further discuss this in the S5 File .

Finally, to some the small sample size of the study may be a limitation, especially those concerned about a replication crisis in Social Psychology [ 87 ]. We would respond, however, that the intensive nature of this study in terms of researcher hours and treatment condition makes it difficult to scale-up. Thus, a multi-site replication is likely the best approach to assessing the veracity of these findings [ 88 , 89 ]. We encourage such replication and have provided all materials necessary on the corresponding author’s Dataverse ( http://dx.doi.org/10.7910/DVN/YVCPDT ). Additional lessons relevant to replication work and laboratory experiments in political science can be found in Mallinson (2018) [ 90 ].

Conclusions

While researchers have examined the roles of social influence (public compliance) and new information (private acceptance) on opinion change, the two are less often examined concurrently and the explicit causal arrows are more often assumed than tested through an experiment. Furthermore, social conformity is a complex concept to measure through surveys or interviews alone. Live interaction provides an optimal means to understand social pressures. Our experiment was designed specifically to further unpack the causal mechanisms underlying opinion change and test whether a person’s values and identity are subject to social pressure. Furthermore, the selection of the topic of study, the firing of an important symbol of Penn State, also allowed us to explicate the extent to which information and social pressure challenge a person’s deeply held values and identity. We find that while information has an important role in changing people’s opinions on a highly salient topic that is attached to a group identity, the social delivery of that information plays a large and independent role. Most individuals that changed their opinion did so out of some combination of the two forces, but there were people who only changed their opinion overtly in order to gain social acceptance as well as those who did not want to give the appearance of changing their mind, but still wanted to be right.

These findings have important implications for research on social and political behavior. They reinforce the understanding that citizens and elites cannot be simply viewed as rational utility maximizers independent of group dynamics. Yet, at the same time, the desire to be right and information remain critical components of opinion change. Furthermore, there are important individual differences such as ideology, self-esteem, and personality that appear to have a role in conformity. Exposure to politics and political discussion are fundamentally social, and therefore behavior is conditioned on the combination of the information one receives, and the social influence of the person or group providing that information interacting with one’s disposition. All should be considered when examining any inter-personal, social or political outcome. Be it a deliberative setting like a jury or a town hall meeting or informal gatherings of citizens, or political elites for that matter, changes in behavior are not simply due to rational information-driven updating, and even when they are, that updating may be pushed by the social forces that we experience in our interactions with other humans in variegated ways dependent upon the characteristics of the individual (for example, see [ 91 ]). This was the case for simple and objective stimuli, like Asch’s lines, and it is also the case in our context-laden experiment that focuses on the complexities of personal identity and opinion. That is, the conformity of social and political values relies on the same psychological mechanisms underlying general conformity.

Beyond theoretical and empirical importance for the study of social and political behavior, these findings also hold normative importance for democratic society. The normative implications are perhaps best exemplified by the organizational and personal turmoil that followed the revelation of child abuse by priests in the Catholic Church. Politics forms important aspects of the social and personal identities of elites and citizens, more so today than ever before [ 92 , 93 ]. People include their political party, positions on particular issues (e.g., environmentalism), and membership in political, religious, social and academic organizations, among other things, as key aspects of their identities. Our experiment helps us better understand how individuals behave when part of that identity is challenged.

That being said, no design is perfect, and this experiment only unpacks part of the causal mechanism. Like the early work on social conformity, it serves as a foundation for future studies to extend upon and further explicate the causal mechanism. For example, an extension on this design, such as controlling variation in the type and number of confederates [ 44 , 94 ], could help us better understand the nature and amount of pressure necessary to induce conformity across a variety of individual characteristics. For example, a potentially fruitful avenue of extension would be to provide the participant with one supportive confederate who verbally changes their opinion during the discussion. Having support reduces conformity pressure, but deviation by that support should increase it. Additionally, while we identify individuals whose behavior was prompted by either social pressure or information, the largest group responded to a combination of the two. Further parsing out the interaction between information, persuasion, pressure and the complexity of human dynamics will require an even more complex research design on a larger scale. The numerous extensions of Asch’s original experiment demonstrate the wealth of potential extensions of this design that can help unpack this black box. Doing so requires an incremental approach that will be time and resource intensive. This study provides the foundation for those next steps.

Supporting information

S1 file. information sheet provided to both treatment and control groups..

https://doi.org/10.1371/journal.pone.0196600.s001

S2 File. Confederate talking points.

https://doi.org/10.1371/journal.pone.0196600.s002

S3 File. Randomization.

https://doi.org/10.1371/journal.pone.0196600.s003

S4 File. Use of deception in the study design.

https://doi.org/10.1371/journal.pone.0196600.s004

S5 File. Breakdown of opinion change in the treatment and control groups.

https://doi.org/10.1371/journal.pone.0196600.s005

Acknowledgments

An earlier version of this paper was presented at the 2013 American Political Science Association Annual Meeting in Chicago, Illinois, and the April 2014 Center for American Political Responsiveness Brown Bag in State College, Pennsylvania. We would like to thank the editor, the anonymous reviewer, Ralf Kurvers, Rose McDermott, and conference attendees for their helpful comments and suggestions on this manuscript. We are also grateful to Ralf Kurvers for providing Fig 1 . We would like to thank our research assistants, Ronald Festa, Emilly Flynn, Christina Grier, Christopher Ojeda, Kimberly Seufer, and Matthew Wilson, that helped make this experimental protocol a success.

• 1. Asch SE. Effects of Group Pressure Upon the Modification and Distortion of Judgments. In: Guetzkow H, editor. Groups, leadership and men; research in human relations. Oxford, England: Carnegie Press; 1951.
• View Article
• Google Scholar
• 3. Sherif M, Harvey OJ, White BJ, Hood WR, Sherif CW. Intergroup Conflict and Cooperation: The Robbers Cave Experiment. Norman, OK: Institute of Group Relations; 1954/1961.
• PubMed/NCBI
• 16. Sherif M. The Psychology of Social Norms. Oxford, England: Harper; 1936.
• 18. Alwin DF, Cohen RL, Newcomb TM. Political attitudes over the life span: The Bennington women after fifty years. Madison, WI: University of Wisconsin Press; 1991. xxviii, 422 p. p.
• 19. Converse PE. The Nature of Belief Systems in Mass Publics. In: Apter D, editor. Ideology and Discontent. New York, NY: Free Press; 1964.
• 21. Zuckerman AS. The Social Logic of Politics: Personal Networks as Contexts for Political Behavior Philadelphia, PA: Temple University Press; 2005.
• 22. Sinclair B. The Social Citizen: Peer Networks and Political Behavior. Chicago, IL: University of Chicago Press; 2012.
• 24. Huckfeldt R, Johnson PE, Sprague J. Political Disagreement: The Survival of Diverse Opinions within Communication Networks. Cambridge: Cambridge University Press; 2004.
• 26. Cialdini RB, Trost MR. Social influence: Social norms, conformity and compliance. In: Gilbert DT, Fiske ST, Gardner L, editors. The Handbook of Social Psychology. 4 ed. Hoboken, NJ: Wiley and Son; 1998. p. 151–92.
• 28. Ellickson RC. The Evolution of Social Norms: A Perspective from the Legal Academy. In: Hecter M, Opp K-D, editors. Social Norms. NY: Russell Sage Foundation; 2001. p. 35–75.
• 29. Horne C. Sociological Perspectives on the Emergence of Social Norms. In: Hecter M, Opp K-D, editors. Social Norms. NY: Russell Sage Foundation; 2001. p. 3–34.
• 31. Fearon JD. Deliberation as Discussion. In: Elster J, editor. Deliberative Democracy. Cambridge: Cambridge University Press; 1998.
• 32. Gastil J. Democracy in Small Groups: Participation, Decision Making, and Communication. Philadelphia, PA: New Society Publishers; 1993.
• 33. Habermas J. Between Facts and Norms: Contributions to a Discourse Theory of Law and Democracy. Cambridge, MA: MIT Press; 1996.
• 34. Rawls J. The Law of Peoples. Cambridge, MA: Harvard University Press; 1999.
• 37. Hibbing JR, Theiss-Morse E. Stealth Democracy: Americans’ Beliefs about How Government Should Work: Cambridge University Press; 2002.
• 38. Abramson J. We, the Jury. New York, NY: Basic Books; 1994.
• 39. Kalven HJ, Zeisel H. The American Jury. Chicago, IL: University of Chicago Press; 1970.
• 40. Verba S. Small Groups and Political Behavior. Princeton, NJ: Princeton University Press; 1961.
• 54. Hock RR. Forty Sudies That Changed Psychology. Upper Saddle River, NJ: Pearson Prentice Hall; 2006.
• 73. Mooney CZ. The Public Clash of Private Values: The Politics of Morality Policy. New York, NY: Chatham House Publishers; 2001.
• 74. CBS. Obama: ‘Soul-searching’ from Penn State Scandal.2011 February 3, 2014. http://www.cbsnews.com/news/obama-soul-searching-from-penn-state-scandal/ .
• 75. Keppel G, Wickens TD. Design and Analysis: A Researcher’s Handbook. Englewood Cliffs, NJ: Prentice Hall; 2004.
• 80. Rosenberg M. Society and the Adolescent Self-Image. Princeton, NJ: Princeton University Press; 1965.
• 82. Freeh SS, LLP. Report of the Special Investigative Counsel Regarding the Actions of The Pennsylvania State University Related to the Child Sexual Abuse Committed by Gerald A. Sandusky. 2012.
• 86. Milgram S. Obedience to Authority: An Experimental View. 1st ed. New York: Harper & Row; 1974. xvii, 224 p. p.
• 91. Grimmer J, Messing S, Westwood SJ. Estimating Heterogeneous Treatment Effects and the Effects of Heterogeneous Treatments with Ensemble Methods. n.d.

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

• Publications
• Account settings

Preview improvements coming to the PMC website in October 2024. Learn More or Try it out now .

• Advanced Search
• Journal List
• HHS Author Manuscripts

Social influence on positive youth development: A developmental neuroscience perspective

Susceptibility to social influence is associated with a host of negative outcomes during adolescence. However, emerging evidence implicates the role of peers and parents in adolescents’ positive and adaptive adjustment. Hence, in this chapter we highlight social influence as an opportunity for promoting social adjustment, which can redirect negative trajectories and help adolescents thrive. We discuss influential models about the processes underlying social influence, with a particular emphasis on internalizing social norms, embedded in social learning and social identity theory. We link this behavioral work to developmental social neuroscience research, rooted in neurobiological models of decision-making and social cognition. Work from this perspective suggests that the adolescent brain is highly malleable and particularly oriented towards the social world, which may account for heightened susceptibility to social influences during this developmental period. Functional magnetic resonance imaging (fMRI) has been used to investigate the neural processes underlying social influence from peers and family as they relate to positive, adaptive outcomes in adolescence. Regions of the brain involved in social cognition, cognitive control, and reward processing are implicated in social influence. This chapter underscores the need to leverage social influences during adolescence, even beyond the family and peer context, to promote positive developmental outcomes. By further probing the underlying neural mechanisms as an additional layer to examining social influence on positive youth development, we will be able to gain traction on our understanding of this complex phenomenon.

I. A developmental social neuroscience perspective on social influence

If your friends jumped off a cliff, would you too? Everyone has heard this phrase at some point in their lives, either in the position of a worried parent or not-so-worried teenager. Indeed, a vast literature indicates that health-compromising risky behaviors increase when adolescents are with their peers (reviewed in Van Hoorn, Fuligni, Crone, & Galván, 2016 ). Emerging evidence from developmental neuroscience suggests that the adolescent brain is highly plastic and undergoes a major “social reorientation” ( Nelson, Leibenluft, McClure, & Pine, 2005 ), which may render adolescents particularly susceptible to social influences. While the focus of most research, popular media, and parental worries has been directed towards seeing social influence susceptibility as negative, leading teens to engage in dangerous behaviors, recent attention has sought to understand how adolescents’ heightened social influence susceptibility may be redirected towards positive, adaptive behaviors.

In this chapter, we review emerging evidence highlighting how social influences from both peers and family can play a positive role in adolescents’ adjustment. We first define social influence, focusing on two influential theories, social learning theory and social identity theory, both of which discuss social influence in terms of internalizing group norms. We then review literature highlighting several sources of social influence, including dyadic friendships, cliques, social networks, parents, siblings, and the larger family unit. Given the important neural changes occurring in adolescence, we describe the important role of maturational changes in the developing brain that may underlie susceptibility to social influence. We discuss prominent models of adolescent brain development and then review emerging research highlighting how family and peer influence are represented at the neural level. Finally, we conclude with future directions underscoring the need to capitalize on social influences from peers and parents during adolescence, examine different sources of social influence in the context of the larger social network, and expand our knowledge on the neural mechanisms underlying social influence.

II. Defining Social Influence

What is social influence? At the most basic level, social influence “comprises the processes whereby people directly or indirectly influence the thoughts, feelings, and actions of others” ( Turner, 1991 , pg. 1). When most people think of social influence, images of peers cheering on their friends to drink, do drugs, or engage in risky and reckless behavior likely come to mind. Popular misconceptions about social influence that saturate the media and parents’ worries too often focus on these very explicit, overt, and negative examples. But what many do not realize is that social influence is much more subtle and complex, and cannot often be identified so easily. In fact, direct peer pressure is not associated with adolescents’ smoking intentions, whereas the perceived behaviors of peers are ( Vitoria, et al., 2009 ). Moreover, social influence has many positive implications, for instance, exposing youth to positive social norms such as school engagement, cooperating with peers, donating money, and volunteering for a good cause. In this section, we will review prominent theories of social influence with a particular emphasis on the internalization of social norms, embedded in social learning and social identity theory.

A. Social Norms

A social norm is “a generally accepted way of thinking, feeling, or behaving that is endorsed and expected because it is perceived as the right and proper thing to do. It is a rule, value or standard shared by the members of a social group that prescribes appropriate, expected or desirable attitudes and conduct in matters relevant to the group” ( Turner, 1991 , pg. 3). Group norms are further defined as “regularities in attitudes and behavior that characterize a social group and differentiate it from other social groups” ( Hogg & Reid, 2006 , pg. 7). Norms are therefore shared thoughts, attitudes, and values, governing appropriate behavior by describing what one ought to do, and in essence prescribe moral obligations ( Cialdini & Trost, 1998 ). Social norms are communicated by what people do and say in their everyday lives, which can be indirect (e.g., inferring norms from others’ behaviors) but also direct (e.g., intentionally talking about what is and is not normative of the group; Hogg & Reid, 2006 ). Deviation from the social norms of a group can result in loss of social status or exclusion, particularly if the social norm is important to the group ( Festinger, 1950 ). Thus, norms serve to reinforce conformity by promoting the need for social acceptance and avoidance of social punishments (e.g., Deutsch & Gerard, 1955 ).

Social norms have a profound impact on influencing attitudes and behaviors, even though people are typically unaware of how influential social norms are ( Nolan et al., 2008 ). In fact, people are strongly influenced by social norms even when they explicitly reject such norms ( McDonald, Fielding, & Louis, 2013 ). In a classic study, Prentice and Miller (1993) asked Princeton undergraduates how comfortable they versus the average Princeton undergraduates are with drinking. Results across several studies converged on the same conclusion – individuals believe others are more comfortable with drinking than themselves. This phenomenon is referred to as pluralistic ignorance (e.g., Prentice & Miller, 1996 ), which occurs when people personally reject a group norm, yet they incorrectly believe that everyone else in the group engages in the behavior. This introduces a “perceptual paradox” – in reality the behavior is not the norm since nobody engages in it, yet it is the group norm because everyone thinks everyone else does engage in the behavior ( Hogg & Reid, 2006 ). Adolescents also misjudge the behaviors of their peers and close friends. Referred to as the false consensus effect, adolescents misperceive their peers’ attitudes and behaviors to be more similar to their own or even overestimate their peers’ engagement in health-risk behaviors ( Prinstein & Wang, 2005 ). Thus, adolescents overestimate the prevalence of their peers’ behaviors and use their (mis)perceptions of social norms as a standard by which to compare their own behavior.

B. Social Learning Theory

Social learning theory provides the basis for how social norms are learned and internalized during adolescence. Although this theory was originally developed to describe criminality and deviant behavior, its propositions can also be applied to positive social learning. Akers ( 1979 , 2001 , 2011 ) identified four core constructs of social learning: differential association, differential reinforcement, imitation or modeling, and definitions . Differential association refers to the direct association with groups who express certain norms, values, and attitudes. The groups with whom one is associated provides the social context in which all social learning occurs. The most important groups include family and friends, but can also include more secondary sources such as the media ( Akers & Jensen, 2006 ). According to Sutherland’s differential association theory ( Sutherland, Cressey, & Luckenbill, 1992 ), learning takes place according to the frequency, duration, priority, and intensity of adolescents’ social interactions. Adolescents will learn from and internalize social norms if (1) associations occur earlier in development (priority), (2) they associate frequently with others who engage in the behavior (frequency), (3) interactions occur over a long period of time (duration), and (4) interactions involve individuals with whom one is close (e.g., friends and family) as opposed to more casual or superficial interactions (intensity). The more one’s patterns of differential association are balanced towards exposure to prosocial, positive behavior and attitudes, the greater the probability that one will also engage in positive behaviors. Association with groups provides the social context in which exposure to differential reinforcement, imitation of models, and definitions for behaviors take place ( Akers, 1979 ).

Differential reinforcement refers to the balance of past, present, and anticipated future rewards and punishments for a given behavior ( Akers & Jensen, 2006 ), and includes the reactions and sanctions of all important social groups, especially those of peers and family, but can also include other groups such as schools and churches ( Akers, 1979 ; Krohn et al., 1985 ). In particular, behaviors are strengthened through rewards (i.e., positive reinforcement; e.g., peer acceptance of behaviors) and avoidance of punishments (i.e., negative reinforcement; e.g., peer rejection of behaviors) or weakened though receiving punishments (i.e., positive punishment; e.g., being grounded by parents) and loss of rewards (i.e., negative punishment; e.g., having the family car taken away; Akers, 1979 ). Behaviors that are reinforced, either through social rewards or the avoidance of social punishments, are more likely to be repeated, whereas behaviors that elicit social punishments are less likely to be repeated ( Akers, 2001 ). Thus, through differential reinforcement, individuals are conditioned to internalize the social norms that are valued by the group.

Social behavior is also shaped by imitating or modeling others’ behavior. Individuals learn behaviors by observing those around them ( Bandura, 1977 , 1986 ), particularly close others such as parents, siblings, or friends. The magnitude of social learning, and imitation in particular, is strengthened the more similar the individuals are ( Bandura, 1986 , 2001 ). Social influence has an effect on youth when adolescents are exposed to the behaviors and norms of others (i.e., mere exposure) and observe the positive outcomes others receive from such behaviors (i.e., vicarious learning). Adolescents then internalize such social norms and model the behaviors in future instances.

Finally, definitions are the attitudes, rationalizations, or meanings that one attaches to a given behavior that define the behavior as good or bad, right or wrong, justified or unjustified, appropriate or inappropriate ( Akers & Jensen, 2006 ). The more individuals have learned that specific attitudes or behaviors are good or desirable (positive definition) or as justified (neutralizing definition) rather than as undesirable (negative definition), the more likely they are to engage in the behavior ( Akers, 1979 ). These definitions are learned through imitation and subsequent differential reinforcement by members of their peer and family groups. Although there may be norm conflict in terms of the definitions promoted by one’s peers (e.g., positive definition for alcohol) and parents (e.g., negative definition for alcohol), the relative weight of such definitions will determine whether an adolescent endorses the social norm and engages in the behavior. An individual will engage in the behavior when the positive and neutralizing definitions of the behavior offset the negative definitions ( Akers, 1979 ).

C. Social Identity Theory

Group identification is essential for understanding the effects of social norms ( Turner, 1991 ). According to social identity theory, social influence occurs when individuals internalize contextually salient group norms, which set the stage for their self-definition, attitudes, and behavioral regulation ( Tajfel, 1981 ; Tajfel & Turner, 1979 ; Hogg & Reid, 2006 ). From a social identity perspective, norms reflect a shared group prototype, which are individuals’ cognitive representations of group norms ( Hogg & Reid, 2006 ). Group prototypes describe normative behaviors and prescribe behavior, indicating how one ought to behave as a group member. Thus, strong group identification can lead to social influence and conformity because individuals endorse the behaviors they should engage in based on the social norms prescribed by group prototypes ( Terry & Hogg, 1996 ).

The family is the first and primary social group to which most individuals belong ( Bahr et al., 2005 ), whereas friends become an increasingly salient social identity during adolescence, a developmental period marked by a need to belong and affiliate with peers ( Crockett et al., 1984 ; Newman & Newman, 2001 ; Kroger, 2000 ; Furman & Buhrmester, 1992 ; Hart & Fegley, 1995 ). Importantly, the social environment can activate certain identities and determine whether an individual will be influenced ( Oakes, 1987 ). Across development (e.g., from childhood to adolescence) and across contexts (e.g., at school versus at home), different social identities (e.g., family versus peers) will be more or less salient, affecting whether group norms are strongly internalized and activated.

Adolescents are not only influenced by a single salient group but also by the norms of multiple groups ( McDonald et al., 2013 ), including family, close friends, out-group peers, and the broader societal norms. When more than one social identity is activated, norm-conflict may occur, especially if there are inconsistencies across group norms ( McDonald et al., 2013 ). A particularly prominent example of this likely occurs in adolescents’ daily lives when the norms and valued behaviors of the peer group (e.g., drinking alcohol is fun) conflict with the norms internalized at home (e.g., drinking alcohol is unacceptable behavior). Although seemingly bad, norm conflict can potentially increase motivation to engage in a behavior, because the norm conflict reinforces the need to personally act ( McDonald et al., 2013 ). As an example, if a teen sees a peer being bullied at school, and her close friends are cheering on the bully to continue picking on the teen, but another group of her peers is expressing concern for the teen, an adolescent may be moved to act and stick up for the victim due to this conflict, because she sees the need to personally act. Thus, when multiple group identities are activated and norm-conflict occurs, teens may be motivated to engage in a positive behavior ( McDonald et al., 2013 ).

III. Social influence on positive youth development

Social learning and social identity theories highlight that a myriad of social influences affect positive adjustment during adolescence. Sources of social influence include peers, family, teachers, other attachment figures (e.g., coach of sports team, youth group leader) and even (social) media ( Akers 1979 ; Bandura, 2001 ; McDonald et al., 2013 ). In this chapter, we specifically focus on social influences from peers and family and their interactions, given the saliency of developmental changes in these social relationships during adolescence ( Bronfenbrenner & Morris, 2006 ). Although peers are often referred to as a unified construct (i.e., persons of the same age, status, or ability as another specified person) previous research has assessed a wide range of peers that fall under this umbrella. Hence, we make a distinction between best friend dyads, smaller peer groups such as cliques, and larger peer groups of unknown others. Family influences similarly encompass multiple layers, and here we review influences from parents, siblings and their interactions within the larger family unit. Finally, we will discuss literature that examines these social influences simultaneously.

A. Peer influence on positive adolescent development

Peer influence has predominantly negative connotations and received most attention in the context of problem behaviors during adolescence. Indeed, extant research has shown that hanging out with the wrong crowd may increase deviant behaviors through processes of social reinforcement or “peer contagion” (reviewed in Dishion & Tipsord, 2011 ). For example, in videotaped interactions between delinquent adolescent males, rule-breaking behaviors (e.g., mooning the camera, drug use, obscene gestures) were socially reinforced through laughter, and this was predictive of greater delinquent behavior two years later ( Dishion, Spracklen, Andrews, & Patterson, 1996 ). Importantly however, the very same social learning process reinforced normative and prosocial talk (e.g., non-rule breaking topics such as school, money, family and peer-related issues) in non-delinquent adolescent dyads. This highlights the benefits of hanging out with the right crowd, and shows that imitation and social reinforcement in the peer context can also shape positive development. This section provides an overview of behavioral research that has examined peer socialization of prosocial behaviors during adolescence, as well as the application of peer processes in interventions to promote positive adjustment outcomes.

Peer influence in close friendships.

Prosocial behavior is a broad and multidimensional construct that includes cooperation, donation, and volunteering ( Padilla-Walker & Carlo, 2014 ). Given the association between prosocial engagement during adolescence and a range of adult positive adjustment outcomes (e.g., mental health, self-esteem, and better peer relations; reviewed in Do, Guassi Moreira, & Telzer, 2017 ), it is crucial to understand how peers can promote these behaviors. There is consistent evidence that best friends influence prosocial behaviors. In adolescent best friend dyads, a friend’s prosocial behavior is related to an individual’s prosocial goal pursuit, which, in turn, is associated with an individual’s prosocial behavior (e.g., cooperating, sharing and helping) ( Barry & Wentzel, 2006 ). These effects are moderated by friendship characteristics, including friendship quality ( Barry & Wentzel, 2006 ) and closeness between friends ( Padilla-Walker, Fraser, Black, & Bean, 2015 ; see Brown, Bakken, Ameringer, & Mahon, 2008 for a comprehensive chapter on pathways of peer influence). In particular, a friend’s prosocial behavior is most likely to influence adolescent’s own prosocial behavior when there is a strong positive relationship and greater closeness between friends, consistent with Sutherland’s differential association theory ( Sutherland et al., 1992 ). Moreover, not only do actual behaviors , but also perceived peer expectations about positive behaviors in the classroom predict greater prosocial goal pursuit and subsequent sharing, cooperating, and helping ( Wentzel, Filisetti, & Looney, 2007 ). These results underscore that getting along with peers is a powerful social motive to behave in positive, prosocial ways. Together, this work suggests that social influence on prosocial behavior is likely explained by processes of social learning ( Bandura, 2001 ).

Peer influence in small groups.

Experimental techniques allow one to manipulate peer effects on prosocial behaviors to better understand the mechanisms of social influence. In one study, we employed a public goods game, in which participants allocated tokens between themselves and a group of peers ( Van Hoorn, Van Dijk, Meuwese, Rieffe, & Crone, 2016a ). After making decisions individually, participants were ostensibly observed by a group of ten online peer spectators, who provided either prosocial feedback (i.e., likes for donating to the group) or antisocial feedback (i.e., likes for selfish decisions) on their decisions. Adolescents changed their behavior in line with the norms of the spectator group and showed greater prosocial behavior after feedback from prosocial spectators, but became more selfish with antisocial spectators ( Van Hoorn et al., 2016a ). Results from this study were corroborated by other experimental work showing that peers also positively influence intentions to volunteer ( Choukas-Bradley, Giletta, Cohen, & Prinstein, 2015 ). Moreover, adolescents conformed more to high-status peers’ intentions to volunteer than low-status peers’ intentions to volunteer, suggesting that adolescents are more susceptible to salient peers, consistent with social identity theory ( Hogg & Reid, 2006 ). In sum, experimental studies show that social norms are influential in the domain of prosocial behaviors (cooperation and intentions to volunteer), and can serve both as a vulnerability and an opportunity in adolescent development.

Peer influence in social networks.

Finally, other research has utilized social network analysis to study peer effects in the context of the larger group and highlights that specific characteristics of the larger social group may mitigate or magnify peer effects. For example, findings from one study illustrate that highly central (i.e., high status in larger network, trend setters in school) social groups within the larger network endorsed prosocial as well as aggressive and deviant behaviors, whereas groups with lower centrality (i.e., groups with low acceptance in the larger network) showed magnified socialization of deviant behaviors only ( Ellis & Zabartany, 2007 ). Moreover, adolescents tend to shift between different social groups, and there is evidence for socialization of prosocial behaviors from the attracting social group (i.e., the group to be joined), but not the departing social group (i.e., the group left behind) ( Berger & Rodkin, 2012 ). These results suggest that although adolescent’s membership in different peer groups can influence their engagement in positive and negative behaviors, there is often flexibility in the peer groups adolescents choose to identify with. Thus, to fully grasp peer effects, it is important to study multiple levels of the peer context, taking into account the dynamics between dyads, groups, and the larger social network.

Practical implications of positive peer effects.

The studies reviewed above provide a promising foundation for interventions that employ peer processes in order to potentially increase positive behaviors, as well as redirect negative behaviors during adolescence. One intervention that has shown promising effects is the Good Behavior Game, which teamed up non-disruptive and disruptive children ( Van Lier, Huizink, & Vuijk, 2011 ). When one child reinforced positive and prosocial classroom behaviors, their entire team was rewarded, resulting in more positive peer relations and reduced rates of tobacco experimentation three years later. Another study aimed to redirect collective school norms concerning harassment and utilized social networks to identify social referents (e.g., widely known adolescents or leaders of subgroups) within the school network ( Paluck & Shepherd, 2012 ). They then successfully used these social referents within the school setting to change their peers’ perceptions of norms concerning harassment over the school year, which reduced peer victimization. Collectively, these interventions take advantage of peer processes to change social norms and subsequently promote positive psychosocial outcomes.

B. Family influence on positive adolescent development

A considerable portion of research on social influence during adolescence focuses on the growing effect of peer relations, while deemphasizing the role of the family during this developmental transition. However, characterizing social influence during adolescence is hardly this simple. The family context continues to impact the attitudes, decisions, and behaviors of adolescents, particularly in guiding them toward positive adjustment (e.g., Van Ryzin, Fosco, & Dishion, 2012 ). The family context is a dynamic system that constantly affects the way in which adolescents think, behave, and make decisions. The family systems model presents these processes as each family member having continuous and reciprocal influence on one another throughout development ( Cox & Paley, 1997 ; Minuchin, 1985 ). For example, the family context influences each family member’s expectations, needs, desires, and goals. And together, each individual contributes to the family culture, including allocation of resources as well as family rituals, boundaries, and communication ( Parke, 2004 ). To put it simply, the whole is greater than the sum of parts, and the family is no exception during adolescent development ( Cox & Paley, 1997 ). In this section, we review research on families as a salient context for positive adolescent development and provide examples of parents, siblings, and multiple family members together in contributing toward adolescent adjustment.

Parental influence.

The importance of parental influence on positive adolescent development has been well established using longitudinal studies with multiple-informant questionnaires. Many studies converge on the finding that parental management predicts adolescent psychosocial adjustment. Authoritative parenting, which is characterized by frequent involvement and supervision, is associated with higher levels of adolescent academic competence and orientation and lower delinquency compared to other parenting styles ( Steinberg et al., 1994 ). Specifically, parents who are involved in their child’s school life (e.g., attendance, open-house) and who engage in intellectual activities (e.g., reading, discussing current events) tend to have adolescents who display high academic competence and school achievement ( Grolnick & Slowiaczek, 1994 ). In addition to managing and being involved in the lives of adolescents, parent-child relationship quality also affects adolescent development. Adolescent perceptions of closeness and trust with their parents predict better academic competence, engagement, and achievement ( Murray, 2009 ), as well as decreases in depressive symptoms for girls ( Guassi Moreira & Telzer, 2015 ).

Another approach to investigating parental influence on adolescent development includes examining parental beliefs and behaviors specific to the domain of interest, such as verbally promoting academics or athletics, or buffering against risky sexual behavior. When mothers take interest in, or value a specific behavior, such as doing well in school, their adolescents are also more likely to take interest ( Dotterer, McHale, & Crouter, 2009 ), which is an example of attitude definitions in social learning theory ( Akers & Jensen, 2006 ). One study examined maternal influences on adolescent beliefs and behaviors in the domains of reading, math, art, and athletics across childhood and adolescence. Mothers who displayed relevant beliefs, such as valuing the domain and their child’s competence in the domain, as well as demonstrated relevant behaviors themselves, such as modeling and encouragement, had adolescents who valued and engaged more in each domain ( Simpkins, Fredricks, & Eccles, 2012 ). Collectively, these studies show the power of parental influence on adolescent development through involvement, closeness, and displaying positive beliefs and behaviors. Clearly, parents continue to impact their children’s decisions across adolescence through parental values and parent-child conversations about the adolescent’s friends, whereabouts, and daily lives.

Sibling influence.

Recently, there has been a surge in research examining sibling relationships due to their salient influence on adolescent health and well-being ( Conger, 2013 ). Sibling influences can be especially impactful during developmental transitions ( Cox, 2010 ), helping adolescents navigate new roles and adjust to social and physical changes ( Eccles, 1999 ). Siblings primarily influence each other through two mechanisms: social learning, which is the process of observing and selectively integrating modeled behaviors, and through deidentification, which is the process of actively behaving differently from one another ( Whiteman, Beccera, & Killoren, 2009 ). However, these mechanisms largely depend on one factor—perceptions of support (for a review see, Dirks, Persram, Recchia, & Howe, 2015 ).

Although research on sibling relationships has traditionally focused on conflict and rivalry as it contributes to negative child and adolescent outcomes, accumulating research suggests that siblings positively influence adolescent development through sibling relationships built upon support ( Conger, 2013 ). Adolescents who perceive general closeness and academic support from siblings are more likely to report positive school attitudes and high academic motivation ( Alfaro & Umaña-Taylor, 2010 ; Milevsky & Levitt, 2005 ). In addition, experiencing support from a sibling is associated with later feelings of competence, autonomy, and relatedness during adolescence, as well as life satisfaction during the transition into emerging adulthood ( Hollifield & Conger, 2015 ). Further, in the face of stressful life events, perceived affection and closeness from a sibling can buffer against the progression of internalizing behaviors across adolescence ( Buist et al., 2014 ; Gass, Jenkins, & Dunn, 2007 ). These are just a sample of studies that highlight how powerful sibling relationships can be in socializing adolescents toward prosocial behavior and maintaining well-being. Future work should tap into siblings as a natural resource to bolster positive adolescent development.

The influence of multiple family members.

Although we have reviewed literature examining one parent or one sibling, research has also investigated the combined influence of multiple family members, which reflects the essence of the family systems model ( Cox, 2010 ). Parental and sibling influences are intertwined in adolescent’s daily lives, and thus, are important to investigate together to better inform our understanding of positive adolescent development ( Tucker & Updegraff, 2009 ). Mothers, fathers, and siblings can all contribute to adolescent psychosocial adjustment by providing supervision, acceptance, and opportunities for autonomy ( Kurdek & Fine, 1995 ). For example, high levels of parental involvement and high levels of sibling companionship are associated with lower substance use during adolescence ( Samek, Rueter, Keyes, McGue, & Ianoco, 2015 ). In addition, both observed parental support, and sibling-reported sibling relationship quality, positively contribute to academic engagement during adolescence, and educational attainment in emerging adulthood ( Melby, Conger, Fang, Wickrama, & Conger, 2008 ). Parents and siblings can also work together to buffer adolescents against negative life events. One study found that for adolescent victims of bullying who also experienced low parental conflict and low sibling victimization, boys reported lower levels of depression and girls reported lower levels of delinquency compared to adolescents who experienced high dissatisfaction at home ( Sapouna & Wolke, 2013 ). Moreover, sometimes siblings can provide support when parents come up short. Older siblings can buffer the negative effect of hostile parental behaviors on adolescent externalizing behavior by providing younger siblings with a warm and supportive relationship ( Conger, Conger, & Elder, 1994 ). Together, these studies suggest that adolescent development is heavily influenced by the family context, and by each family member. The social susceptibility and flexibility present during adolescence allows teens to benefit from the influence of multiple family members, even when one source of family influence is compromised. Thus, both parents and siblings need to be examined together to better inform our understanding of how the family can positively influence adolescent decision-making and well-being, including both the nature of influence (e.g., support, involvement) and the degree to which the influence is present (e.g., absent versus helicopter parenting).

C. Family and peer influence on positive adolescent development

Despite extensive research examining how family and peers uniquely influence a wide range of adolescent behaviors, less is known about how these sources of influence simultaneously guide adolescent decision making in positive ways. Indeed, adolescents often face the need to reconcile potential differences in the attitudes and behaviors endorsed by their family relative to peers. Extant research examining social conformity across development supports the reference group theory ( Shibutani, 1955 ), which suggests that individuals adopt the perspectives of different social reference groups (e.g., family or peers) based on their perceived relevance in guiding that decision. Using this theoretical framework, we review literature examining the social contexts in which adolescents rely more on their family or peer influence when faced with conflicting information, which can, in turn, reinforce the development of positive social norms and relationships, as well as promote adaptive decision making.

Susceptibility to social conformity.

Susceptibility to parent versus peer pressures changes with age, resulting in different rates of social conformity across development. One of the earliest methods used to explore how family and peer influence interact and contribute to positive adolescent behaviors was cross-pressures tests, where adolescents respond to hypothetical situations in which their parent and/or peers suggest conflicting actions. From childhood to adolescence, there is a general increase in the tendency for youth to conform to the perspectives of their peers when parents and peers offer conflicting advice ( Utech & Hoving, 1969 ). This supports other work showing the social value of peers is also increasing with age ( Bandura & Kupers, 1964 ), suggesting that, relative to parents, peers may be more successful at reinforcing certain norms or behaviors across development. Consistent with social learning and social identity theories, these results suggest that over the course of adolescence, youth may be shifting their attitudes to align with whichever reference group (e.g., parents or peers) is more salient (i.e., social identity), whose norms may become differentially reinforced over time (i.e., social learning). However, distinct developmental trajectories emerge when adolescents are evaluating different types of behaviors. For example, one study examined parent and peer conformity to prosocial behaviors and found both parent and peer conformity to prosocial behaviors declined from childhood to adolescence (albeit results for peer conformity to prosocial behaviors are inconsistent) ( Berndt, 1979 ). The fact that youth are conforming to their parent or peer influence less often in considering prosocial actions illustrates their increasing ability to make positive decisions independently with age, without the need for a reference group. Not only do these findings suggest that youth seek guidance from parents or peers differently based on the type of behavior under consideration, but they also highlight childhood and early adolescence as an important developmental transition for promoting positive social influence, either by parents or peers.

Flexibility of norms and behaviors.

Evidence from qualitative interview studies demonstrates the flexibility and potential mechanisms by which interacting sources of social influence shape youth’s norms and behaviors. The degree to which parent or peer pressures impact adolescent decision making varies systematically across domains, such that adolescents are more likely to seek guidance about future- or career-oriented topics (e.g., applying for college) from parents and about status- or identity-related topics (e.g., attending social events) from peers ( Biddle, Bank, & Marlin, 1980 ; Brittain, 1963 ; Sebald & White, 1980 ). Interestingly, adolescents rely more heavily on parents’ advice when their choices are perceived to be more difficult, such as in situations involving ethical or legal concerns (e.g., reporting a peer’s crime; Brittain, 1963 ). Another study examined the relative impact of parent and peer norms (e.g., do your parents/peers think you should/shouldn’t do well in school?) versus behaviors (e.g., did your parents/peers do well in school?) on adolescents’ own norms and behaviors as it related to school achievement and alcohol use ( Biddle et al., 1980 ). Adolescents’ alcohol use was more strongly influenced by peers’ behaviors, whereas school achievement was more strongly influenced by parental norms ( Biddle et al., 1980 ). While adolescents can adapt to parent and peer pressures under the appropriate circumstances (e.g., different domains), the extent to which adolescents internalize those pressures—insofar that parent/peer pressures are adopted as adolescents’ own norms—may determine whether those pressures result in more positive or negative decisions.

Parents often influence their adolescents’ peer group affiliations, which also affects the strength and type of norms and behaviors that youth are exposed to. Positive parenting practices lead youth to engage in more adaptive behaviors (e.g., academic achievement), which, in turn, promote affiliation with better peer groups (e.g., “populars” over “druggies;” Brown, Mounts, Lamborn, & Steinberg, 1993 ). In fact, peer pressures are generally stronger within positive domains (e.g., school achievement) compared to negative domains (e.g., misconduct), especially among social groups that are well interconnected (i.e., less alienated) within the school structure ( Clasen & Brown, 1987 ). These studies highlight the significant role that parents can play in promoting prosocial peer affiliations, which may subsequently facilitate opportunities for peers to positively influence youth’s decision making.

Protective role of positive relationships.

In addition to promoting prosocial peer affiliations, positive social figures can buffer adolescents against negative social pressures over time. Positive family influence can attenuate the potentially negative impact of peers on adolescents’ well-being. Indeed, warm family relationships and environments promote resilience to peer bullying ( Bowes, Maughan, Caspi, Moffitt, & Arseneault, 2010 ) and mitigate the effect of peer pressure on alcohol use ( Nash, Mcqueen, & Bray, 2005 ) among youth. As peers become increasingly important across adolescence, positive peer influence can similarly protect against aversive family experiences. For example, family adversity (e.g., harsh discipline) is not associated with child externalizing behaviors for youth with high levels of positive peer relationships ( Criss, Pettit, Bates, Dodge, & Lapp, 2002 ). This highlights the potential of strong peer support in redirecting negative developmental trajectories, particularly among vulnerable youth.

In some cases, peers may serve as a stronger buffer against poor developmental outcomes than parents. One study examined how the perceived expectations of mothers and friends influenced adolescents’ engagement in antisocial and prosocial behaviors ( Padilla-Walker & Carlo, 2007 ). Adolescents indicated how strongly they personally agreed with the importance of engaging in several prosocial behaviors (e.g., helping people), as well as rated how much they felt their mother versus friends expected them to engage in these same prosocial behaviors. Adolescent boys who perceived their peers to have stronger expectations of their prosocial engagement actually participated in fewer antisocial behaviors; there was no effect of maternal expectations or personal values on their antisocial behaviors. Thus, positive peer influence may be more protective against antisocial behaviors for adolescent boys relative to girls. In contrast, both the perceived expectations of mothers and friends were related to adolescents’ personal prosocial values, which subsequently influenced their prosocial behaviors. Although peers may be a stronger protective factor against negative behaviors compared to family, adolescents rely on the social norms of both their family and peers to inform their own values and choices about engaging in more adaptive, positive behaviors (a la social identity theory). In the following sections, we review prominent neurobiological theories, which describe how heightened social influence susceptibility during adolescence may reflect maturational changes in how the brain responds to social information.

IV. Neurobiological Models of Adolescents’ Social Influence Susceptibility

Often described as a car in full throttle with ineffective brakes, the adolescent brain was originally thought to be defective in some way (see Payne, 2012 ). However, based on functional and structural magnetic resonance imaging (MRI) research, we now know that the teenage brain is rapidly changing and adapting to its environment in ways that promote skill acquisition, learning, and social growth (see Telzer, 2016 ). Indeed, the adolescent period is marked by dramatic changes in brain development, second only to that seen in infancy. Such changes in the brain uniquely sensitize adolescents to social stimuli in their environment, and may underlie social influence susceptibility – for better or for worse.

Social influence susceptibility may reflect a (1) heightened orientation to social cues, (2) greater sensitivity to social rewards and punishments, and (3) compromised cognitive control. Indeed, adolescence is characterized by changes in neural circuitry underlying each of these processes (see Figure 1 ). For instance, complex social behaviors, including the ability to think about others’ mental states such as their thoughts and feelings, to reason about others’ mental states to inform one’s own behaviors, and to predict what another person will do next during a social interaction ( Frith & Frith, 2007 ; Blakemore, 2008 ) involve the recruitment of brain regions including the temporoparietal junction (TPJ), posterior superior temporal sulcus (pSTS), and the dorsomedial prefrontal cortex (DMPFC). Moreover, the medial prefrontal cortex (MPFC) is involved in thinking about the self and close others ( Kelley et al., 2002 ; Johnson et al., 2002 ). These brain regions tend to be more activated among adolescents relative to adults when processing social information ( Blakemore, den Ouden, Choudhury, & Frith, 2007 ; Burnett, Bird, Moll, Frith, & Blakemore, 2009 ; Gunther Moor et al., 2012 ; Pfeifer et al., 2009 ; Van den Bos, Van Dijk, Westenberg, Rombouts, & Crone, 2011 ; Wang, Lee, Sigman, & Dapretto, 2006 ; Somerville et al., 2013 ), underscoring adolescence as a key period of social sensitivity ( Blakemore, 2008 ; Blakemore & Mills, 2014 ).

Neural regions involved in social cognition (yellow), cognitive control (blue), and affective processing (red).

Brain regions involved in affective processing include the ventral striatum (VS), which is implicated in reward processing, including the receipt and anticipation of primary and secondary rewards ( Delgado, 2007 ), the orbitofrontal cortex (OFC), which is involved in the valuation of rewards and hedonic experiences ( Saez et al., 2017 ; Kringelbach, 2005 ), and the amygdala, which is involved in detecting salient cues in the environment, responding to punishments, and is activated to both negative and positive emotional stimuli ( Hamann, Ely, Hoffman, & Kilts, 2002 ). Compared to children and adults, adolescents show heightened sensitivity to rewards in the VS ( Galvan et al., 2006 ; Ernst et al., 2006 ; Eshel et al., 2007 ), particularly in the presence of peers ( Chein et al., 2010 ). Adolescents also show heightened VS and amygdala activation to socially appetitive stimuli ( Perino et al., 2016 ; Somerville et al., 2011 ). Thus, adolescents may be uniquely attuned to salient social rewards in their environment.

Finally, brain regions involved in regulatory processes include lateral and medial areas of the prefrontal cortex (e.g., VLPFC, DLPFC, MPFC, ACC). These regions are broadly involved in cognitive control, emotion regulation, goal directed inhibitory control, and serve as a neural brake system ( Wessel et al., 2013 ). Both age-related increases and decreases in PFC activity have been reported across development, such that some studies find that adolescents show heightened PFC activation compared to adults, whereas other studies report adolescent suppression of the PFC ( Bunge et al., 2002 ; Booth et al., 2003 ; Durston et al., 2006 ; Marsh et al., 2006 ; Rubia et al., 2007 ; Velanova et al., 2009 ). Such discrepant developmental patterns of activation have been theorized to underlie flexibility and learning, promoting exploratory behavior in adolescence (see Crone and Dahl, 2012 ).

Based on emerging developmental cognitive neuroscience research, many theoretical models have been proposed to describe adolescents’ neurobiological sensitivity to social context (see Schriber & Guyer, 2016 ). While several of these models explain neural changes that underlie vulnerabilities during adolescence (e.g., heightened risk taking and psychopathology; Casey, Jones, & Hare, 2008 ; Steinberg, 2008 ; Ernst, Pine, & Hardin, 2006 ), these models can be useful heuristics for broadly describing adolescent brain development and social sensitivity, as well as opportunities for positive adjustment (but see Pfeifer & Allen, 2012 , 2016 , for why these models are too simplified).

A. Imbalance Model

The Imbalance Model ( Somerville, Jones, & Casey, 2010 ; Casey et al., 2008 ) proposes that the subcortical network, comprising neural regions associated with the valuation of rewards (e.g., ventral striatum (VS)), matures relatively early, leading to increased reward seeking during adolescence, whereas the cortical network, comprising neural regions involved in higher order cognition and impulse control (e.g., ventral and dorsal lateral prefrontal cortices (VLPFC, DLPFC)), gradually matures over adolescence and into adulthood. The differential rates of maturation in the cognitive control and affective systems creates a neurobiological imbalance during adolescence, which is thought to bias adolescents towards socioemotionally salient and rewarding contexts during a developmental period when they are unable to effectively regulate their behavior (see Figure 2 ).

Imbalance Model of adolescent brain development. Earlier developmental of affective, reward-related activation (red line) and relatively later and more protracted development of cognitive control (blue line) result in a neurobiological imbalance during adolescence (depicted by the grey box).

B. Dual Systems Model

The Dual Systems Model discusses a balance between “hot” and “cool” systems ( Metcalfe & Mischel, 1999 ). The cool system focuses on the cognitive control system, which is emotionally neutral, rational, and strategic, allowing for flexible, goal-directed behaviors, whereas the hot system focuses on the emotional system, which is emotionally reactive and driven by desires (see Casey, 2015 ). During adolescence, the hot system is overactive, and the cool system is not yet fully mature. Similar to the Imbalance Model, the Dual Systems Model describes relatively early and rapid developmental increases in the brain’s socioemotional “hot” system (e.g., VS, amygdala, orbitofrontal cortex) that leads to increased reward- and sensation-seeking in adolescence, coupled with more gradual and later development of the brain’s cognitive control “cool” system (e.g., lateral PFC) that does not reach maturity until the late 20s or even early 30s ( Steinberg, 2008 ; Shulman et al., 2016 ). The temporal gap between these systems is thought to create a developmental window of vulnerability in adolescence during which youth may be highly susceptible to peer influence due to the socioemotional nature of peer contexts ( Steinberg, 2008 ). Although children still have relatively immature cognitive control, they do not yet evidence this heightened orientation towards reward-driven behaviors, and adults have relative maturity of cognitive control and strengthened connectivity across brain networks that facilitate top-down regulation of reward-driven activation. Therefore, the temporal gap between affective and regulatory development is only present in adolescence (see Figure 3 ).

Dual Systems Model of adolescent brain development. (A) Adolescence is characterized by hyperactivation of the “hot” socioemotional system (red circle) coupled with later developing cognitive control (blue circle), and immature connectivity (dotted line) between systems, resulting in an ability to engage in effective regulation. (B) Childhood is characterized by not yet maturing “hot” or “cold” systems, whereas adulthood is characterized by mature “hot” and “cold” systems, coupled with effective connectivity (double arrow) between systems.

C. Triadic Neural Systems Model

The Triadic Neural Systems Model includes the cognitive control system as well as two affective systems, an approach, reward-driven system, which centers on the VS, and an avoidance/emotion system, which centers on the amygdala, a brain region involved in withdrawal from aversive cues and avoidance of punishments ( Ernst, 2014 ). Whereas the VS supports reward processes and approach behavior, the amygdala serves as a “behavioral brake” to avoid potential harm ( Amaral, 2002 ), and the PFC serves to orchestrate the relative contributions of the approach and avoidance systems (see Ernst et al., 2006 ). The balance between reward-driven behaviors and harm-avoidant behaviors is tilted, such that adolescents are more oriented to rewards and less sensitive to potential harms, and the immature regulatory system fails to adaptively balance the two affective systems (see Figure 4 ). Thus, adolescents will be more likely to approach, but not avoid, risky and potentially harmful situations, whereas adults’ more mature regulatory system effectively balances approach and avoidance behaviors, thereby decreasing the likelihood of risk behaviors.

Triadic Systems Model of adolescent neurodevelopment. (A) Adolescents show heightened approach behaviors (ventral striatum), are less sensitivity to harm (amygdala), and have an immature regulatory system (PFC) that does not effectively balance the approach and avoidance systems. (B) Adults have mature regulatory capabilities that effectively balance the approach and avoidance systems.

D. Social Information Processing Network

The Social Information Processing Network model (SIPN; Nelson et al., 2005 ; Nelson, Jarcho, & Guyer, 2016 ) proposes that social stimuli are processed by three nodes in sequential order. The detection node first categorizes a stimulus as social and detects its basic social properties. This node includes regions such as the superior temporal sulcus (STS), intraparietal sulcus, fusiform face area, temporal pole, and occipital cortical regions. After a stimulus has been identified, it is processed by the affective node , which codes for rewards and punishments and determines whether stimuli should be approached or avoided. This node includes regions such as the amygdala, VS, and orbitofrontal cortex. Finally, social stimuli are processed in the cognitive-regulatory node , which performs complex cognitive processing, including theory of mind (i.e., mental state reasoning), cognitive inhibition, and goal-directed behaviors. This node includes regions such as the medial prefrontal cortex (MPFC) and dorsal and ventral prefrontal cortices. These three nodes function as an interactive network, largely in a unidirectional way, from detection to affective to cognitive, but there are also bidirectional pathways. Similar to all of the models discussed above, the affective node is particularly reactive and sensitive during adolescence, whereas the cognitive-regulatory node shows more protracted development into adulthood. Each of the models discussed so far suggest that differential neural development and overreliance on subcortical, reward-related regions drives adolescents to seek out (social) rewards in their environment at a developmental period when self-control is still maturing. While social contexts may tip the balance in terms of affective and cognitive control-related activation, these models do not take into consideration neural regions that specifically code for higher-order social cognition.

E. Neurobiological Susceptibility to Social Context Framework

Perhaps the most promising model for understanding adolescents’ susceptibility to social influence, particularly in regards to positive social influence, stems from the Neurobiological Susceptibility to Social Context Framework ( Schriber & Guyer, 2016 ), which is based on other theoretical frameworks including biological sensitivity to context ( Boyce & Ellis, 2005 ) and differential susceptibility to environmental influences ( Belsky & Pluess, 2009 ). This model proposes that individuals vary in their sensitivity to the social environment as a function of biological factors, particularly neural sensitivity to social contexts. While specific neural biomarkers are not specified, Schriber and Guyer (2016) build on the existing models of brain development discussed above to suggest that adolescents with high neurobiological susceptibility can be pushed in a for-better or for-worse fashion, depending on their social environment ( Figure 5 ). In particular, individuals who are not highly sensitive will not be affected by either positive or aversive social environments, whereas highly sensitive individuals will be both more vulnerable to aversive contexts (e.g., negative peer influence effects), but also more responsive to salubrious contexts (e.g., positive peer influence effects). In other words, those who have supportive peers and family will thrive, whereas those who face family or peer rejection will be most vulnerable.

Neurobiological susceptibility to social influence model. Adolescents with high neurobiological susceptibility (blue dashed line) thrive in positive contexts but are vulnerable in negative contexts.

V. Neural Correlates of Peer and Family Influence

While current neurobiological models or cognitive neuroscience research have yet to clearly connect how social influence processes (e.g., social learning theory, social identity theory) map onto neurobiological development, emerging research has begun to highlight how peer and family contexts influence adolescent neurodevelopment. These studies highlight a set of neural candidates to examine as promising indices of adolescents’ susceptibility to social influence. In particular, neural regions involved in (1) affective processing of social rewards and punishments (e.g.,VS, amygdala), (2) social-cognition and thinking about others’ mental states (e.g.,TPJ, MPFC), and (3) cognitive control that facilitates behavioral inhibition (e.g., VLPFC, anterior cingulate cortex (ACC)) show sensitivity to peer and family contexts (see Figure 1 ). Below we review recent research unpacking the neurobiological correlates of peer and family influence, highlighting studies that focus on positive social influence.

A. Peer Relationships and Neurobiological Development in Adolescence

Prior research has largely focused on the supposed monolithic negative influence of peers (e.g., deviancy training) at both the behavioral (e.g., Dishion et al., 1996 ) and neural level ( Chein, Albert, O’Brien, Uckert, & Steinberg, 2011 ). This research supports the widely held notion that adolescents are more likely to take risks in the presence of their peers, and this is modulated by heightened ventral striatum activation, suggesting that peers increase the salient and rewarding nature of taking risks ( Chein et al., 2011 ). However, it is essential to also examine positive peer influences. If adolescents are highly sensitive to peer influence due to heightened neurobiological sensitivity to social context, then in addition to being pushed to engage in negative behaviors (e.g., risk taking), peers should be able to push teens to engage in more positive behaviors (e.g., prosocial behaviors).

Positive peer influence.

In a recent neuroimaging study, we examined whether peer presence and positive feedback affected adolescents’ prosocial behaviors (donation of tokens to their group in a public goods game) and associated neural processing ( Van Hoorn, Van Dijk, Güroğlu, & Crone, 2016 ). Adolescents donated significantly more to a public goods group when they were being observed by their peers, and even more so when receiving positive feedback (i.e., thumbs up) from their peers. Prosocial decision-making in the presence of peers was associated with enhanced activity in several social brain regions, including the dorsal medial prefrontal cortex (dmPFC), TPJ, precuneus and STS. Effects in the dmPFC were more pronounced in early adolescents (12–13 year olds) than mid-adolescents (15–16 year olds), suggesting that early adolescence may be a window of opportunity for prosocial peer influence. Interestingly, these findings revealed that social brain regions, rather than affective reward-related regions, underlie prosocial peer influence. These findings underscore early adolescents as particularly sensitive to social influence, but in a way that promotes positive, prosocial behavior.

Researchers have also examined how the context of risk-promoting or risk-averse social norms affects adolescents’ risk taking. In a recent study, researchers had adolescents complete a cognitive control task during an fMRI scan, and used a “brain as predictor of behavior” approach to test how the neural correlates of cognitive control affect adolescents’ conformity to peer influence ( Cascio et al., 2015 ). One week following the scan, adolescents returned to the lab to undergo a simulated driving session in the presence of either a high- (e.g., indicating their driving behavior is more risky than the participant) or low- (e.g., indicating their behavior is less risky and more cautious than the participant) risk-promoting peer. Adolescents made fewer risky choices in the presence of low-risk peers compared to high-risk peers. At the neural level, adolescents who recruited regions involved in cognitive control(e.g., lateral PFC ) during the cognitive control task were more influenced by their cautious peers, such that cognitive control-related activation was associated with safer driving in the presence of cautious peers. Such activation was not associated with being influenced by risky peers or driving behavior when alone. Engagement of the PFC during the cognitive control task may represent a neurobiological marker for more thoughtful and deliberative thinking, allowing adolescents to override the tendency to be risky and instead conform to their more cautious peers’ behavior. This study highlights that social influence susceptibility may be a regulated process as opposed to a lack of inhibition, and also points to the positive side of peer conformity.

Supportive peer friendships.

In addition to examining how peers may influence adolescents to engage in more positive behaviors, researchers have examined the role of supportive peer friendships in buffering adolescents from negative outcomes. The need for social connection and peer acceptance is one of the most fundamental and universal human needs ( Baumeister & Leary, 1995 ). As peer relationships increase in importance during adolescence, close friendships become their primary source of social support ( Furman & Buhrmester, 1992 ). When adolescents do not feel socially connected, it poses serious threats to their well-being. Fortunately, social connection and close friendships can buffer adolescents from the distress associated with negative peer relations. In a recent study, we tested the stress-buffering model of social relationships ( Cohen, Gottlieb, & Underwood, 2001 ) to examine whether supportive peer relationships can attenuate the negative implications of chronic peer conflict ( Telzer, Fuligni, Lieberman, Miernicki, & Galvan, 2015 ). Adolescents reporting chronic peer conflict engaged in more risk-taking behavior, and at the neural level, showed increased activation in the ventral striatum when making risky choices. But those adolescents reporting high peer support were completely buffered from these effects – those experiencing high peer conflict did not engage in more risk taking or show heightened ventral striatum activation during risky decisions when they had a close friend. These findings highlight the vital role that supportive friends play. Even in the face of peer conflict, having a close friend can provide the means to feel connected to a social group and receive emotional support and guidance, which may provide them with a means of coping with stress.

B. Family Relationships and Neurobiological Development in Adolescence

In addition to investigating the role of peers on positive adolescent adjustment, developmental social neuroscientists have also examined the influence of the family. In the following section, we review neuroimaging work on how the family context contributes to adolescent adjustment through family norms and values, positive family relationships, and parental monitoring.

Familial norms and values.

One way researchers have examined familial influence on positive youth adjustment and brain development is to examine the internalization of family values. Often referred to as “familism” or “family obligation,” youth from Latin American families, for example, stress the importance of spending time with the family, high family unity, family social support, and interdependence for daily activities ( Cuellar, Arnold, & Maldonado, 1995 ; Fuligni, 2001 ). The internalization of strong family obligation values is associated with lower rates of substance use ( Telzer, Gonzales, & Fuligni, 2014 ) and depression ( Telzer, Tsai, Gonzales, & Fuligni, 2015 ) in Mexican-American adolescents, underscoring family obligation as an important cultural resource. At the neural level, we found that higher family obligation values were associated with greater activation in the DLPFC during a cognitive control task, which was associated with better decision making skills ( Telzer, Fuligni, Lieberman, & Galvan, 2013a ), suggesting that by putting their family’s needs first and delaying personal gratification for their family, youth may develop more effective cognitive control, helping them to avoid the impulse to engage in risky behaviors. In addition, higher family obligation values were associated with lower activation in the VS during a risk-taking task, which was associated with less self-reported risk-taking behavior ( Telzer et al., 2013a ). Youth with stronger family obligation values report more negative consequences for engaging in risk taking, as it may reflect poorly upon their family ( German, Gonzales, & Dumka, 2009 ). Thus, risk taking itself may become less rewarding, as evidenced by dampened VS activation.

We also examined whether the rewarding and meaningful nature of family obligation itself offsets the rewards of risk taking. First, we found that engaging in family obligation (i.e., making decisions that benefit the family) recruits the VS, even more so than gaining a personal reward for the self, suggesting that decisions to make sacrifices for the family are personally meaningful and rewarding ( Telzer, Fuligni, & Galvan, 2016 ). Secondly, we correlated VS activation during the family obligation task with VS activation during the risk-taking task described above. Adolescents who had heightened VS activation during the family obligation task showed less activation in the same brain region during the risk-taking task, suggesting that the rewarding nature of family obligation may make risk taking comparatively less rewarding ( Telzer et al., 2016 ). Importantly, increased activation in the VS during the family obligation task predicted longitudinal declines in risky behaviors and depression, whereas increased VS activation during the risk taking task predicted increases in psychopathology ( Telzer, Fuligni, Lieberman, & Galvan, 2013b , Telzer et al., 2015 ). Thus, finding meaning in social, other-focused behaviors (i.e., family obligation) can promote positive youth adjustment, whereas being oriented towards more self-focused behaviors (i.e., risk taking) is a vulnerability. Together, these findings suggest that the internalization of important family values is rewarding and meaningful, buffering adolescents from both risk taking and depression.

Positive family relationships.

Besides family values, the quality of family relationships also influences adolescents’ positive adjustment – high family support and cohesion and low conflict are associated with a host of positive outcomes, including better school performance, lower substance use, and lower internalizing symptoms ( Melby et al., 2008 ; Samek et al., 2015 ; Telzer & Fuligni, 2013 ). According to social control theories, adolescents who are close to their parents feel obligated to act in non-deviant ways, whereas adolescents in conflictual families do not feel obligated to conform to their parents’ expectations and will be more likely to engage in risky behaviors ( Bahr et al., 2005 ). Thus, strong family relationship quality can buffer adolescents from risk taking, perhaps by making risk taking less rewarding. In one longitudinal fMRI study, we examined changes in the quality of family relationships, paying particular attention to three dimensions of positive family interactions: high parental support (e.g., their parents respected their feelings), adolescents’ spontaneous disclosure (e.g., telling their parents about their friends), and low family conflict (e.g., having a fight or argument with their parents). Adolescents who reported improvements in the quality of their family interactions showed longitudinal declines in risk taking, which was mediated by declines in VS activation during a risk-taking task (Qu, Fuligni, Galvan, Lieberman, & Telzer, 2015). This study suggests that increases in positive family relationships may provide adolescents with a supportive environment, increasing their desire to follow their parents’ expectations, which may dampen their subjective sensitivity to rewards during risk taking. In addition to examining cohesion and conflict, this study assessed adolescents’ disclosure to their parents. Given that adolescents spend increasingly less time with their parents than do children ( Lam, McHale, & Crouter, 2012 ; 2014 ), voluntary disclosure of their activities may provide opportunities for parents to give their children advice and supervision, helping them develop the skills to avoid risks and devalue the rewarding nature of risk taking.

Parental monitoring.

In addition to adolescents’ spontaneous disclosure, parental monitoring plays a key influence on adolescents’ decisions to avoid deviant behaviors. Yet, during the adolescent years, parents tend to decrease their supervision of their children, and adolescents are more likely to make maladaptive decisions during unsupervised time or in the presence of their peers ( Richardson, Radziszewska, Dent, & Flay, 1993 ; Beck, Shattuck, & Raleigh, 2001 ; Borawski, Ievers-Landis, Lovegreen, & Trapl, 2003 ). In a recent study, we tested how the presence of parents changes the way adolescents make decisions in a risky context. During an fMRI scan, adolescents played a risky driving game twice: once alone, and once with their mother present and watching. Whereas adolescents take greater risks when their friends are watching them during this same task ( Chein et al., 2011 ), we found that adolescents made significantly fewer risks when their mother was present ( Telzer, Ichien, & Qu, 2015 ).

At the neural level, the presence of friends is associated with more VS activation (Chein et al. 2010), whereas the presence of mothers is associated with less VS activation when making risky choices ( Telzer, Ichien, & Qu, 2015 ). Importantly, this protective role is specific to mothers, as we did not find the same decrease in risk taking or ventral striatum activation when an unknown adult was present (Guassi Moreira & Telzer, in press). Together, these findings suggest that peers may increase the rewarding nature of risk taking, whereas mothers may take the fun away. In addition, neural regions involved in cognitive control (e.g., VLPFC, MPFC), were more activated when their mother was present than when alone or in the presence of an unknown adult, suggesting that maternal presence may facilitate more mature and effective neural regulation via top-down inhibitory control from prefrontal regions. Finally, after making a risky decision, adolescents recruited regions involved in mentalizing (e.g., TPJ) more when their mother was present than an unknown adult, suggesting that adolescents are more sensitive to their mother’s perspective following a brief instance of misbehavior (i.e., running the yellow light). Together, these findings suggest that the presence of mothers alters the way adolescents make risky decisions and may provide an important scaffolding role, helping adolescents avoid risks by decreasing the rewarding nature of risks and promoting more effective cognitive control.

C. Simultaneous Role of Family and Peer Relationships on Adolescent Brain Development

Although few neuroimaging studies have examined the simultaneous influence of family and peers on adolescent development, there is emerging evidence suggesting that adolescents’ choices are affected, in part, by differential neural sensitivity to family versus peers. In order to capture how behavior and brain function change in the context of family and peers, researchers have mainly examined within-person differences between decisions that affect a family member (primarily parents) compared to decisions that affect peers. In addition, novel research designs have recently stimulated investigations of the simultaneous influence of both parent and peer influence on adolescent decision making, which are also discussed in this section.

Emotional reactivity to peers and parents.

Prior research consistently characterizes adolescence as a time of social reorientation from parent to peer influences, a process thought to be supported by developmental changes within several affective and social cognitive brain regions ( Nelson et al., 2005 ). However, only recently has research emerged showing that this social reorientation at the behavioral level is paralleled by functional changes at the neural level, such that simply processing peer versus parent faces elicits different neural responses in regions involved in socioemotional processing during adolescence. In a study examining adolescents’ emotion perception of their mother’s, father’s and an unknown peer’s faces, adolescents exhibited greater activation in regions implicated in social (PCC, pSTS, TPJ) and affective (VS, amygdala, hippocampus) processing when viewing their peer relative to parent faces (no difference between processing maternal or paternal stimuli; Saxbe, Del Piero, Immordino-Yang, Kaplan, & Margolin, 2015 ). This illustrates that the neural correlates underlying socioemotional processing change over the course of adolescence as the salience of peers increases relative to family. Moreover, although adolescents, on average, showed greater activation in the PCC and precuneus to peer versus parent faces, those who showed less of this effect (i.e., did not show greater activation in these regions to peer over parent faces) engaged in lower levels of risk-taking behaviors and affiliation with deviant peers. Thus, less recruitment of regions involved in social cognition (e.g., mentalizing) toward peers relative to parents may help to diminish the social value of peer influence on negative behaviors during adolescence.

Vicarious rewards for peers and parents.

Differential neural sensitivity to peers versus parents can be leveraged to promote adaptive decision making during adolescence, specifically by encouraging vicarious learning about other-oriented behaviors. Even in the absence of a personally experienced reward, the act of seeing or imagining others experience rewards (i.e., vicarious rewards) elicits activation in reward-related regions (VS) and promotes prosocial motivations ( Mobbs et al., 2009 ). Given the heightened salience of peer and parent influence during adolescence, it is important to explore whether exposure to vicarious rewards that affect close others might reinforce positive choices. Vicarious learning, especially through observing the positive behaviors and outcomes of close others, can facilitate the internalization of positive social norms and increase motivation to model similar behaviors in the future, which is consistent with social learning theory. A recent study examined VS activation during a risk-taking task, where the potential gains and losses could affect adolescents’ mothers or best friends ( Braams & Crone, 2016 ). Striatal activation peaked in adolescence compared to childhood and young adulthood when youth took risks to win money for their mothers, but not for their peers. Self-report data further demonstrated a positive association between relationship quality and the extent to which adolescents enjoyed taking risks to win money for both their mothers and best friends. Therefore, developmental changes in reward sensitivity and relationship quality can affect adolescents’ motivation to engage in risky behaviors that affect others over time. Indeed, a new perspective from developmental neuroscience proposes that, in some contexts, adolescents may be taking risks with the explicit intention of helping others ( Do et al., 2017 ), a process that may be supported by neural reactivity in reward-related regions to the experience of vicarious rewards for close others.

Balancing conflicting social influence from peers and parents.

A common feature of adolescent decision making is the balance of conflicting social information from parents and peers. This is an important area of inquiry, as peer and family values and norms often differ, resulting in norm conflicts that inevitably affect adolescent decision making and beg for reconciliation. In one of the first developmental studies to examine the neural correlates of both parental and peer influence on attitude change, we first asked adolescents, their primary caregiver, and several peers from their schools to each independently evaluate artwork stimuli prior to their scan ( Welborn et al., 2015 ). Artwork was selected, as it tends to be neutral stimuli where attitudes may be swayed by influence. Adolescents completed an fMRI session a few weeks later, where they were shown their parents’ or peers’ real evaluations of the same pieces of artwork before re-evaluating the stimuli. Adolescents were more likely to change their own attitudes to bring them in line with those of their parents compared to their peers. At the neural level, adolescents exhibited greater activation in regions involved in mentalizing (TPJ, precuneus), reward processing (ventral medial prefrontal cortex, VMPFC), and self-control (VLPFC) when they were influenced by both their peers and parent, with no difference between the source of social influence. Moreover, greater activation in these task-responsive regions predicted a greater likelihood for youth to shift their attitudes in favor of the corresponding source of influence. Thus, although family and peers influence adolescents through similar neural mechanisms (involved in mentalizing, reward processing, and regulation), individual differences in this neurobiological sensitivity might differentially predict adolescents’ tendency to adopt the attitudes and/or behaviors of their family or peers.

While prior research has examined neural differences between social influence from family and peers, no study to date has delineated how youth incorporate the simultaneous influence of their family and peers into their decisions and behaviors. When there is a discrepancy between peers’ and parents’ attitudes about a behavior, adolescents often need to simultaneously weigh the relative value of these conflicting attitudes when deciding whether to personally endorse that behavior, which may differ depending on if it is positive or negative. Over time, their decision to conform to the attitudes of one influence over the other can have important implications for reinforcing their participation in those behaviors. For example, an adolescent who endorses drug use as a means of conforming to the attitudes favored by their peers, but is discouraged by their parents, may be more likely to do drugs over time. We recently examined this process in an fMRI study, where we showed adolescents their parents’ and peers’ evaluations of positive and negative behaviors at the same time, each of which differed from each other and were manipulated to conflict with adolescents’ initial evaluations (Do, McCormick, & Telzer, unpublished data). To measure the extent to which adolescents were affected by conflicting social information, adolescents indicated whether they agreed with their parent or peers’ evaluations of each behavior. On average, adolescents showed differences in neural activation within affective and reward-related regions when agreeing more with their peers than parents (collapsed across both positive and negative behaviors), highlighting the important role of these regions in reconciling conflicting social information from parents and peers, and ultimately agreeing with the peer. Overall, this research highlights the need to further investigate how interactions between family and peer influence differentially affect adolescent decision making, with the goal of identifying opportunities to leverage adolescents’ increased social and neurobiological susceptibility in favor of positive developmental outcomes.

VI. Conclusions and future directions

Social influences from peers and family have a profound impact on positive youth adjustment. Although susceptibility to social influence is often viewed as a vulnerability in adolescent development, particularly in the peer domain (and arguably so, given the evidence for peer-related increases in risk taking behaviors), we reviewed empirical support that underscores the positive side of susceptibility to social influence. Peers and families provide an opportunity for social adjustment, with the potential to redirect negative trajectories and increase positive outcomes. With empirical evidence showing that social influence relates to positive adjustment, it is key to capitalize on the social context and use this time as a period of investment, perhaps especially during middle school when adolescents are thought to be most socially sensitive ( Knoll, Magis-Weinberg, Speekenbrink, & Blakemore, 2015 ; Van Hoorn et al., 2016b). Indeed, recent prevention programs designed to decrease problem behaviors (e.g., tobacco use, peer victimization) and/or increase positive behaviors (e.g., prosocial behaviors), have successfully applied aspects of social learning and social identity theories in the promotion of positive classroom norms and use of socially salient referent peers to change negative attitudes ( Van Luijk et al., 2011 ; Paluck & Shepherd, 2012 ). Despite increasing attention to the positive side of social influences and its application in interventions, further research is needed to fully capture the inherent complexities of the social influence process and its relation to positive youth adjustment. With increased understanding of the social influence processes involved in deviancy training, we could modify and apply them to prosocial training, in which youth are exposed to more positive social influences.

Emerging evidence from developmental neuroscience has identified neurobiological processes through which peers and family influence decision-making and positive adjustment via changes in functional brain activity. Indeed, social influences from peers and parents are neurally represented in the adolescent brain by activity in a collection of cognitive, affective and social brain areas. Adolescents’ decisions and positive adjustment outcomes are likely affected by differential neural sensitivity to family and peers, and future studies should further probe the neural mechanisms of simultaneous and interactive influence from these two salient social sources. Given that social influence often occurs on a more implicit and unconscious level, the developmental social neuroscience perspective provides an informative additional layer of assessment that complements behavioral self-report and experimental methods.

While the peer and family contexts are especially critical in understanding positive adolescent development ( Van Ryzin et al., 2012 ), this is admittedly a narrow view of the social context. Other salient persons in the immediate environment may also be potent sources of social influence, such as sports team coaches, teachers, and mentors. Large individual differences exist in such proximal social contexts, and it is important to consider these individual differences within the larger social network (i.e., school context, neighborhoods, and larger community; Bronfenbrenner & Morris, 2006 ). Some youth may have access to mentoring opportunities in their local neighborhood (both setting an example as mentor and learning as mentee), whereas others do not, which may greatly impact the form and power of social influence. While those with no access to mentoring opportunities are perhaps more exposed to social influences from parents and siblings at home, youth with a larger social network who play sports or music with peers may be more exposed to peer norms. Hence, in order to help youth thrive, it is important for future work to study the complex influences from the social context on positive youth development. And perhaps, the question posed at the start of the chapter will eventually be complemented with “If your friends would [insert something positive here], then would you too?” .

Acknowledgments

This work was supported by the National Institutes of Health (R01DA039923 to Telzer) and the National Science Foundation (SES 1459719 to Telzer).

• Akers RL (2001). Social learning theory. In Paternoster R & Bachman R (Eds.), Explaining Criminals and Crime: Essays in Contemporary Criminological Theory (pp. 192–210). Los Angeles, CA: Roxbury. [ Google Scholar ]
• Akers RL (2011). Social Learning and Social Structure: A General Theory of Crime and Deviance Piscataway, NJ: Transaction Publishers. [ Google Scholar ]
• Akers RL, Krohn MD, Lanza-Kaduce L, & Radosevich M (1979). Social learning and deviant behavior: A specific test of a general theory. American Sociological Review , 44 ( 4 ), 636–655. 10.2307/2094592 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Akers RL, & Jensen GF (2006). The empirical status of social learning theory of crime and deviance: The past, present, and future. Taking stock: The Status of Criminological Theory , 15 , 37–76. [ Google Scholar ]
• Alfaro EC, & Umaña-Taylor AJ (2010). Latino adolescents’ academic motivation: The role of siblings. Hispanic Journal of Behavioral Sciences , 32 ( 4 ), 549–570. 10.1177/0739986310383165 [ CrossRef ] [ Google Scholar ]
• Amaral DG (2002). The primate amygdala and the neurobiology of social behavior: Implications for understanding social anxiety. Biological Psychiatry , 51 ( 1 ), 11–17. 10.1016/S0006-3223(01)01307-5 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Bahr SJ, Hoffmann JP, & Yang X (2005). Parental and peer influences on the risk of adolescent drug use. Journal of Primary Prevention , 26 ( 6 ), 529–551. 10.1007/s10935-005-0014-8 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Bandura A (1977). Self-efficacy: Toward a unifying theory of behavioral change. Psychological Review , 84 ( 2 ), 191–215. 10.1037//0033-295x.84.2.191 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Bandura A (1986). Social Foundations of Thought and Action: A Social Cognitive Theory Englewood Cliffs, NJ: Prentice-Hall. [ Google Scholar ]
• Bandura A (2001). Social cognitive theory: An agentic perspective. Annual Review of Psychology , 52 ( 1 ), 1–26. 10.1146/annurev.psych.52.1.1 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Bandura A, & Kupers CJ (1964). The transmission of patterns of self-reinforcement through modeling. Journal of Abnormal and Social Psychology , 69 , 1–9. 10.1037/h0041187 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Barry CM, & Wentzel KR (2006). Friend influence on prosocial behavior: The role of motivational factors and friendship characteristics. Developmental Psychology , 42 ( 1 ), 153–163. 10.1037/0012-1649.42.1.153 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Baumeister RF, & Leary MR (1995). The need to belong: desire for interpersonal attachments as a fundamental human motivation. Psychological Bulletin , 117 ( 3 ), 497–529. 10.1037//0033-2909.117.3.497 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Beck KH, Shattuck T, & Raleigh R (2001). Parental predictors of teen driving risk. American Journal of Health Behavior , 25 ( 1 ), 10–20. 10.5993/ajhb.25.1.2 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Belsky J, & Pluess M (2009). Beyond diathesis stress: Differential susceptibility to environmental influences. Psychological Bulletin , 135 ( 6 ), 885–908. 10.1037/a0017376 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Berger C, & Rodkin PC (2012). Group influences on individual aggression and prosociality: Early adolescents who change peer affiliations. Social Development , 21 ( 2 ), 396–413. 10.1111/j.1467-9507.2011.00628.x [ CrossRef ] [ Google Scholar ]
• Berndt TJ (1979). Developmental changes in conformity to peers and parents. Developmental Psychology , 15 ( 6 ), 608–616. 10.1037//0012-1649.15.6.608 [ CrossRef ] [ Google Scholar ]
• Biddle BJ, Bank BJ, & Marlin MM (1980). Parental and peer influence on adolescents. Social Forces , 58 ( 4 ), 1057–1079. 10.1038/nrn2353 [ CrossRef ] [ Google Scholar ]
• Blakemore SJ (2008). The social brain in adolescence. Nature Reviews Neuroscience , 9 ( 4 ), 267–277. 10.1038/nrn2353 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Blakemore SJ, & Mills KL (2014). Is adolescence a sensitive period for sociocultural processing? Annual Review of Psychology , 65 ( 1 ), 187–207. 10.1146/annurev-psych-010213-115202 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Blakemore SJ, den Ouden H, Choudhury S, & Frith C (2007). Adolescent development of the neural circuitry for thinking about intentions. Social Cognitive and Affective Neuroscience , 2 ( 2 ), 130–139. 10.1093/scan/nsm009 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Booth JR, Burman DD, Meyer JR, Lei Z, Trommer BL, Davenport ND, … & Mesulam MM. (2003). Neural development of selective attention and response inhibition. NeuroImage , 20 ( 2 ), 737–751. 10.1016/s1053-8119(03)00404-x [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Borawski EA, Ievers-Landis CE, Lovegreen LD, & Trapl ES (2003). Parental monitoring, negotiated unsupervised time, and parental trust: The role of perceived parenting practices in adolescent health risk behaviors. Journal of Adolescent Health , 33 ( 2 ), 60–70. 10.1016/s1054-139x(03)00100-9 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Bowes L, Maughan B, Caspi A, Moffitt TE, & Arseneault L (2010). Families promote emotional and behavioural resilience to bullying: Evidence of an environmental effect. Journal of Child Psychology and Psychiatry , 51 ( 7 ), 809–817. 10.1111/j.1469-7610.2010.02216.x [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Boyce WT, & Ellis BJ (2005). Biological sensitivity to context: I. An evolutionary–developmental theory of the origins and functions of stress reactivity. Development and Psychopathology , 17 ( 2 ), 271–301. 10.1017/s0954579405050145 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Braams BR, & Crone EA (2016). Peers and parents: A comparison between neural activation when winning for friends and mothers in adolescence. Social Cognitive and Affective Neuroscience , 33 ( 5 ), nsw136 10.1093/scan/nsw136 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Brittain CV (1963). Adolescent choices and parent-peer cross-pressures. American Sociological Review , 28 ( 3 ), 385–391. 10.2307/2090349 [ CrossRef ] [ Google Scholar ]
• Bronfenbrenner U, & Morris PA (2006). The bioecological model of human development. Handbook of Child Psychology 10.1002/9780470147658.chpsy0114 [ CrossRef ]
• Brown BB, Mounts N, Lamborn SD, & Steinberg L (1993). Parenting practices and peer group affiliation in adolescence. Child Development , 64 ( 2 ), 467–482. 10.2307/1131263 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Brown BB, Bakken JP, Ameringer SW, & Mahon MD (2008). A comprehensive conceptualization of the peer influence process in adolescence. In Prinstein MJ & Dodge KA (Eds.), Understanding peer influence in children and adolescents (pp. 17–44). New York: Guilford. [ Google Scholar ]
• Buist KL, Paalman CH, Branje SJT, Deković M, Reitz E, Verhoeven M, … Hale WW. (2014). Longitudinal effects of sibling relationship quality on adolescent problem behavior: A cross-ethnic comparison. Cultural Diversity & Ethnic Minority Psychology , 20 ( 2 ), 266–75. 10.1037/a0033675 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Bunge SA, Dudukovic NM, Thomason ME, Vaidya CJ, & Gabrieli JD (2002). Immature frontal lobe contributions to cognitive control in children: evidence from fMRI. Neuron , 33 ( 2 ), 301–311. 10.1016/s0896-6273(01)00583-9 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Burnett S, Bird G, Moll J, Frith C, & Blakemore SJ (2009). Development during adolescence of the neural processing of social emotion. Journal of Cognitive Neuroscience , 21 ( 9 ), 1736–1750. 10.1162/jocn.2009.21121 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Cascio CN, Carp J, O’Donnell MB, Tinney FJ Jr, Bingham CR, Shope JT, … & Falk EB. (2015). Buffering social influence: Neural correlates of response inhibition predict driving safety in the presence of a peer. Journal of Cognitive Neuroscience , 27 ( 1 ), 83–95. 10.1162/jocn_a_00693 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Casey BJ (2015). Beyond simple models of self-control to circuit-based accounts of adolescent behavior. Annual Review of Psychology , 66 ( 1 ), 295–319. 10.1146/annurev-psych-010814-015156 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Casey BJ, Jones RM, & Hare TA (2008). The adolescent brain. Annals of the New York Academy of Sciences , 1124 ( 1 ), 111–126. 10.1196/annals.1440.010 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Chein J, Albert D, O’Brien L, Uckert K, & Steinberg L (2011). Peers increase adolescent risk taking by enhancing activity in the brain’s reward circuitry. Developmental Science , 14 ( 2 ), F1–F10. 10.1111/j.1467-7687.2010.01035.x [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Choukas-Bradley S, Giletta M, Cohen GL, & Prinstein MJ (2015). Peer influence, peer status, and prosocial behavior: An experimental investigation of peer socialization of adolescents’ intentions to volunteer. Journal of Youth and Adolescence , 44 ( 12 ), 2197–2210. 10.1007/s10964-015-0373-2 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Cialdini RB, & Trost MR (1998). Social influence: Social norms, conformity, and compliance. In Gilbert D, Fiske S, & Lindzey G (Eds.), Handbook of Social Psychology (pp. 151–192). New York, NY: McGraw-Hill. [ Google Scholar ]
• Clasen DR, & Brown BB (1987). Understanding peer pressure in middle school. Middle School Research Selected Studies , 12 ( 1 ), 65–75. 10.1080/08851700.1987.11670280 [ CrossRef ] [ Google Scholar ]
• Cohen S, Gottlieb BH, Underwood LG (2001). Social relationships and health: Challenges for measurement and intervention. Advances in Mind-Body Medicine , 17 , 129–41. 10.1093/med:psych/9780195126709.003.0001 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Conger KJ (2013). Beyond tattling: What can siblings tell us about adolescent behavior? The Journal of Adolescent Health : 53 ( 2 ), 151–3. 10.1016/j.jadohealth.2013.06.011 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Conger KJ, Conger RD, & Elder JH Jr. (1994). Sibling relations during hard times. In Conger RD & Elder GH Jr. (Eds.), Families in Troubled Times: Adapting to Change in Rural America Hawthorne, NY: Aldine. [ Google Scholar ]
• Cox MJ (2010). Family systems and sibling relationships. Child Development Perspectives , 4 ( 2 ), 95–96. 10.1111/j.1750-8606.2010.00124.x [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Cox MJ, & Paley B (1997). Families as systems. Annual Review of Psychology , 48 , 243–67. 10.1146/annurev.psych.48.1.243 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Criss MM, Pettit GS, Bates JE, Dodge KA, & Lapp AL (2002). Family adversity, positive peer relationships, and children’s externalizing behavior: A longitudinal perspective on risk and resilience. Child Development , 73 ( 4 ), 1220–1237. 10.1111/1467-8624.00468 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Crockett L, Losoff M, and Petersen AC (1984). Perceptions of the peer group and friendship in early adolescence. Journal of Early Adolescence , 4 ( 2 ), 155–181. 10.1177/0272431684042004 [ CrossRef ] [ Google Scholar ]
• Crone EA, & Dahl RE (2012). Understanding adolescence as a period of social-affective engagement and goal flexibility. Nature Reviews Neuroscience , 13 ( 9 ), 636–650. 10.1038/nrn3313 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Cuellar I, Arnold B, & Maldonado R (1995). Acculturation rating scale for Mexican Americans-II: A revision of the original ARSMA scale. Hispanic Journal of Behavioral Sciences , 17 ( 3 ), 275–304. 10.1177/07399863950173001 [ CrossRef ] [ Google Scholar ]
• Delgado MR (2007). Reward‐related responses in the human striatum. Annals of the New York Academy of Sciences , 1104 ( 1 ), 70–88. 10.1196/annals.1390.002 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Deutsch M, & Gerard HB (1955). A study of normative and informational social influences upon individual judgment. Journal of Abnormal and Social Psychology , 51 ( 3 ), 629–636. 10.1037/h0046408 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Dirks MA, Persram R, Recchia HE, & Howe N (2015). Sibling relationships as sources of risk and resilience in the development and maintenance of internalizing and externalizing problems during childhood and adolescence. Clinical Psychology Review 42 , 145–155. 10.1016/j.cpr.2015.07.003 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Dishion TJ, Spracklen KM, Andrews DW, & Patterson GR (1996). Deviancy training in male adolescent friendships. Behavior Therapy , 27 ( 3 ), 373–390. 10.1016/s0005-7894(96)80023-2 [ CrossRef ] [ Google Scholar ]
• Dishion TJ, & Tipsord JM (2011). Peer contagion in child and adolescent social and emotional development. Annual Review of Psychology , 62 , 189–214. 10.1146/annurev.psych.093008.100412 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Do KT, Guassi Moreira JF, & Telzer EH (2017). But is helping you worth the risk? Defining prosocial risk taking in adolescence. Developmental Cognitive Neuroscience , 25 , 260–271. 10.1016/j.dcn.2016.11.008 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Dotterer AM, McHale SM, & Crouter AC (2009). The development and correlates of academic interests from childhood through adolescence. Journal of Educational Psychology , 101 ( 2 ), 509–519. 10.1037/a0013987 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Durston S, Davidson MC, Tottenham N, Galvan A, Spicer J, Fossella JA, & Casey BJ (2006). A shift from diffuse to focal cortical activity with development. Developmental Science , 9 ( 1 ), 1–8. 10.1111/j.1467-7687.2005.00454.x [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Eccles JS (1999). The development of children ages 6 to 14. The Future of Children , 9 ( 2 ), 30–44. 10.2307/1602703 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Ellis WE, & Zarbatany L (2007). Peer group status as a moderator of group influence on children’s deviant, aggressive, and prosocial behavior. Child Development , 78 ( 4 ), 1240–1254. 10.1111/j.1467-8624.2007.01063.x [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Ernst M (2014). The triadic model perspective for the study of adolescent motivated behavior. Brain and Cognition , 89 , 104–111. 10.1016/j.bandc.2014.01.006 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Ernst M, Pine DS, & Hardin M (2006). Triadic model of the neurobiology of motivated behavior in adolescence. Psychological Medicine , 36 ( 3 ), 299–312. 10.1017/s0033291705005891 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Eshel N, Nelson EE, Blair RJ, Pine DS, & Ernst M (2007). Neural substrates of choice selection in adults and adolescents: development of the ventrolateral prefrontal and anterior cingulate cortices. Neuropsychologia , 45 ( 6 ), 1270–1279. 10.1016/j.neuropsychologia.2006.10.004 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Festinger L (1950). Informal social communication. Psychological Review , 57 ( 5 ), 271–82. 10.1037/h0056932 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Frith CD, & Frith U (2007). Social cognition in humans. Current Biology , 17 ( 16 ), R724–R732. 10.1016/j.cub.2007.05.068 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Fuligni AJ (2001). Family obligation and the academic motivation of adolescents from Asian, Latin American, and European backgrounds. New Directions for Child and Adolescent Development , 2001 ( 94 ), 61–76. 10.1002/cd.31 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Furman W, & Buhrmester D (1992). Age and sex differences in perceptions of networks of personal relationships. Child Development , 63 ( 1 ), 103–115. 10.1111/j.1467-8624.1992.tb03599.x [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Galvan A, Hare TA, Parra CE, Penn J, Voss H, Glover G, & Casey BJ (2006). Earlier development of the accumbens relative to orbitofrontal cortex might underlie risk-taking behavior in adolescents. Journal of Neuroscience , 26 ( 25 ), 6885–6892. 10.1523/jneurosci.1062-06.2006 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Gass K, Jenkins J, & Dunn J (2007). Are sibling relationships protective? A longitudinal study. Journal of Child Psychology and Psychiatry and Allied Disciplines , 48 ( 2 ), 167–75. 10.1111/j.1469-7610.2006.01699.x [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Germán M, Gonzales NA, & Dumka L (2009). Familism values as a protective factor for Mexican-origin adolescents exposed to deviant peers. The Journal of Early Adolescence , 29 ( 1 ), 16–42. 10.1177/0272431608324475 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Grolnick WS, & Slowiaczek ML (1994). Parents’ involvement in children’s schooling: A multidimensional conceptualization and motivational model. Child Development , 65 ( 1 ), 237–252. 10.1111/j.1467-8624.1994.tb00747.x [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Guassi Moreira JF, & Telzer EH (2015). Changes in family cohesion and links to depression during the college transition. Journal of Adolescence , 43 , 72–82. 10.1016/j.adolescence.2015.05.012 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Guassi Moreira JF, & Telzer EH (in press). Mother still knows best: Maternal influence uniquely modulates adolescent reward sensitivity during risk taking. Developmental Science 10.1111/desc.12484 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Gunther Moor BG, Güroğlu B, Op de Macks ZA, Rombouts SARB, Van der Molen MW, & Crone EA (2012). Social exclusion and punishment of excluders: Neural correlates and developmental trajectories. NeuroImage , 59 ( 1 ), 708–717. 10.1016/j.neuroimage.2011.07.028 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Hamann SB, Ely TD, Hoffman JM, & Kilts CD (2002). Ecstasy and agony: activation of the human amygdala in positive and negative emotion. Psychological Science , 13 ( 2 ), 135–141. 10.1111/1467-9280.00425 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Hart D, & Fegley S (1995). Prosocial behavior and caring in adolescence: Relations to self‐understanding and social judgment. Child Development , 66 ( 5 ), 1346–1359. 10.1111/j.1467-8624.1995.tb00939.x [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Hogg MA, & Reid SA (2006). Social identity, self‐categorization, and the communication of group norms. Communication Theory , 16 ( 1 ), 7–30. 10.1111/j.1468-2885.2006.00003.x [ CrossRef ] [ Google Scholar ]
• Hollifield CR, & Conger KJ (2015). The role of siblings and psychological needs in predicting life satisfaction during emerging adulthood. Emerging Adulthood , 3 ( 3 ), 143–153. 10.1177/2167696814561544 [ CrossRef ] [ Google Scholar ]
• Johnson SC, Baxter LC, Wilder LS, Pipe JG, Heiserman JE, & Prigatano GP (2002). Neural correlates of self‐reflection. Brain , 125 ( 8 ), 1808–1814. 10.1016/s1053-8119(01)91765-3 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Kelley WM, Macrae CN, Wyland CL, Caglar S, Inati S, & Heatherton TF (2002). Finding the self? An event-related fMRI study. Journal of Cognitive Neuroscience , 14 ( 5 ), 785–794. 10.1162/08989290260138672 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Knoll LJ, Magis-Weinberg L, Speekenbrink M, & Blakemore SJ (2015). Social influence on risk perception during adolescence. Psychological Science , 26 ( 5 ), 583–592. 10.1177/0956797615569578 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Kringelbach ML (2005). The human orbitofrontal cortex: Linking reward to hedonic experience. Nature Reviews Neuroscience , 6 ( 9 ), 691–702. 10.1038/nrn1747 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Kroger J (2000). Ego identity status research in the new millennium. International Journal of Behavioral Development , 24 ( 2 ), 145–148. 10.1080/016502500383250 [ CrossRef ] [ Google Scholar ]
• Krohn MD, Skinner WF, Massey JL, & Akers RL (1985). Social learning theory and adolescent cigarette smoking: A longitudinal study. Social Problems , 32 ( 5 ), 455–473. 10.1525/sp.1985.32.5.03a00050 [ CrossRef ] [ Google Scholar ]
• Kurdek LA, & Fine MA (1995). Mothers, fathers, stepfathers, and siblings as providers of supervision, acceptance, and autonomy to young adolescents. Journal of Family Psychology , 9 ( 1 ), 95–99. 10.1037//0893-3200.9.1.95 [ CrossRef ] [ Google Scholar ]
• Laible DJ, Carlo G, & Raffaelli M (2000). The differential relations of parent and peer attachment to adolescent adjustment. Journal of Youth and Adolescence , 29 ( 1 ), 45–59. 10.1023/A:1005169004882 [ CrossRef ] [ Google Scholar ]
• Lam CB, McHale SM, & Crouter AC (2012). Parent–child shared time from middle childhood to late adolescence: Developmental course and adjustment correlates. Child Development , 83 ( 6 ), 2089–2103. 10.1111/j.1467-8624.2012.01826.x [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Lam CB, McHale SM, & Crouter AC (2014). Time with peers from middle childhood to late adolescence: Developmental course and adjustment correlates. Child Development , 85 ( 4 ), 1677–1693. 10.1111/cdev.12235 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Marsh R, Zhu H, Schultz RT, Quackenbush G, Royal J, Skudlarski P, & Peterson BS (2006). A developmental fMRI study of self‐regulatory control. Human Brain Mapping , 27 ( 11 ), 848–863. 10.1002/hbm.20225 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• McDonald RI, Fielding KS, & Louis WR (2013). Energizing and de-motivating effects of norm-conflict. Personality and Social Psychology Bulletin , 39 ( 1 ), 57–72. 10.1177/0146167212464234 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Melby JN, Conger RD, Fang S-A, Wickrama KAS, & Conger KJ (2008). Adolescent family experiences and educational attainment during early adulthood. Developmental Psychology , 44 ( 6 ), 1519–36. 10.1037/a0013352 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Metcalfe J, & Mischel W (1999). A hot/cool-system analysis of delay of gratification: dynamics of willpower. Psychological Review , 106 ( 1 ), 3–19. 10.1037//0033-295x.106.1.3 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Milevsky A, & Levitt MJ (2005). Sibling support in early adolescence: Buffering and compensation across relationships. European Journal of Developmental Psychology , 2 ( 3 ), 299–320. 10.1080/17405620544000048 [ CrossRef ] [ Google Scholar ]
• Minuchin P (1985). Relationships within the family: A systems perspective on development. In Hinde RA & Stevenson-Hinde J (Eds.), Relationships within Families: Mutual Influences (pp. 1–6). Oxford: Clarendon Press. [ Google Scholar ]
• Mobbs D, Yu R, Meyer M, Passamonti L, Seymour B, Calder AJ, … Dalgleish T (2009). A key role for similarity in vicarious reward. Science , 324 ( 5929 ), 900 10.1126/science.1170539 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Murray C (2009). Parent and teacher relationships as predictors of school engagement and functioning among low-income urban youth. The Journal of Early Adolescence , 29 ( 3 ), 376–404. 10.1177/0272431608322940 [ CrossRef ] [ Google Scholar ]
• Nash SG, Mcqueen A, & Bray JH (2005). Pathways to adolescent alcohol use: Family environment, peer influence, and parental expectations. Journal of Adolescent Health , 37 ( 1 ), 19–28. 10.1016/j.jadohealth.2004.06.004 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Nelson EE, Leibenluft E, McClure EB, & Pine DS (2005). The social re-orientation of adolescence: A neuroscience perspective on the process and its relation to psychopathology. Psychological Medicine , 35 ( 2 ), 163–174. 10.1017/s0033291704003915 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Nelson EE, Jarcho JM, & Guyer AE (2016). Social re-orientation and brain development: An expanded and updated view. Developmental Cognitive Neuroscience , 17 , 118–127. 10.1016/j.dcn.2015.12.008 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Newman BM, & Newman PR (2001). Group identity and alienation: Giving the we its due. Journal of Youth and Adolescence , 30 ( 5 ), 515–538. 10.1023/a:1010480003929 [ CrossRef ] [ Google Scholar ]
• Nolan JM, Schultz PW, Cialdini RB, Goldstein NJ, & Griskevicius V (2008). Normative social influence is underdetected. Personality and Social Psychology Bulletin , 34 ( 7 ), 913–923. 10.1177/0146167208316691 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Oakes J (1987). Tracking in secondary schools: A contextual perspective. Educational Psychologist , 22 ( 2 ), 129–153. 10.1207/s15326985ep2202_3 [ CrossRef ] [ Google Scholar ]
• Padilla-Walker LM, & Carlo G (2007). Personal values as a mediator between parent and peer expectations and adolescent behaviors. Journal of Family Psychology , 21 ( 3 ), 538–541. 10.1037/0893-3200.21.3.538 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Padilla-Walker LM, & Carlo G (2014). The study of prosocial behavior. Prosocial Development: A Multidimensional Approach (pp. 3–16). Oxford: Clarendon Press; 10.1093/acprof:oso/9780199964772.003.0001 [ CrossRef ] [ Google Scholar ]
• Padilla-Walker LM, Fraser AM, Black BB, & Bean RA (2015). Associations between friendship, sympathy, and prosocial behavior toward friends. Journal of Research on Adolescence , 25 ( 1 ), 28–35. 10.1111/jora.12108 [ CrossRef ] [ Google Scholar ]
• Paluck EL, & Shepherd H (2012). The salience of social referents: A field experiment on collective norms and harassment behavior in a school social network. Journal of Personality and Social Psychology , 103 ( 6 ), 899–915. 10.1037/a0030015 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Parke RD (2004). Development in the family. Annual Review of Psychology , 55 ( 1 ), 365–399. 10.1146/annurev.psych.55.090902.141528 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Payne MA (2012). All gas and no brakes! Journal of Adolescent Research , 27 ( 1 ), 3–17. 10.1177/0743558411412956 [ CrossRef ] [ Google Scholar ]
• Perino MT, Miernicki ME, & Telzer EH (2016). Letting the good times roll: adolescence as a period of reduced inhibition to appetitive social cues. Social Cognitive and Affective Neuroscience , 11 ( 11 ), 1762–1771. 10.1093/scan/nsw096 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Pfeifer JH, Masten CL, Borofsky LA, Dapretto M, Fuligni AJ, & Lieberman MD (2009). Neural correlates of direct and reflected self‐appraisals in adolescents and adults: When social perspective‐taking informs self‐perception. Child Development , 80 ( 4 ), 1016–1038. 10.1111/j.1467-8624.2009.01314.x [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Pfeifer JH, & Allen NB (2012). Arrested development? Reconsidering dual-systems models of brain function in adolescence and disorders. Trends in Cognitive Sciences , 16 ( 6 ), 322–329. 10.1016/j.tics.2012.04.011 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Pfeifer JH, & Allen NB (2016). The audacity of specificity: Moving adolescent developmental neuroscience towards more powerful scientific paradigms and translatable models. Developmental Cognitive Neuroscience , 17 , 131–137. 10.1016/j.dcn.2015.12.012 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Prentice DA, & Miller DT (1993). Pluralistic ignorance and alcohol use on campus: Some consequences of misperceiving the social norm. Journal of Personality and Social Psychology , 64 ( 2 ), 243–256. 10.1037//0022-3514.64.2.243 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Prentice DA, & Miller DT (1996). Pluralistic ignorance and the perpetuation of social norms by unwitting actors. Advances in Experimental Social Psychology , 28 , 161–209. 10.1016/S0065-2601(08)60238-5 [ CrossRef ] [ Google Scholar ]
• Prinstein MJ, & Wang SS (2005). False consensus and adolescent peer contagion: Examining discrepancies between perceptions and actual reported levels of friends’ deviant and health risk behaviors. Journal of Abnormal Child Psychology , 33 ( 3 ), 293–306. 10.1007/s10802-005-3566-4 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Qu Y, Fuligni AJ, Gálvan A, Lieberman MD, & Telzer EH (2016). Links between parental depression and longitudinal changes in youths’ neural sensitivity to rewards. Social Cognitive Affective Neuroscience , 11 ( 8 ), 1262–1271. 10.1093/scan/nsw035 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Richardson JL, Radziszewska B, Dent CW, & Flay BR (1993). Relationship between after-school care of adolescents and substance use, risk taking, depressed mood, and academic achievement. Pediatrics , 92 ( 1 ), 32–38. [ PubMed ] [ Google Scholar ]
• Rubia K, Smith AB, Taylor E, & Brammer M (2007). Linear age‐correlated functional development of right inferior fronto‐striato‐cerebellar networks during response inhibition and anterior cingulate during error‐related processes. Human Brain Mapping , 28 ( 11 ), 1163–1177. 10.1002/hbm.20347 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Saez RA, Saez A, Paton JJ, Lau B, & Salzman CD (2017). Distinct roles for the amygdala and orbitofrontal cortex in representing the relative amount of expected reward. Neuron , 95 ( 1 ), 70–77. 10.1016/j.neuron.2017.06.012 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Samek DR, Rueter MA, Keyes MA, McGue M, & Iacono WG (2015). Parent involvement, sibling companionship, and adolescent substance use: A longitudinal, genetically informed design. Journal of Family Psychology , 29 ( 4 ), 614–623. 10.1037/fam0000097 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Sapouna M, & Wolke D (2013). Resilience to bullying victimization: The role of individual, family and peer characteristics. Child Abuse and Neglect , 37 ( 11 ), 997–1006. 10.1016/j.chiabu.2013.05.009 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Saxbe D, Del Piero L, Immordino-Yang MH, Kaplan J, & Margolin G (2015). Neural correlates of adolescents’ viewing of parents’ and peers’ emotions: Associations with risk-taking behavior and risky peer affiliations. Social Neuroscience , 10 ( 6 ), 592–604. 10.1080/17470919.2015.1022216 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Scriber RA, & Guyer AE (2016). Adolescent neurobiological susceptibility to social context. Developmental Cognitive Neuroscience , 19 , 1–18. 10.1016/j.dcn.2015.12.009 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Sebald H, & White B (1980). Teenagers’ divided reference groups: Uneven alignment with parents and peers. Adolescence (Vol. XV No, Vol. 15 ). Roslyn Hts., NY: Libra Publishers. [ Google Scholar ]
• Shibutani T (1955). Reference groups as perspectives. American Journal of Sociology , 60 ( 6 ), 562–569. 10.1086/221630 [ CrossRef ] [ Google Scholar ]
• Shulman EP, Smith AR, Silva K, Icenogle G, Duell N, Chein J, & Steinberg L (2016). The dual systems model: Review, reappraisal, and reaffirmation. Developmental Cognitive Neuroscience , 17 , 103–117. 10.1016/j.dcn.2015.12.010 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Simpkins SD, Fredricks JA & Eccles JS (2012). Charting the Eccles’ expectancy-value model from mothers’ beliefs in childhood to youths’ activities in adolescence. Developmental Psychology , 48 ( 4 ), 1019–32. 10.1037/a0027468 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Somerville LH, Jones RM, & Casey BJ (2010). A time of change: Behavioral and neural correlates of adolescent sensitivity to appetitive and aversive environmental cues. Brain and Cognition , 72 , 124–133. 10.1016/j.bandc.2009.07.003 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Somerville LH, Hare T, & Casey BJ (2011). Frontostriatal maturation predicts cognitive control failure to appetitive cues in adolescents. Journal of Cognitive Neuroscience , 23 ( 9 ), 2123–2134. 10.1162/jocn.2010.21572 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Somerville LH, Jones RM, Ruberry EJ, Dyke JP, Glover G, & Casey BJ (2013). The medial prefrontal cortex and the emergence of self-conscious emotion in adolescence. Psychological Science , 24 ( 8 ), 1554–1562. 10.1177/0956797613475633 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Steinberg L (2008). A social neuroscience perspective on adolescent risk-taking. Developmental Review , 28 ( 1 ), 78–106. 10.1016/j.dr.2007.08.002 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Steinberg L, Lamborn SD, Darling N, Mounts NS, Dornbusch M, & Dornbusch SM (1994). Over-time changes in adjustment and competence among adolescents from authoritative, authoritarian, indulgent, and neglectful families. Child Development , 65 ( 3 ), 754–770. 10.1111/j.1467-8624.1994.tb00781.x [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Sutherland EH, Cressey DR, & Luckenbill DF (1992). Principles of Criminology Rowman & Littlefield. [ Google Scholar ]
• Tajfel H (1981). Human Groups and Social Categories: Studies in Social Psychology CUP Archive. [ Google Scholar ]
• Tajfel H, & Turner JC (1979). An integrative theory of intergroup conflict. The Social Psychology of Intergroup Relations , 33 ( 47 ), 74. [ Google Scholar ]
• Telzer EH (2016). Dopaminergic reward sensitivity can promote adolescent health: A new perspective on the mechanism of ventral striatum activation. Developmental Cognitive Neuroscience , 17 , 57–67. 10.1016/j.dcn.2015.10.010 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Telzer EH, & Fuligni AJ (2013). Positive daily family interactions eliminate gender differences in internalizing symptoms during adolescence. Journal of Youth and Adolescence , 42 ( 10 ), 1498–511. 10.1007/s10964-013-9964-y [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Telzer EH, Fuligni AJ, Lieberman MD, & Gálvan A (2013a). Meaningful family relationships: Neurocognitive buffers of adolescent risk taking. Journal of Cognitive Neuroscience , 25 ( 3 ), 374–387. 10.1162/jocn_a_00331 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Telzer EH, Fuligni AJ, Lieberman MD, & Gálvan A (2013b). Ventral striatum activation to prosocial rewards predicts longitudinal declines in adolescent risk taking. Developmental Cognitive Neuroscience , 3 , 45–52. 10.1016/j.dcn.2012.08.004 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Telzer EH, Gonzales N, & Fuligni AJ (2014). Family obligation values and family assistance behaviors: Protective and risk factors for adolescent substance use. Journal of Youth and Adolescence , 43 ( 2 ), 270–283. 10.1007/s10964-013-9941-5 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Telzer EH, Fuligni AJ, Lieberman MD, Miernicki ME, & Gálvan, A. (2015). The quality of adolescents’ peer relationships modulates neural sensitivity to risk taking. Social Cognitive Affective Neuroscience , 10 ( 3 ), 389–398. 10.1093/scan/nsu064 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Telzer EH, Ichien NI, & Qu Y (2015). Mothers know best: Redirecting adolescent reward sensitivity to promote safe behavior during risk taking. Social Cognitive Affective Neuroscience , 10 ( 10 ), 1383–1391. 10.1093/scan/nsv026 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Telzer EH, Tsai KM, Gonzales N, & Fuligni AJ (2015). Mexican-American adolescents’ family obligation values and behaviors: Links to internalizing symptoms across time and family context. Developmental Psychology , 51 ( 1 ), 75–86. 10.1037/a0038434 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Telzer EH, Fuligni AJ, & Gálvan A (2016). Identifying a cultural resource: Neural correlates of familial influence on risk taking among Mexican-origin adolescents. In Chiao JY, Li S-C, Seligman R, & Turner R (Eds). The Oxford Handbook of Cultural Neuroscience Oxford University Press, New York, NY: 10.1093/oxfordhb/9780199357376.013.15 [ CrossRef ] [ Google Scholar ]
• Terry DJ, & Hogg MA (1996). Group norms and the attitude-behavior relationship: A role for group identification. Personality and Social Psychology Bulletin , 22 ( 8 ), 776–793. 10.1177/0146167296228002 [ CrossRef ] [ Google Scholar ]
• Tucker CJ, & Updegraff K (2009). The relative contributions of parents and siblings to child and adolescent development. In Kramer L & Conger KJ (Eds.), Siblings as Agents of Socialization. New Directions in Child and Adolescent Development (pp. 13–28). San Francisco, CA: Jossey-Bass; 10.1002/cd [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Turner JC (1991). Social influence Thomson Brooks/Cole Publishing Co. [ Google Scholar ]
• Utech DA, & Hoving KL (1969). Parents and peers as competing influences in the decisions of children of differing ages. Journal of Social Psychology , 78 ( 2 ), 267–274. 10.1080/00224545.1969.9922366 [ CrossRef ] [ Google Scholar ]
• Van den Bos Van Dijk, Westenberg M, Rombouts SARB, & Crone EA (2011). Changing brains, changing perspectives. Psychological Science , 22 ( 1 ), 60–70. 10.1177/0956797610391102 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Van Hoorn J, Van Dijk E, Meuwese R, Rieffe C, & Crone EA (2016a). Peer influence on prosocial behavior in adolescence. Journal of Research on Adolescence , 26 ( 1 ), 90–100. 10.1111/jora.12173 [ CrossRef ] [ Google Scholar ]
• Van Hoorn J, Fuligni AJ, Crone EA, & Galván A (2016). Peer influence effects on risk-taking and prosocial decision-making in adolescence: Insights from neuroimaging studies. Current Opinion in Behavioral Sciences , 10 , 59–64. 10.1016/j.cobeha.2016.05.007 [ CrossRef ] [ Google Scholar ]
• Van Hoorn J, Van Dijk E, Güroğlu B, & Crone EA (2016). Neural correlates of prosocial peer influence on public goods game donations during adolescence. Social Cognitive and Affective Neuroscience , 11 ( 6 ), 923–933. 10.1093/scan/nsw013 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Van Lier PA, Huizink A, & Vuijk P (2011). The role of friends’ disruptive behavior in the development of children’s tobacco experimentation: Results from a preventive intervention study. Journal of Abnormal Child Psychology , 39 ( 1 ), 45–57. 10.1007/s10802-010-9446-6 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Van Ryzin MJ, Fosco GM, & Dishion TJ (2012). Family and peer predictors of substance use from early adolescence to early adulthood: An 11-year prospective analysis. Addiction Behavior , 37 ( 12 ), 1314–24. 10.1016/j.addbeh.2012.06.020 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Velanova K, Wheeler ME, & Luna B (2009). The maturation of task set-related activation supports late developmental improvements in inhibitory control. Journal of Neuroscience , 29 ( 40 ), 12558–12567. 10.1523/jneurosci.1579-09.2009 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Vitória PD, Salgueiro MF, Silva SA, & Vries H (2009). The impact of social influence on adolescent intention to smoke: Combining types and referents of influence. British Journal of Health Psychology , 14 ( 4 ), 681–699. 10.1348/135910709×421341 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Wang AT, Lee SS, Sigman M, & Dapretto M (2006). Developmental changes in the neural basis of interpreting communicative intent. Social Cognitive and Affective Neuroscience , 1 ( 2 ), 107–121. 10.1093/scan/nsl018 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Welborn BL, Lieberman MD, Goldenberg D, Fuligni AJ, Galvan A, & Telzer EH (2015). Neural mechanisms of social influence in adolescence. Social Cognitive and Affective Neuroscience , 11 ( 1 ), 100–109. 10.1093/scan/nsv095 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Wentzel KR, Filisetti L, & Looney L (2007). Adolescent prosocial behavior: The role of self‐processes and contextual cues. Child Development , 78 ( 3 ), 895–910. 10.1111/j.1467-8624.2007.01039.x [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Wessel JR, Conner CR, Aron AR, & Tandon N (2013). Chronometric electrical stimulation of right inferior frontal cortex increases motor braking. Journal of Neuroscience , 33 ( 50 ), 19611–19619. 10.1523/jneurosci.3468-13.2013 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Whiteman SD, Becerra JM, & Killoren SE (2009). Mechanisms of sibling socialization in normative family development. In Kramer & Conger L (Eds.), Siblings as Agents of Socialization. New Directions in Child and Adolescent Development (pp. 29–43). San Francisco, CA: Jossey-Bass; 10.1002/cd.255 [ PubMed ] [ CrossRef ] [ Google Scholar ]

ORIGINAL RESEARCH article

Following the majority: social influence in trusting behavior.

• 1 Key Laboratory for NeuroInformation of Ministry of Education, School of Life Science and Technology, University of Electronic Science and Technology of China, Chengdu, China
• 2 Key Laboratory of Cognition and Personality (MOE), Faculty of Psychology, Southwest University, Chongqing, China

When making decisions, people may change their behavior, sometimes against their personal preference, according to the opinions of peers. However, the effect of social influence on trust is still unknown. In our study, we used the event-related functional magnetic resonance imaging to investigate brain activity in social influence during a trust game. The behavioral results revealed that people tend to conform to others’ opinions and behaviors in a trust game. Decreased activations were observed in superior temporal gyrus during processing of social influences. Moreover, brain regions supporting value processing and reward learning were activated when subjects decided to follow the majority. These regions include the ventral medial prefrontal cortex, ventral striatum, and parahippocampal gyrus. Finally, our exploratory analysis revealed an increase in functional connectivity between the prefrontal cortex and the ventral striatum during conformity in trusting behavior. These findings indicate that the neural basis of social influence in trusting behavior are similar to the mechanisms implicated in reward learning. The brain regions involved in reward learning might reflect the reward value of agreeing with others in our study.

Introduction

Our opinions and behaviors are often affected by the majority ( Asch, 1956 ; Turner, 1991 ). People tend to change their opinions and behaviors in order to follow with social norms, even if the majority decision is against their personal preference ( Cialdini and Goldstein, 2004 ; Morgan and Laland, 2012 ; Haun et al., 2013 ). Psychologists defined this phenomenon as “social conformity.” It refers to individuals’ action of adopting the opinions, behaviors, and judgments of others ( Turner, 1991 ). Asch (1951) used a simple line judgment task to investigate social conformity. Since then social psychologists began to explore the causes of social conformity. Based on previous study, there are three types of intrinsic motivations underlying social conformity: a desire to obtain social approval of others, a desire to make a correct choice, and a desire to keep a positive self-concept ( Cialdini and Goldstein, 2004 ).

Recent studies have investigated the effect of conformity in many judgment tasks as well as the neural basis of conformity. By using mental rotation task and music rating task, Berns et al. (2005 , 2010 ) found that the opinions of peers could change participants’ initial judgments and affect neural activity within relatively low-level processing brain areas related to each task. In addition, previous literatures have reported that the brain regions associated with reward processing and behavioral adjustment were closely associated with social influence. Mason et al. (2009) exposed subjects to popular, unpopular and novel symbols and reported that the medial prefrontal cortex (mPFC) was involved in normative social influence by comparing socially and not socially marked symbols, while the striatum (the caudate) might be a possible index of informational social influence by comparing popular and unpopular symbols. Wei et al. (2013) also found that confliction with group norms during an ultimatum game activated the bilateral insula, bilateral middle frontal gyrus (MFG) and mPFC. Additionally, Klucharev et al. (2009) found that conflicting group opinions triggered a neuronal response in the nucleus accumbens and the rostral cingulate zone (RCZ). These brain regions are often associated with reward processing and behavioral adjustment, which is similar to prediction error signal ( Berns et al., 2001 ; Holroyd and Coles, 2002 ; Ridderinkhof et al., 2004 ). Neural activity in these regions could predict participants’ subsequent conforming behaviors ( Klucharev et al., 2009 ). By using stock task and music choice task, Burke et al. (2010) and Campbell-Meiklejohn et al. (2010) found that neural activity in the ventral striatum was involved in social influence, suggesting that the opinions of others could modulate the basic value signals in known reinforcement learning neural circuitry ( Campbell-Meiklejohn et al., 2010 ).

Conformity effect was also found in economic decisions, such as ultimatum game ( Wei et al., 2013 ), dictator game ( Wei et al., 2017 ), risk taking ( Gardner and Steinberg, 2005 ), stock market participation ( Hong et al., 2004 ), consuming decision and investment decision ( Bursztyn et al., 2014 ). These results indicated that the opinion of majority could influence people’s own preferences in economic decision context. Trust plays an important role in economic decision interactions ( Cochard et al., 2004 ). Previous study suggested that, for the trusting behaviors, genetics only explain about 20% of the cross-sectional variation while environmental factors would explain 80% of the variation ( Cesarini et al., 2008 ; Ahern et al., 2014 ). One potential environment factor is social conformity. Prior studies have found that individuals tended to change their rating of trustworthiness toward social norm in a trustworthiness judgment task ( Campbell-Meiklejohn et al., 2012 ; Simonsen et al., 2014 ). In present study, we used trust game to explore whether peers’ decision could change the choices of individuals. Trust game is widely used to measure trusting behavior. There are two players in the classic trust game: an investor and a trustee. Both players are endowed with $10. The investor decides whether give the money to the trustee. If the investor gives the money to the trustee, the endowment would be multiplied by experimenter then. In the end, the trustee decides whether to give any portion of the money she/he received back to the investor or just keep it. In our study, we developed a modified trust game. In this task, participants were able to see peer’ choices when they made the trust decision.

Firstly, we hypothesized that the choices of the majority would affect subjects’ trust preference. Subjects may invest the money to the trustee when they see that the majority of the group trusts the trustee. Conversely, participants may distrust the trustee if they see that the majority does not trust the trustee. Otherwise, subjects will insist on their own trust preferences if social influence has no effects on trust decision. Secondly, we predicted that participants may conform to the opinion of the majority with a relatively high level of decision confidence, since they may have high reward expectancy in the trust social influence condition. Finally, previous literatures had reported that social influence might affect participants’ behaviors through the neural underpinnings of reward learning and behavioral adjustment, such as ventral medial prefrontal cortex (vmPFC) and anterior cingulate cortex (ACC), and also brain structures underlying social reward processing especially the striatum ( Izuma et al., 2008 ; Mason et al., 2009 ; Klucharev et al., 2011 ; Wei et al., 2013 ). Therefore, we hypothesized that the activity in brain reward circuits such as the vmPFC and caudate may be associated with social influence. Recent brain imaging studies have suggested evidence that enhanced functional connectivity between the prefrontal cortex and ventral striatum during reward processing ( Camara et al., 2008 ). Hence, we hypothesized that a psychophysiological interaction (PPI) analysis may confirm an enhanced functional connectivity between the prefrontal cortex and ventral striatum during conformity in the trust social influence condition.

Materials and Methods

Participants.

Twenty-seven healthy right-handed participants (mean age = 21.1, female = 16, male = 11) participated in the experiment. These participants were recruited from Southwest University through advertisements in the online student forums, none of them came from department of psychology or economics. All were native Mandarin speakers, with no neurological illness as confirmed by psychiatric clinical assessment or psychological disorders, and with (corrected to) normal vision. Written informed consent was obtained in accordance with the regulations of the Ethics Committee of Southwest University. This study was approved by the Ethics Committee of Southwest University.

Stimulus Materials

Peers’ choices were presented in the form of a table to the participants. The number “1” refers to a choice to send the endowment to the stranger and the number “2” indicates a choice to keep the endowment. There were four conditions of social influence: trust influence (three or four group members decided to send the endowment to the stranger); moderate (two group members decided to trust the stranger while the other two decided keep their endowments); distrust influence (three or four group members decided to keep the endowment); and no information (the boxes corresponding to each group members’ choices were replaced with “×”). There were 70 offers in total. The offer stimuli consisted of the number of the trustee (randomly from 1 to 70), the choices available, and the social information (peers’ choices). The former was presented in the upper portion of the screen. The choices available were presented in the center of the screen and the latter in the lower part of the picture.

Experimental Procedures

Participants were told that they would play an on-line monetary game with four other participants, who would be in a separate behavioral laboratory. They would see the choices of the other peers on the computer screen during the decision phase of the experiment. Participants acted as an investor and play the game independently with 70 different strangers (trustees). These trustees were randomly selected from the university and played the game on the other floor. Participants and their group members did not know anything about these seventy trustees. At the beginning of each trial, both players (investor and trustee) were endowed with ¥10. The investor was asked to decide whether to send the endowment. The endowment would be tripled if the investor decided to invest. Then the trustee was asked to decide whether to send half of the money back (¥15). The investor would not know the outcome (i.e., trustees’ choice) during the task. Subjects were told that they will receive ¥50 for participating in the experiment plus the additional money earned from ten of their trust decisions, chosen at random, in the trust game. Subjects earned on average about ¥60 for their participated in the experiment which was not based on investment outcome. We asked participants whether she/he believed the existence of trustees after they finished the task. All the participants reported that they believed the existence of trustees. After the data of all the participants were collected, participants received payment and were told that the peers and trustees did not exist.

Participants then received details about the procedure of the experiment. At the beginning of each trial, they saw a fixation point for a 2–4 s jittered duration that varied pseudo-randomly. Then, the decision screen was presented for 3 s. They used the index and middle fingers of their right hands to separately respond to the offer by pressing one of two buttons on an MRI-compatible button box (“1” to invest and “2” to keep the endowment). Peers’ choices were placed in the lower part of the decision interface. Subsequently, confidence ratings were provided for 2 s. Finally, the word “next” displayed for 1 s, indicating that the next trial was about to begin. The sequence of events in a trial is illustrated in Figure 1 .

Figure 1. Demonstration of sequence of events in a trial (take trust influence condition for example).

There were seventy trials in present experiment. The duration of a trial is approximately 9 seconds. In 10 of the trials, participants were informed that two peers decided to send the money to the trustee while the other two decided to keep the endowments. These trials were used solely to maintain the believability of the interaction between the participant and the four peers. They were excluded in the final analysis. In one-third of the remaining trials (20 trials), participants could not see the group’s choices (the no information, or baseline condition; we told participants that the decisions in these trials were not made by all the four peers). For the 20 trials of the trust influence condition, three or four peers’ choices were to send the endowments to the trustee. For the 20 trials of the distrust influence condition, one or none of the group members decided to invest. Before performing the task in the scanner, all participants completed a training session. They were told that the computer for the pre-experiment training is not connected to the local network, therefore they could not receive anything information about the peers’ choices.

We used a PC running E-Prime 2.0 to display the stimuli and acquire the responses of the participants, as well as the reaction times (RTs). In the scanner, there was a mirror placed on the top of the image acquisition coil. Participants saw the experiment task via this mirror that reflected the screen mounted at the back of the scanner.

Image Acquisition

Functional MRI data were acquired using a 3T Siemens Trio scanner. Each scan contains 355 functional volumes, using an echo-planar imaging (EPI) sequence with the following parameters: TR/TE = 2000/30 ms, flip angle = 90°, acquisition matrix = 64 × 64, FOV = 192 mm × 192 mm, axial slices = 32, slice thickness/gap = 3mm/1 mm, voxel size = 3 mm × 3 mm × 3 mm. The first three images were discarded for the saturation effect.

Data Analysis

Behavioral data analysis.

We used statistical product and service solutions (SPSS) to analyze the behavioral data. We predicted that the choices of the majority may influence participants’ decision. A repeated measure (social influence: baseline, trust influence, distrust influence) ANOVA was used to analyze the RTs in the decision phase, as well as the rate of trust. Since we predicted that subjects may have high reward expectancy in the trust social influence condition, we conducted a 3 (social influence: baseline, trust influence, distrust influence) × 2 (choices: trust, distrust) ANOVA on the mean confidence rating.

fMRI Data Analysis

Image preprocessing was performed with statistical parametric mapping 8 (SPM8; Welcome Department of Imaging Neuroscience, University of London, United Kingdom). Functional images were first corrected for motion artifacts. Then images were interpolated to correct for slice timing, and spatially normalized into the Montreal Neurological Institute (MNI)-space using the SPM8 EPI template, and resampled into 3 mm × 3 mm × 3 mm voxels. Images were smoothed using an 8 mm 3 full-width-at-half-maximum (FWHM) Gaussian kernel. A 0.01 Hz–0.08 Hz band-pass filter, which was composed of a discrete cosine-basis function with a cutoff period of 128 s for the high-pass filter was applied to the time courses of all brain voxels.

We conducted analysis on functional magnetic resonance imaging data of the decision phase. General linear model analysis was performed with SPM8. Three regressors were entered based on social information (baseline, trust influence and distrust influence). These regressors were then convolved with the standard hemodynamic response function. In addition, the realignment parameters were included in the model to regress out potential movement artifacts. For a whole-brain analysis, the result was thresholded at p < 0.05 (FDR correction), cluster size > 10. The effect of social influence was estimated by contrasting the trust influence effect ( trust influence condition > no information ). For more detailed insights into the neural mechanisms underlying social conformity in trusting behavior, we did an exploratory analysis, analyzed the conforming behavior contrast ( conformity vs. non-conformity ) in trust influence condition ( trust influence condition – conformity > trust influence condition – non-conformity ). Activations in this analysis were thresholded at p < 0.05 (FDR correction), cluster size > 10.

Finally, an exploratory PPI analysis was performed in order to identify brain regions that showed significantly increased coordination (i.e., increased functional connectivity) with the ventral striatum activity related to conformity compared to non-conformity in the trust influence condition ( Friston et al., 1997 ). Based on our fMRI results and previous literature, the region of interest (ROI) was defined as a sphere with 6-mm-radius centered at the peak voxel in the ventral striatum (MNI coordinates: [10, 18, -9]) ( Campbell-Meiklejohn et al., 2010 ). The time series was extracted from each subject in the ventral striatum. And the PPI regressor was calculated as the element-by-element product of the mean-corrected activity of ROI and a vector coding for differential task effects of conformity-trust influence versus non-conformity-trust influence. The PPI regressors reflected the interaction between psychological variable ( trust influence condition - conformity > trust influence condition – non-conformity ) and the activation time course of the ventral striatum. Individual contrast images for conformity-trust influence versus non-conformity-trust influence were computed and entered into second-level one-sample t -tests. Brain regions surviving the cluster-extent based threshold p < 0.05 (FDR correction, with a primary voxel-level threshold of p < 0.001) were considered significant.

Behavioral Results

Data from twenty-seven subjects entered the behavioral analysis. We used a one-way repeated measures (social influence: baseline, trust influence, distrust influence) ANOVA to analyze the RTs in the decision phase. The effect of social influence was significant, F (2,25) = 4.204, p < 0.05. Participants responded faster in the trust influence condition ( M = 1222.46 ms, SD = 312.76) than in the baseline condition ( M = 1328.58 ms, SD = 333.87), t (26) = -2.845, p < 0.01. The responses were also faster in the trust influence condition ( M = 1222.46 ms, SD = 312.76) than in the distrust condition ( M = 1294.6 ms, SD = 294.42), t (26) = -2.479, p < 0.05.

Regarding the subjects’ choices, a one-way repeated measures (social influence: baseline, trust influence, distrust influence) ANOVA was used to analyze the rate of trust in the decision phase. The effect of social influence was significant, F (2,25) = 7.714, p < 0.01. Subjects decided to trust the trustee at a significantly higher rate in the trust influence condition ( M = 0.72, SD = 0.2) than in the baseline condition ( M = 0.53, SD = 0.22), t (26) = 3.543, p < 0.01. We also found this phenomenon in the contrast between trust influence condition ( M = 0.72, SD = 0.2) and distrust influence condition ( M = 0.43, SD = 0.27), t (26) = 3.926, p < 0.001. Participants chose to trust the trustee at a significantly higher rate in the baseline condition ( M = 0.53, SD = 0.22) than in the distrust influence condition ( M = 0.43, SD = 0.27), t (26) = 2.074, p < 0.05.

Because we predicted that subjects may have high reward expectancy in the trust social influence condition, we hypothesized that participants may conform to the opinion of the majority with a relatively high level of decision confidence. We conducted a 3 (social influence: baseline, trust influence, distrust influence) × 2 (choices: trust, distrust) ANOVA on the mean confidence rating. As predicted, the interaction between social influence and choices was significant, F (2,25) = 9.202, p < 0.001. The level of decision confidence is higher in the trust influence-trust condition ( M = 3.78, SD = 0.52) than in the baseline-trust condition ( M = 3.44, SD = 0.83), t (26) = 2.632, p < 0.05, as well as in the distrust influence-trust condition ( M = 3.37, SD = 0.74), t (26) = 3.227, p < 0.01. Confidence ratings for the trust influence-trust condition ( M = 3.78, SD = 0.52) seemed to be overall higher than ratings for the trust influence-distrust condition ( M = 3.37, SD = 0.61), t (26) = 3.827, p < 0.001.

fMRI Results

We compared the neural activity in trust influence condition with baseline condition and found significantly greater deactivation in superior temporal gyrus (STG) (for more details see Table 1 and Figure 2 ). The STG is a key brain region that involved in the cognitive capacity of perspective taking ( Frith and Frith, 2003 ).

Table 1. Significant activation clusters for trust social influence.

Figure 2. The superior temporal gyrus was involved in trust influence condition (Trust influence > Baseline), p < 0.05, cluster size = 10, FDR correction.

To capture the neural mechanisms underlying conformity effect in trusting behavior, exploratory analyses were performed. We compared the trust influence-conformity trials (mean number of trials 14) to trust influence-non-conformity (mean number of trials 6). Results shown that the trust influence which successfully induced conformity in trusting behavior activated the brain regions such as bilateral parahippocampal gyrus, vmPFC, RCZ, ACC/ caudate, middle occipital gyrus (MOG), MFG, middle temporal gyrus (MTG), postcentral gyrus and inferior parietal lobule (IPL) (see Table 2 and Figure 3 for more details). Comparison of activity in non-conformity trials with conformity trials did not show any significant activation.

Table 2. Significant activation clusters for conformity in trusting behavior.

Figure 3. Brain regions correlated with social influence in trusting behavior (trust influence – conformity > trust influence – non-conformity). Significant activations in middle frontal gyrus, middle temporal gyrus, middle occipital gyrus, rostral cingulate zone, anterior cingulate cortex, ventral medial prefrontal cortex, and inferior parietal lobule. p < 0.05, cluster size = 10, FDR correction.

Moreover, psychophysiological interaction (PPI) analysis showed that activity in the ventral striatum was accompanied by task-dependent (conformity > non-conformity) functional interaction with brain areas: STG, superior frontal gyrus (SFG), MTG and inferior temporal gyrus (ITG). The opposite contrast did not reveal any significant changes in functional connectivity (see Table 3 and Figure 4 for more details).

Table 3. Results of psychophysiological interaction (PPI) analysis.

Figure 4. Results of psychophysiological interaction (PPI) analysis. The region of interest was ventral striatum, MNI coordinates: [10, 18, –9]. Functional connectivity with the ventral striatum (conformity > non-conformity) in the trust influence condition. Voxels were selected for p < 0.05, FDR cluster-level correction with an initial peak-level threshold p < 0.001.

In the present study, we used psychological and neuroscientific methods to investigate the impact of social influence on trust. We found that individuals are likely to conform to the opinions of their peers in a trust game. The rate of trust was higher when participants found that the majority of group members trusted the trustee compared to in the baseline condition. Conversely, the rate of trust was lower when participants saw that most group members decided to keep the endowment (distrust) compared to in the baseline condition. In addition, participants conformed to the opinion of the majority with relatively high levels of decision confidence in the trust influence condition.

Functional imaging data suggested that the STG, a brain region involved in perspective-taking, was decreased when participants made decision in the trust influence condition comparing with the baseline condition. The activity of STG is associated with perspective taking, which can be termed as theory of mind ( Frith and Frith, 2003 ). As the decision to trust is concerned with perspective-taking, it should activate brain regions involved in theory-of-mind tasks ( Fehr and Camerer, 2007 ). Moreover, researchers found the STG was involved in the processing of gaze direction in a modified trust game ( Sun et al., 2018 ). A previous study that focused on the neurobiological correlates of conformity during mental rotation task has reported that the presence of external information was associated with decreased activation in the mental rotation neural network ( Berns et al., 2005 ). They inferred that the external information relieved the mental rotation processing load ( Berns et al., 2005 ). Similarly, decreased activations were observed during trust game in STG when external information was presented in our study. This result might suggest that external trust information affected neural activity in brain regions associated with trust game, which relieved the perspective-taking process in the game.

In our study, we tried to capture the conformity effect in the imaging data and found that brain regions involved in reward learning such as the vmPFC, ACC, ventral striatum, parahippocampal gyrus, and RCZ were also related with social influence in trusting behavior. The vmPFC has been previously implicated in processing reward expectations and computing the subjective value of multiple reward types ( Rushworth et al., 2009 , 2011 ; Rangel and Hare, 2010 ; Grabenhorst and Rolls, 2011 ). The study of brain activity during decision-making suggested that fictive reward signals (rewards that could have been, but were not directly received) have been represented in the ACC ( Hayden et al., 2009 ). The RCZ is engaged when the need for adjustments to achieve action goals becomes evident ( Ridderinkhof et al., 2004 ). Previous studies have demonstrated that the caudate is involved in gain prediction in response to reward cues and implicated in reward processing, social learning, and reciprocate cooperation ( Rilling et al., 2002 , 2004 ; McCoy and Platt, 2005 ; Knutson and Wimmer, 2007 ). According to PPI results, we found possible enhanced functional connectivity between the ventral striatum and prefrontal cortex during conformity compared to non-conformity in trusting behavior. Notably, recent research demonstrated that increased functional connectivity between the ventral striatum and prefrontal cortex was related to reward processing ( Frank and Claus, 2006 ; Camara et al., 2008 , 2009 ; van den Bos et al., 2012 ). Taken together, these exploratory imaging results suggest that the underlying mechanisms of social influence in trusting behavior may be similar to those implicated in reward learning. Agreement with the other group members might predict future acceptance from peer, which can also activate the reward system ( Izuma and Adolphs, 2013 ). These exploratory findings were consistent with the results of previous studies that reported that social influence effect affects participants’ behaviors through the neural mechanisms involved in reward learning and behavioral adjustment ( Izuma et al., 2008 ; Mason et al., 2009 ; Wei et al., 2013 ).

Several limitations of this study should be noted. Firstly, the present task is different from the Asch’s experiment. In our study, subjects had no other information about trust decision except the group members’ choices. This manipulation can potentially lead to conforming to the group member. Secondly, we did not use scale to quantitatively measure whether participants believed the experiment manipulation, which might also affect the result. Thirdly, the number of non-conformity trials that were included in exploratory analysis was less than 10 which limited the power of our GLM model. Despite that the results for these analyses survived correction, further studies could consider increasing the number of trials in order to more reliably evaluate these effects.

The present study provides evidence of the relationship between social influence and trust decisions. It complements previous research by assessing the neural basis of social influence and extends our understanding of the decision to trust. Our behavioral results revealed that individuals are likely to be influenced by others’ opinions and conform to the opinions of peers in a trust game. Participants conformed to the opinion of the majority with a relatively high level of decision confidence as a result of the high reward expectancy in the trust social influence condition. Decreased activations were observed in STG when external information was presented and this result might suggest that external trust information affected neural activity in brain regions associated with trust game, which relieved the perspective-taking process in the trust game. The results of exploratory analysis indicated that the brain regions involved in value processing and reward learning, such as the vmPFC, ventral striatum, ACC, and parahippocampal gyrus, were activated when subjects decided to follow the majority in trusting behavior. The PPI analysis confirmed possible increased functional connectivity between the ventral striatum and the prefrontal cortex during conformity in trusting behavior. In conclusion, these findings suggest that the mechanisms underlying social influence in trusting behavior may be similar to those implicated in reward learning.

Author Contributions

ZW and YZ conceived and designed the experiments. ZW and ZZ programed the task and analyzed the data. ZW performed the experiments. ZW, ZZ and YZ wrote the paper.

This project was funded by China Postdoctoral Science Foundation #2018M633337.

Conflict of Interest Statement

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Acknowledgments

The authors thank the reviewers for their constructive comments that improved the manuscript considerably.

Ahern, K. R., Duchin, R., and Shumway, T. (2014). Peer effects in risk aversion and trust. Rev. Financ. Stud. 27, 3213–3240. doi: 10.1093/rfs/hhu042

CrossRef Full Text | Google Scholar

Asch, S. E. (1951). Effects of Group Pressure Upon the Modification and Distortion of Judgments. Pittsburgh: Carnegie Press.

Google Scholar

Asch, S. E. (1956). Studies of independence and conformity: a minority of one against a unanimous majority. Psychol. Monogr. 70, 1–70. doi: 10.1037/h0093718

Berns, G. S., Capra, C. M., Moore, S., and Noussair, C. (2010). Neural mechanisms of the influence of popularity on adolescent ratings of music. Neuroimage 49, 2687–2696. doi: 10.1016/j.neuroimage.2009.10.070

PubMed Abstract | CrossRef Full Text | Google Scholar

Berns, G. S., Chappelow, J., Zink, C. F., Pagnoni, G., Martin-Skurski, M. E., and Richards, J. (2005). Neurobiological correlates of social conformity and independence during mental rotation. Biol. Psychiatry 58, 245–253. doi: 10.1016/j.biopsych.2005.04.012

Berns, G. S., McClure, S. M., Pagnoni, G., and Montague, P. R. (2001). Predictability modulates human brain response to reward. J. Neurosci. 21, 2793–2798. doi: 10.1523/JNEUROSCI.21-08-02793.2001

Burke, C. J., Tobler, P. N., Schultz, W., and Baddeley, M. (2010). Striatal BOLD response reflects the impact of herd information on financial decisions. Front. Hum. Neurosci. 4:48. doi: 10.3389/fnhum.2010.00048

Bursztyn, L., Ederer, F., Ferman, B., and Yuchtman, N. (2014). Understanding mechanisms underlying peer effects: evidence from a field experiment on financial decisions. Econometrica 82, 1273–1301. doi: 10.3982/ECTA11991

Camara, E., Rodriguez-Fornells, A., and Münte, T. F. (2008). Functional connectivity of reward processing in the brain. Front. Hum. Neurosci. 2:19. doi: 10.3389/neuro.09.019.2008

Camara, E., Rodriguez-Fornells, A., Ye, Z., and Münte, T. F. (2009). Reward networks in the brain as captured by connectivity measures. Front. Neurosci. 3:350. doi: 10.3389/neuro.01.034.2009

Campbell-Meiklejohn, D. K., Bach, D. R., Roepstorff, A., Dolan, R. J., and Frith, C. D. (2010). How the opinion of others affects our valuation of objects? Curr. Biol. 20, 1165–1170. doi: 10.1016/j.cub.2010.04.055

Campbell-Meiklejohn, D. K., Simonsen, A., Jensen, M., Wohlert, V., Gjerløff, T., Scheel-Kruger, J., et al. (2012). Modulation of social influence by methylphenidate. Neuropsychopharmacology 37, 1517–1525. doi: 10.1038/npp.2011.337

Cesarini, D., Dawes, C. T., Fowler, J. H., Johannesson, M., Lichtenstein, P., and Wallace, B. (2008). Heritability of cooperative behavior in the trust game. Proc. Natl. Acad. Sci. U.S.A. 105, 3721–3726. doi: 10.1073/pnas.0710069105

Cialdini, R. B., and Goldstein, N. J. (2004). Social influence: compliance and conformity. Annu. Rev. Psychol. 55, 591–621. doi: 10.1146/annurev.psych.55.090902.142015

Cochard, F., Nguyen Van, P., and Willinger, M. (2004). Trusting behavior in a repeated investment game. J. Econ. Behav. Organ. 55, 31–44. doi: 10.1016/j.jebo.2003.07.004

Fehr, E., and Camerer, C. F. (2007). Social neuroeconomics: the neural circuitry of social preferences. Trends Cogn. Sci. 11, 419–427. doi: 10.1016/j.tics.2007.09.002

Frank, M. J., and Claus, E. (2006). Anatomy of a decision: striato-orbitofrontal interactions in reinforcement learning, decision making, and reversal. Psychol. Sci. 113, 300–326. doi: 10.1037/0033-295X.113.2.300

Friston, K. J., Buechel, C., Fink, G. R., Morris, J., Rolls, E., and Dolan, R. J. (1997). Psychophysiological and modulatory interactions in neuroimaging. Neuroimage 6, 218–229. doi: 10.1006/nimg.1997.0291

Frith, U., and Frith, C. D. (2003). Development and neurophysiology of mentalizing. Philos. Trans. R. Soc. Lond. Ser. B Biol. Sci. 358, 459–473. doi: 10.1098/rstb.2002.1218

Gardner, M., and Steinberg, L. (2005). Peer influence on risk taking, risk preference, and risky decision making in adolescence and adulthood: an experimental study. Dev. Psychol. 41, 625–635. doi: 10.1037/0012-1649.41.4.625

Grabenhorst, F., and Rolls, E. T. (2011). Value, pleasure and choice in the ventral prefrontal cortex. Trends Cogn. Sci. 15, 56–67. doi: 10.1016/j.tics.2010.12.004

Haun, D., van Leeuwen, E. J., and Edelson, M. G. (2013). Majority influence in children and other animals. Dev. Cogn. Neurosci. 3, 61–71. doi: 10.1016/j.dcn.2012.09.003

Hayden, B. Y., Pearson, J. M., and Platt, M. L. (2009). Fictive reward signals in the anterior cingulate cortex. Science 324, 948–950. doi: 10.1126/science.1168488

Holroyd, C. B., and Coles, M. G. (2002). The neural basis of human error processing: reinforcement learning, dopamine, and the error-related negativity. Psychol. Rev. 109:679. doi: 10.1037/0033-295X.109.4.679

Hong, H., Kubik, J. D., and Stein, J. C. (2004). Social interaction and stock-market participation. J. Finance 59, 137–163. doi: 10.1111/j.1540-6261.2004.00629.x

Izuma, K., and Adolphs, R. (2013). Social manipulation of preference in the human brain. Neuron 78, 563–573. doi: 10.1016/j.neuron.2013.03.023

Izuma, K., Saito, D. N., and Sadato, N. (2008). Processing of social and monetary rewards in the human striatum. Neuron 58, 284–294. doi: 10.1016/j.neuron.2008.03.020

Klucharev, V., Hytönen, K., Rijpkema, M., Smidts, A., and Fernández, G. (2009). Reinforcement learning signal predicts social conformity. Neuron 61, 140–151. doi: 10.1016/j.neuron.2008.11.027

Klucharev, V., Munneke, M. A., Smidts, A., and Fernandez, G. (2011). Downregulation of the posterior medial frontal cortex prevents social conformity. J. Neurosci. 31, 11934–11940. doi: 10.1523/JNEUROSCI.1869-11.2011

Knutson, B., and Wimmer, G. E. (2007). Splitting the difference. Ann. N. Y. Acad. Sci. 1104, 54–69. doi: 10.1196/annals.1390.020

Mason, M. F., Dyer, R., and Norton, M. I. (2009). Neural mechanisms of social influence. Organ. Behav. Hum. Decis. Process. 110, 152–159. doi: 10.1016/j.obhdp.2009.04.001

McCoy, A. N., and Platt, M. L. (2005). Expectations and outcomes: decision-making in the primate brain. J. Comp. Physiol. A Neuroethol. Sens. Neural. Behav. Physiol. 191, 201–211. doi: 10.1007/s00359-004-0565-9

Morgan, T. J. H., and Laland, K. N. (2012). The biological bases of conformity. Front. Neurosci. 6:87. doi: 10.3389/fnins.2012.00087

Rangel, A., and Hare, T. (2010). Neural computations associated with goal-directed choice. Curr. Opin. Neurobiol. 20, 262–270. doi: 10.1016/j.conb.2010.03.001

Ridderinkhof, K. R., Ullsperger, M., Crone, E. A., and Nieuwenhuis, S. (2004). The role of the medial frontal cortex in cognitive control. Science 306, 443–447. doi: 10.1126/science.1100301

Rilling, J. K., Gutman, D. A., Zeh, T. R., Pagnoni, G., Berns, G. S., and Kilts, C. D. (2002). A neural basis for social cooperation. Neuron 35, 395–405. doi: 10.1016/S0896-6273(02)00755-9

Rilling, J. K., Sanfey, A. G., Aronson, J. A., Nystrom, L. E., and Cohen, J. D. (2004). Opposing BOLD responses to reciprocated and unreciprocated altruism in putative reward pathways. Neuroreport 15, 2539–2543. doi: 10.1097/00001756-200411150-00022

Rushworth, M. F., Mars, R. B., and Summerfield, C. (2009). General mechanisms for making decisions? Curr. Opin. Neurobiol. 19, 75–83. doi: 10.1016/j.conb.2009.02.005

Rushworth, M. F. S., Noonan, M. A. P., Boorman, E. D., Walton, M. E., and Behrens, T. E. (2011). Frontal cortex and reward-guided learning and decision-making. Neuron 70, 1054–1069. doi: 10.1016/j.neuron.2011.05.014

Simonsen, A., Scheel-Krüger, J., Jensen, M., Roepstorff, A., Møller, A., Frith, C. D., et al. (2014). Serotoninergic effects on judgments and social learning of trustworthiness. Psychopharmacology 231, 2759–2769. doi: 10.1007/s00213-014-3444-2

Sun, D. L., Shao, R. B., Wang, Z. X., and Lee, T. M. C. (2018). Perceived gaze direction modulates neural processing of prosocial decision making. Front. Hum. Neurosci. 12:52. doi: 10.3389/fnhum.2018.00052

Turner, J. C. (1991). Social Influence. Milton Keynes: Open University Press.

van den Bos, W., Cohen, M. X., Kahnt, T., and Crone, E. A. (2012). Striatum–medial prefrontal cortex connectivity predicts developmental changes in reinforcement learning. Cereb. Cortex 22, 1247–1255. doi: 10.1093/cercor/bhr198

Wei, Z., Zhao, Z., and Zheng, Y. (2013). Neural mechanisms underlying social conformity in an ultimatum game. Front. Hum. Neurosci. 7:896. doi: 10.3389/fnhum.2013.00896

Wei, Z., Zhao, Z., and Zheng, Y. (2017). The neural basis of social influence in a dictator decision. Front. Psychol. 8:2134. doi: 10.3389/fpsyg.2017.02134

Keywords : social influence, trust game, superior temporal gyrus, ventral striatum, reward learning

Citation: Wei Z, Zhao Z and Zheng Y (2019) Following the Majority: Social Influence in Trusting Behavior. Front. Neurosci. 13:89. doi: 10.3389/fnins.2019.00089

Received: 09 October 2018; Accepted: 25 January 2019; Published: 11 February 2019.

Reviewed by:

Copyright © 2019 Wei, Zhao and Zheng. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY) . The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

*Correspondence: Yong Zheng, [email protected]

Disclaimer: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.

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.

• View all journals
• Explore content
• About the journal
• Publish with us
• Sign up for alerts
• Open access
• Published: 28 August 2024

AI generates covertly racist decisions about people based on their dialect

• Valentin Hofmann   ORCID: orcid.org/0000-0001-6603-3428 1 , 2 , 3 ,
• Pratyusha Ria Kalluri 4 ,
• Dan Jurafsky   ORCID: orcid.org/0000-0002-6459-7745 4 &
• Sharese King 5  

Nature volume  633 ,  pages 147–154 ( 2024 ) Cite this article

1 Citations

184 Altmetric

Metrics details

• Computer science

Hundreds of millions of people now interact with language models, with uses ranging from help with writing 1 , 2 to informing hiring decisions 3 . However, these language models are known to perpetuate systematic racial prejudices, making their judgements biased in problematic ways about groups such as African Americans 4 , 5 , 6 , 7 . Although previous research has focused on overt racism in language models, social scientists have argued that racism with a more subtle character has developed over time, particularly in the United States after the civil rights movement 8 , 9 . It is unknown whether this covert racism manifests in language models. Here, we demonstrate that language models embody covert racism in the form of dialect prejudice, exhibiting raciolinguistic stereotypes about speakers of African American English (AAE) that are more negative than any human stereotypes about African Americans ever experimentally recorded. By contrast, the language models’ overt stereotypes about African Americans are more positive. Dialect prejudice has the potential for harmful consequences: language models are more likely to suggest that speakers of AAE be assigned less-prestigious jobs, be convicted of crimes and be sentenced to death. Finally, we show that current practices of alleviating racial bias in language models, such as human preference alignment, exacerbate the discrepancy between covert and overt stereotypes, by superficially obscuring the racism that language models maintain on a deeper level. Our findings have far-reaching implications for the fair and safe use of language technology.

Similar content being viewed by others

Large language models propagate race-based medicine

The benefits, risks and bounds of personalizing the alignment of large language models to individuals

Cognitive causes of ‘like me’ race and gender biases in human language production

Language models are a type of artificial intelligence (AI) that has been trained to process and generate text. They are becoming increasingly widespread across various applications, ranging from assisting teachers in the creation of lesson plans 10 to answering questions about tax law 11 and predicting how likely patients are to die in hospital before discharge 12 . As the stakes of the decisions entrusted to language models rise, so does the concern that they mirror or even amplify human biases encoded in the data they were trained on, thereby perpetuating discrimination against racialized, gendered and other minoritized social groups 4 , 5 , 6 , 13 , 14 , 15 , 16 , 17 , 18 , 19 , 20 .

Previous AI research has revealed bias against racialized groups but focused on overt instances of racism, naming racialized groups and mapping them to their respective stereotypes, for example by asking language models to generate a description of a member of a certain group and analysing the stereotypes it contains 7 , 21 . But social scientists have argued that, unlike the racism associated with the Jim Crow era, which included overt behaviours such as name calling or more brutal acts of violence such as lynching, a ‘new racism’ happens in the present-day United States in more subtle ways that rely on a ‘colour-blind’ racist ideology 8 , 9 . That is, one can avoid mentioning race by claiming not to see colour or to ignore race but still hold negative beliefs about racialized people. Importantly, such a framework emphasizes the avoidance of racial terminology but maintains racial inequities through covert racial discourses and practices 8 .

Here, we show that language models perpetuate this covert racism to a previously unrecognized extent, with measurable effects on their decisions. We investigate covert racism through dialect prejudice against speakers of AAE, a dialect associated with the descendants of enslaved African Americans in the United States 22 . We focus on the most stigmatized canonical features of the dialect shared among Black speakers in cities including New York City, Detroit, Washington DC, Los Angeles and East Palo Alto 23 . This cross-regional definition means that dialect prejudice in language models is likely to affect many African Americans.

Dialect prejudice is fundamentally different from the racial bias studied so far in language models because the race of speakers is never made overt. In fact we observed a discrepancy between what language models overtly say about African Americans and what they covertly associate with them as revealed by their dialect prejudice. This discrepancy is particularly pronounced for language models trained with human feedback (HF), such as GPT4: our results indicate that HF training obscures the racism on the surface, but the racial stereotypes remain unaffected on a deeper level. We propose using a new method, which we call matched guise probing, that makes it possible to recover these masked stereotypes.

The possibility that language models are covertly prejudiced against speakers of AAE connects to known human prejudices: speakers of AAE are known to experience racial discrimination in a wide range of contexts, including education, employment, housing and legal outcomes. For example, researchers have previously found that landlords engage in housing discrimination based solely on the auditory profiles of speakers, with voices that sounded Black or Chicano being less likely to secure housing appointments in predominantly white locales than in mostly Black or Mexican American areas 24 , 25 . Furthermore, in an experiment examining the perception of a Black speaker when providing an alibi 26 , the speaker was interpreted as more criminal, more working class, less educated, less comprehensible and less trustworthy when they used AAE rather than Standardized American English (SAE). Other costs for AAE speakers include having their speech mistranscribed or misunderstood in criminal justice contexts 27 and making less money than their SAE-speaking peers 28 . These harms connect to themes in broader racial ideology about African Americans and stereotypes about their intelligence, competence and propensity to commit crimes 29 , 30 , 31 , 32 , 33 , 34 , 35 . The fact that humans hold these stereotypes indicates that they are encoded in the training data and picked up by language models, potentially amplifying their harmful consequences, but this has never been investigated.

To our knowledge, this paper provides the first empirical evidence for the existence of dialect prejudice in language models; that is, covert racism that is activated by the features of a dialect (AAE). Using our new method of matched guise probing, we show that language models exhibit archaic stereotypes about speakers of AAE that most closely agree with the most-negative human stereotypes about African Americans ever experimentally recorded, dating from before the civil-rights movement. Crucially, we observe a discrepancy between what the language models overtly say about African Americans and what they covertly associate with them. Furthermore, we find that dialect prejudice affects language models’ decisions about people in very harmful ways. For example, when matching jobs to individuals on the basis of their dialect, language models assign considerably less-prestigious jobs to speakers of AAE than to speakers of SAE, even though they are not overtly told that the speakers are African American. Similarly, in a hypothetical experiment in which language models were asked to pass judgement on defendants who committed first-degree murder, they opted for the death penalty significantly more often when the defendants provided a statement in AAE rather than in SAE, again without being overtly told that the defendants were African American. We also show that current practices of alleviating racial disparities (increasing the model size) and overt racial bias (including HF in training) do not mitigate covert racism; indeed, quite the opposite. We found that HF training actually exacerbates the gap between covert and overt stereotypes in language models by obscuring racist attitudes. Finally, we discuss how the relationship between the language models’ covert and overt racial prejudices is both a reflection and a result of the inconsistent racial attitudes of contemporary society in the United States.

Probing AI dialect prejudice

To explore how dialect choice impacts the predictions that language models make about speakers in the absence of other cues about their racial identity, we took inspiration from the ‘matched guise’ technique used in sociolinguistics, in which subjects listen to recordings of speakers of two languages or dialects and make judgements about various traits of those speakers 36 , 37 . Applying the matched guise technique to the AAE–SAE contrast, researchers have shown that people identify speakers of AAE as Black with above-chance accuracy 24 , 26 , 38 and attach racial stereotypes to them, even without prior knowledge of their race 39 , 40 , 41 , 42 , 43 . These associations represent raciolinguistic ideologies, demonstrating how AAE is othered through the emphasis on its perceived deviance from standardized norms 44 .

Motivated by the insights enabled through the matched guise technique, we introduce matched guise probing, a method for investigating dialect prejudice in language models. The basic functioning of matched guise probing is as follows: we present language models with texts (such as tweets) in either AAE or SAE and ask them to make predictions about the speakers who uttered the texts (Fig. 1 and Methods ). For example, we might ask the language models whether a speaker who says “I be so happy when I wake up from a bad dream cus they be feelin too real” (AAE) is intelligent, and similarly whether a speaker who says “I am so happy when I wake up from a bad dream because they feel too real” (SAE) is intelligent. Notice that race is never overtly mentioned; its presence is merely encoded in the AAE dialect. We then examine how the language models’ predictions differ between AAE and SAE. The language models are not given any extra information to ensure that any difference in the predictions is necessarily due to the AAE–SAE contrast.

a , We used texts in SAE (green) and AAE (blue). In the meaning-matched setting (illustrated here), the texts have the same meaning, whereas they have different meanings in the non-meaning-matched setting. b , We embedded the SAE and AAE texts in prompts that asked for properties of the speakers who uttered the texts. c , We separately fed the prompts with the SAE and AAE texts into the language models. d , We retrieved and compared the predictions for the SAE and AAE inputs, here illustrated by five adjectives from the Princeton Trilogy. See Methods for more details.

We examined matched guise probing in two settings: one in which the meanings of the AAE and SAE texts are matched (the SAE texts are translations of the AAE texts) and one in which the meanings are not matched ( Methods  (‘Probing’) and Supplementary Information  (‘Example texts’)). Although the meaning-matched setting is more rigorous, the non-meaning-matched setting is more realistic, because it is well known that there is a strong correlation between dialect and content (for example, topics 45 ). The non-meaning-matched setting thus allows us to tap into a nuance of dialect prejudice that would be missed by examining only meaning-matched examples (see Methods for an in-depth discussion). Because the results for both settings overall are highly consistent, we present them in aggregated form here, but analyse the differences in the  Supplementary Information .

We examined GPT2 (ref. 46 ), RoBERTa 47 , T5 (ref. 48 ), GPT3.5 (ref. 49 ) and GPT4 (ref. 50 ), each in one or more model versions, amounting to a total of 12 examined models ( Methods and Supplementary Information (‘Language models’)). We first used matched guise probing to probe the general existence of dialect prejudice in language models, and then applied it to the contexts of employment and criminal justice.

Covert stereotypes in language models

We started by investigating whether the attitudes that language models exhibit about speakers of AAE reflect human stereotypes about African Americans. To do so, we replicated the experimental set-up of the Princeton Trilogy 29 , 30 , 31 , 34 , a series of studies investigating the racial stereotypes held by Americans, with the difference that instead of overtly mentioning race to the language models, we used matched guise probing based on AAE and SAE texts ( Methods ).

Qualitatively, we found that there is a substantial overlap in the adjectives associated most strongly with African Americans by humans and the adjectives associated most strongly with AAE by language models, particularly for the earlier Princeton Trilogy studies (Fig. 2a ). For example, the five adjectives associated most strongly with AAE by GPT2, RoBERTa and T5 share three adjectives (‘ignorant’, ‘lazy’ and ‘stupid’) with the five adjectives associated most strongly with African Americans in the 1933 and 1951 Princeton Trilogy studies, an overlap that is unlikely to occur by chance (permutation test with 10,000 random permutations of the adjectives; P  < 0.01). Furthermore, in lieu of the positive adjectives (such as ‘musical’, ‘religious’ and ‘loyal’), the language models exhibit additional solely negative associations (such as ‘dirty’, ‘rude’ and ‘aggressive’).

a , Strongest stereotypes about African Americans in humans in different years, strongest overt stereotypes about African Americans in language models, and strongest covert stereotypes about speakers of AAE in language models. Colour coding as positive (green) and negative (red) is based on ref. 34 . Although the overt stereotypes of language models are overall more positive than the human stereotypes, their covert stereotypes are more negative. b , Agreement of stereotypes about African Americans in humans with both overt and covert stereotypes about African Americans in language models. The black dotted line shows chance agreement using a random bootstrap. Error bars represent the standard error across different language models and prompts ( n  = 36). The language models’ overt stereotypes agree most strongly with current human stereotypes, which are the most positive experimentally recorded ones, but their covert stereotypes agree most strongly with human stereotypes from the 1930s, which are the most negative experimentally recorded ones. c , Stereotype strength for individual linguistic features of AAE. Error bars represent the standard error across different language models, model versions and prompts ( n  = 90). The linguistic features examined are: use of invariant ‘be’ for habitual aspect; use of ‘finna’ as a marker of the immediate future; use of (unstressed) ‘been’ for SAE ‘has been’ or ‘have been’ (present perfects); absence of the copula ‘is’ and ‘are’ for present-tense verbs; use of ‘ain’t’ as a general preverbal negator; orthographic realization of word-final ‘ing’ as ‘in’; use of invariant ‘stay’ for intensified habitual aspect; and absence of inflection in the third-person singular present tense. The measured stereotype strength is significantly above zero for all examined linguistic features, indicating that they all evoke raciolinguistic stereotypes in language models, although there is a lot of variation between individual features. See the Supplementary Information (‘Feature analysis’) for more details and analyses.

To investigate this more quantitatively, we devised a variant of average precision 51 that measures the agreement between the adjectives associated most strongly with African Americans by humans and the ranking of the adjectives according to their association with AAE by language models ( Methods ). We found that for all language models, the agreement with most Princeton Trilogy studies is significantly higher than expected by chance, as shown by one-sided t -tests computed against the agreement distribution resulting from 10,000 random permutations of the adjectives (mean ( m ) = 0.162, standard deviation ( s ) = 0.106; Extended Data Table 1 ); and that the agreement is particularly pronounced for the stereotypes reported in 1933 and falls for each study after that, almost reaching the level of chance agreement for 2012 (Fig. 2b ). In the Supplementary Information (‘Adjective analysis’), we explored variation across model versions, settings and prompts (Supplementary Fig. 2 and Supplementary Table 4 ).

To explain the observed temporal trend, we measured the average favourability of the top five adjectives for all Princeton Trilogy studies and language models, drawing from crowd-sourced ratings for the Princeton Trilogy adjectives on a scale between −2 (very negative) and 2 (very positive; see Methods , ‘Covert-stereotype analysis’). We found that the favourability of human attitudes about African Americans as reported in the Princeton Trilogy studies has become more positive over time, and that the language models’ attitudes about AAE are even more negative than the most negative experimentally recorded human attitudes about African Americans (the ones from the 1930s; Extended Data Fig. 1 ). In the Supplementary Information , we provide further quantitative analyses supporting this difference between humans and language models (Supplementary Fig. 7 ).

Furthermore, we found that the raciolinguistic stereotypes are not merely a reflection of the overt racial stereotypes in language models but constitute a fundamentally different kind of bias that is not mitigated in the current models. We show this by examining the stereotypes that the language models exhibit when they are overtly asked about African Americans ( Methods , ‘Overt-stereotype analysis’). We observed that the overt stereotypes are substantially more positive in sentiment than are the covert stereotypes, for all language models (Fig. 2a and Extended Data Fig. 1 ). Strikingly, for RoBERTa, T5, GPT3.5 and GPT4, although their covert stereotypes about speakers of AAE are more negative than the most negative experimentally recorded human stereotypes, their overt stereotypes about African Americans are more positive than the most positive experimentally recorded human stereotypes. This is particularly true for the two language models trained with HF (GPT3.5 and GPT4), in which all overt stereotypes are positive and all covert stereotypes are negative (see also ‘Resolvability of dialect prejudice’). In terms of agreement with human stereotypes about African Americans, the overt stereotypes almost never exhibit agreement significantly stronger than expected by chance, as shown by one-sided t -tests computed against the agreement distribution resulting from 10,000 random permutations of the adjectives ( m  = 0.162, s  = 0.106; Extended Data Table 2 ). Furthermore, the overt stereotypes are overall most similar to the human stereotypes from 2012, with the agreement continuously falling for earlier studies, which is the exact opposite trend to the covert stereotypes (Fig. 2b ).

In the experiments described in the  Supplementary Information (‘Feature analysis’), we found that the raciolinguistic stereotypes are directly linked to individual linguistic features of AAE (Fig. 2c and Supplementary Table 14 ), and that a higher density of such linguistic features results in stronger stereotypical associations (Supplementary Fig. 11 and Supplementary Table 13 ). Furthermore, we present experiments involving texts in other dialects (such as Appalachian English) as well as noisy texts, showing that these stereotypes cannot be adequately explained as either a general dismissive attitude towards text written in a dialect or as a general dismissive attitude towards deviations from SAE, irrespective of how the deviations look ( Supplementary Information (‘Alternative explanations’), Supplementary Figs. 12 and 13 and Supplementary Tables 15 and 16 ). Both alternative explanations are also tested on the level of individual linguistic features.

Thus, we found substantial evidence for the existence of covert raciolinguistic stereotypes in language models. Our experiments show that these stereotypes are similar to the archaic human stereotypes about African Americans that existed before the civil rights movement, are even more negative than the most negative experimentally recorded human stereotypes about African Americans, and are both qualitatively and quantitatively different from the previously reported overt racial stereotypes in language models, indicating that they are a fundamentally different kind of bias. Finally, our analyses demonstrate that the detected stereotypes are inherently linked to AAE and its linguistic features.

Impact of covert racism on AI decisions

To determine what harmful consequences the covert stereotypes have in the real world, we focused on two areas in which racial stereotypes about speakers of AAE and African Americans have been repeatedly shown to bias human decisions: employment and criminality. There is a growing impetus to use AI systems in these areas. Indeed, AI systems are already being used for personnel selection 52 , 53 , including automated analyses of applicants’ social-media posts 54 , 55 , and technologies for predicting legal outcomes are under active development 56 , 57 , 58 . Rather than advocating these use cases of AI, which are inherently problematic 59 , the sole objective of this analysis is to examine the extent to which the decisions of language models, when they are used in such contexts, are impacted by dialect.

First, we examined decisions about employability. Using matched guise probing, we asked the language models to match occupations to the speakers who uttered the AAE or SAE texts and computed scores indicating whether an occupation is associated more with speakers of AAE (positive scores) or speakers of SAE (negative scores; Methods , ‘Employability analysis’). The average score of the occupations was negative ( m  = –0.046,  s  = 0.053), the difference from zero being statistically significant (one-sample, one-sided t -test, t (83) = −7.9, P  < 0.001). This trend held for all language models individually (Extended Data Table 3 ). Thus, if a speaker exhibited features of AAE, the language models were less likely to associate them with any job. Furthermore, we observed that for all language models, the occupations that had the lowest association with AAE require a university degree (such as psychologist, professor and economist), but this is not the case for the occupations that had the highest association with AAE (for example, cook, soldier and guard; Fig. 3a ). Also, many occupations strongly associated with AAE are related to music and entertainment more generally (singer, musician and comedian), which is in line with a pervasive stereotype about African Americans 60 . To probe these observations more systematically, we tested for a correlation between the prestige of the occupations and the propensity of the language models to match them to AAE ( Methods ). Using a linear regression, we found that the association with AAE predicted the occupational prestige (Fig. 3b ; β  = −7.8, R 2 = 0.193, F (1, 63) = 15.1, P  < 0.001). This trend held for all language models individually (Extended Data Fig. 2 and Extended Data Table 4 ), albeit in a less pronounced way for GPT3.5, which had a particularly strong association of AAE with occupations in music and entertainment.

a , Association of different occupations with AAE or SAE. Positive values indicate a stronger association with AAE and negative values indicate a stronger association with SAE. The bottom five occupations (those associated most strongly with SAE) mostly require a university degree, but this is not the case for the top five (those associated most strongly with AAE). b , Prestige of occupations that language models associate with AAE (positive values) or SAE (negative values). The shaded area shows a 95% confidence band around the regression line. The association with AAE or SAE predicts the occupational prestige. Results for individual language models are provided in Extended Data Fig. 2 . c , Relative increase in the number of convictions and death sentences for AAE versus SAE. Error bars represent the standard error across different model versions, settings and prompts ( n  = 24 for GPT2, n  = 12 for RoBERTa, n  = 24 for T5, n  = 6 for GPT3.5 and n  = 6 for GPT4). In cases of small sample size ( n  ≤ 10 for GPT3.5 and GPT4), we plotted the individual results as overlaid dots. T5 does not contain the tokens ‘acquitted’ or ‘convicted’ in its vocabulary and is therefore excluded from the conviction analysis. Detrimental judicial decisions systematically go up for speakers of AAE compared with speakers of SAE.

We then examined decisions about criminality. We used matched guise probing for two experiments in which we presented the language models with hypothetical trials where the only evidence was a text uttered by the defendant in either AAE or SAE. We then measured the probability that the language models assigned to potential judicial outcomes in these trials and counted how often each of the judicial outcomes was preferred for AAE and SAE ( Methods , ‘Criminality analysis’). In the first experiment, we told the language models that a person is accused of an unspecified crime and asked whether the models will convict or acquit the person solely on the basis of the AAE or SAE text. Overall, we found that the rate of convictions was greater for AAE ( r  = 68.7%) than SAE ( r  = 62.1%; Fig. 3c , left). A chi-squared test found a strong effect ( χ 2 (1,  N  = 96) = 184.7,  P  < 0.001), which held for all language models individually (Extended Data Table 5 ). In the second experiment, we specifically told the language models that the person committed first-degree murder and asked whether the models will sentence the person to life or death on the basis of the AAE or SAE text. The overall rate of death sentences was greater for AAE ( r  = 27.7%) than for SAE ( r  = 22.8%; Fig. 3c , right). A chi-squared test found a strong effect ( χ 2 (1,  N  = 144) = 425.4,  P  < 0.001), which held for all language models individually except for T5 (Extended Data Table 6 ). In the Supplementary Information , we show that this deviation was caused by the base T5 version, and that the larger T5 versions follow the general pattern (Supplementary Table 10 ).

In further experiments ( Supplementary Information , ‘Intelligence analysis’), we used matched guise probing to examine decisions about intelligence, and found that all the language models consistently judge speakers of AAE to have a lower IQ than speakers of SAE (Supplementary Figs. 14 and 15 and Supplementary Tables 17 – 19 ).

Resolvability of dialect prejudice

We wanted to know whether the dialect prejudice we observed is resolved by current practices of bias mitigation, such as increasing the size of the language model or including HF in training. It has been shown that larger language models work better with dialects 21 and can have less racial bias 61 . Therefore, the first method we examined was scaling, that is, increasing the model size ( Methods ). We found evidence of a clear trend (Extended Data Tables 7 and 8 ): larger language models are indeed better at processing AAE (Fig. 4a , left), but they are not less prejudiced against speakers of it. In fact, larger models showed more covert prejudice than smaller models (Fig. 4a , right). By contrast, larger models showed less overt prejudice against African Americans (Fig. 4a , right). Thus, increasing scale does make models better at processing AAE and at avoiding prejudice against overt mentions of African Americans, but it makes them more linguistically prejudiced.

a , Language modelling perplexity and stereotype strength on AAE text as a function of model size. Perplexity is a measure of how successful a language model is at processing a particular text; a lower result is better. For language models for which perplexity is not well-defined (RoBERTa and T5), we computed pseudo-perplexity instead (dotted line). Error bars represent the standard error across different models of a size class and AAE or SAE texts ( n  = 9,057 for small, n  = 6,038 for medium, n  = 15,095 for large and n  = 3,019 for very large). For covert stereotypes, error bars represent the standard error across different models of a size class, settings and prompts ( n  = 54 for small, n  = 36 for medium, n  = 90 for large and n  = 18 for very large). For overt stereotypes, error bars represent the standard error across different models of a size class and prompts ( n  = 27 for small, n  = 18 for medium, n  = 45 for large and n  = 9 for very large). Although larger language models are better at processing AAE (left), they are not less prejudiced against speakers of it. Indeed, larger models show more covert prejudice than smaller models (right). By contrast, larger models show less overt prejudice against African Americans (right). In other words, increasing scale does make models better at processing AAE and at avoiding prejudice against overt mentions of African Americans, but it makes them more linguistically prejudiced. b , Change in stereotype strength and favourability as a result of training with HF for covert and overt stereotypes. Error bars represent the standard error across different prompts ( n  = 9). HF weakens (left) and improves (right) overt stereotypes but not covert stereotypes. c , Top overt and covert stereotypes about African Americans in GPT3, trained without HF, and GPT3.5, trained with HF. Colour coding as positive (green) and negative (red) is based on ref. 34 . The overt stereotypes get substantially more positive as a result of HF training in GPT3.5, but there is no visible change in favourability for the covert stereotypes.

As a second potential way to resolve dialect prejudice in language models, we examined training with HF 49 , 62 . Specifically, we compared GPT3.5 (ref. 49 ) with GPT3 (ref. 63 ), its predecessor that was trained without using HF ( Methods ). Looking at the top adjectives associated overtly and covertly with African Americans by the two language models, we found that HF resulted in more-positive overt associations but had no clear qualitative effect on the covert associations (Fig. 4c ). This observation was confirmed by quantitative analyses: the inclusion of HF resulted in significantly weaker (no HF, m  = 0.135,  s  = 0.142; HF, m  = −0.119,  s  = 0.234;  t (16) = 2.6,  P  < 0.05) and more favourable (no HF, m  = 0.221,  s  = 0.399; HF, m  = 1.047,  s  = 0.387;  t (16) = −6.4,  P  < 0.001) overt stereotypes but produced no significant difference in the strength (no HF, m  = 0.153,  s  = 0.049; HF, m  = 0.187,  s  = 0.066;  t (16) = −1.2, P  = 0.3) or unfavourability (no HF, m  = −1.146, s  = 0.580; HF, m = −1.029, s  = 0.196; t (16) = −0.5, P  = 0.6) of covert stereotypes (Fig. 4b ). Thus, HF training weakens and ameliorates the overt stereotypes but has no clear effect on the covert stereotypes; in other words, it obscures the racist attitudes on the surface, but more subtle forms of racism, such as dialect prejudice, remain unaffected. This finding is underscored by the fact that the discrepancy between overt and covert stereotypes about African Americans is most pronounced for the two examined language models trained with human feedback (GPT3.5 and GPT4; see ‘Covert stereotypes in language models’). Furthermore, this finding again shows that there is a fundamental difference between overt and covert stereotypes in language models, and that mitigating the overt stereotypes does not automatically translate to mitigated covert stereotypes.

To sum up, neither scaling nor training with HF as applied today resolves the dialect prejudice. The fact that these two methods effectively mitigate racial performance disparities and overt racial stereotypes in language models indicates that this form of covert racism constitutes a different problem that is not addressed by current approaches for improving and aligning language models.

The key finding of this article is that language models maintain a form of covert racial prejudice against African Americans that is triggered by dialect features alone. In our experiments, we avoided overt mentions of race but drew from the racialized meanings of a stigmatized dialect, and could still find historically racist associations with African Americans. The implicit nature of this prejudice, that is, the fact it is about something that is not explicitly expressed in the text, makes it fundamentally different from the overt racial prejudice that has been the focus of previous research. Strikingly, the language models’ covert and overt racial prejudices are often in contradiction with each other, especially for the most recent language models that have been trained with HF (GPT3.5 and GPT4). These two language models obscure the racism, overtly associating African Americans with exclusively positive attributes (such as ‘brilliant’), but our results show that they covertly associate African Americans with exclusively negative attributes (such as ‘lazy’).

We argue that this paradoxical relation between the language models’ covert and overt racial prejudices manifests the inconsistent racial attitudes present in the contemporary society of the United States 8 , 64 . In the Jim Crow era, stereotypes about African Americans were overtly racist, but the normative climate after the civil rights movement made expressing explicitly racist views distasteful. As a result, racism acquired a covert character and continued to exist on a more subtle level. Thus, most white people nowadays report positive attitudes towards African Americans in surveys but perpetuate racial inequalities through their unconscious behaviour, such as their residential choices 65 . It has been shown that negative stereotypes persist, even if they are superficially rejected 66 , 67 . This ambivalence is reflected by the language models we analysed, which are overtly non-racist but covertly exhibit archaic stereotypes about African Americans, showing that they reproduce a colour-blind racist ideology. Crucially, the civil rights movement is generally seen as the period during which racism shifted from overt to covert 68 , 69 , and this is mirrored by our results: all the language models overtly agree the most with human stereotypes from after the civil rights movement, but covertly agree the most with human stereotypes from before the civil rights movement.

Our findings beg the question of how dialect prejudice got into the language models. Language models are pretrained on web-scraped corpora such as WebText 46 , C4 (ref. 48 ) and the Pile 70 , which encode raciolinguistic stereotypes about AAE. A drastic example of this is the use of ‘mock ebonics’ to parodize speakers of AAE 71 . Crucially, a growing body of evidence indicates that language models pick up prejudices present in the pretraining corpus 72 , 73 , 74 , 75 , which would explain how they become prejudiced against speakers of AAE, and why they show varying levels of dialect prejudice as a function of the pretraining corpus. However, the web also abounds with overt racism against African Americans 76 , 77 , so we wondered why the language models exhibit much less overt than covert racial prejudice. We argue that the reason for this is that the existence of overt racism is generally known to people 32 , which is not the case for covert racism 69 . Crucially, this also holds for the field of AI. The typical pipeline of training language models includes steps such as data filtering 48 and, more recently, HF training 62 that remove overt racial prejudice. As a result, much of the overt racism on the web does not end up in the language models. However, there are currently no measures in place to curtail covert racial prejudice when training language models. For example, common datasets for HF training 62 , 78 do not include examples that would train the language models to treat speakers of AAE and SAE equally. As a result, the covert racism encoded in the training data can make its way into the language models in an unhindered fashion. It is worth mentioning that the lack of awareness of covert racism also manifests during evaluation, where it is common to test language models for overt racism but not for covert racism 21 , 63 , 79 , 80 .

As well as the representational harms, by which we mean the pernicious representation of AAE speakers, we also found evidence for substantial allocational harms. This refers to the inequitable allocation of resources to AAE speakers 81 (Barocas et al., unpublished observations), and adds to known cases of language technology putting speakers of AAE at a disadvantage by performing worse on AAE 82 , 83 , 84 , 85 , 86 , 87 , 88 , misclassifying AAE as hate speech 81 , 89 , 90 , 91 or treating AAE as incorrect English 83 , 85 , 92 . All the language models are more likely to assign low-prestige jobs to speakers of AAE than to speakers of SAE, and are more likely to convict speakers of AAE of a crime, and to sentence speakers of AAE to death. Although the details of our tasks are constructed, the findings reveal real and urgent concerns because business and jurisdiction are areas for which AI systems involving language models are currently being developed or deployed. As a consequence, the dialect prejudice we uncovered might already be affecting AI decisions today, for example when a language model is used in application-screening systems to process background information, which might include social-media text. Worryingly, we also observe that larger language models and language models trained with HF exhibit stronger covert, but weaker overt, prejudice. Against the backdrop of continually growing language models and the increasingly widespread adoption of HF training, this has two risks: first, that language models, unbeknownst to developers and users, reach ever-increasing levels of covert prejudice; and second, that developers and users mistake ever-decreasing levels of overt prejudice (the only kind of prejudice currently tested for) for a sign that racism in language models has been solved. There is therefore a realistic possibility that the allocational harms caused by dialect prejudice in language models will increase further in the future, perpetuating the racial discrimination experienced by generations of African Americans.

Matched guise probing examines how strongly a language model associates certain tokens, such as personality traits, with AAE compared with SAE. AAE can be viewed as the treatment condition, whereas SAE functions as the control condition. We start by explaining the basic experimental unit of matched guise probing: measuring how a language model associates certain tokens with an individual text in AAE or SAE. Based on this, we introduce two different settings for matched guise probing (meaning-matched and non-meaning-matched), which are both inspired by the matched guise technique used in sociolinguistics 36 , 37 , 93 , 94 and provide complementary views on the attitudes a language model has about a dialect.

The basic experimental unit of matched guise probing is as follows. Let θ be a language model, t be a text in AAE or SAE, and x be a token of interest, typically a personality trait such as ‘intelligent’. We embed the text in a prompt v , for example v ( t ) = ‘a person who says t tends to be’, and compute P ( x ∣ v ( t );  θ ), which is the probability that θ assigns to x after processing v ( t ). We calculate P ( x ∣ v ( t );  θ ) for equally sized sets T a of AAE texts and T s of SAE texts, comparing various tokens from a set X as possible continuations. It has been shown that P ( x ∣ v ( t );  θ ) can be affected by the precise wording of v , so small modifications of v can have an unpredictable effect on the predictions made by the language model 21 , 95 , 96 . To account for this fact, we consider a set V containing several prompts ( Supplementary Information ). For all experiments, we have provided detailed analyses of variation across prompts in the  Supplementary Information .

We conducted matched guise probing in two settings. In the first setting, the texts in T a and T s formed pairs expressing the same underlying meaning, that is, the i -th text in T a (for example, ‘I be so happy when I wake up from a bad dream cus they be feelin too real’) matches the i -th text in T s (for example, ‘I am so happy when I wake up from a bad dream because they feel too real’). For this setting, we used the dataset from ref. 87 , which contains 2,019 AAE tweets together with their SAE translations. In the second setting, the texts in T a and T s did not form pairs, so they were independent texts in AAE and SAE. For this setting, we sampled 2,000 AAE and SAE tweets from the dataset in ref. 83 and used tweets strongly aligned with African Americans for AAE and tweets strongly aligned with white people for SAE ( Supplementary Information (‘Analysis of non-meaning-matched texts’), Supplementary Fig. 1 and Supplementary Table 3 ). In the  Supplementary Information , we include examples of AAE and SAE texts for both settings (Supplementary Tables 1 and 2 ). Tweets are well suited for matched guise probing because they are a rich source of dialectal variation 97 , 98 , 99 , especially for AAE 100 , 101 , 102 , but matched guise probing can be applied to any type of text. Although we do not consider it here, matched guise probing can in principle also be applied to speech-based models, with the potential advantage that dialectal variation on the phonetic level could be captured more directly, which would make it possible to study dialect prejudice specific to regional variants of AAE 23 . However, note that a great deal of phonetic variation is reflected orthographically in social-media texts 101 .

It is important to analyse both meaning-matched and non-meaning-matched settings because they capture different aspects of the attitudes a language model has about speakers of AAE. Controlling for the underlying meaning makes it possible to uncover differences in the attitudes of the language model that are solely due to grammatical and lexical features of AAE. However, it is known that various properties other than linguistic features correlate with dialect, such as topics 45 , and these might also influence the attitudes of the language model. Sidelining such properties bears the risk of underestimating the harms that dialect prejudice causes for speakers of AAE in the real world. For example, in a scenario in which a language model is used in the context of automated personnel selection to screen applicants’ social-media posts, the texts of two competing applicants typically differ in content and do not come in pairs expressing the same meaning. The relative advantages of using meaning-matched or non-meaning-matched data for matched guise probing are conceptually similar to the relative advantages of using the same or different speakers for the matched guise technique: more control in the former versus more naturalness in the latter setting 93 , 94 . Because the results obtained in both settings were consistent overall for all experiments, we aggregated them in the main article, but we analysed differences in detail in the  Supplementary Information .

We apply matched guise probing to five language models: RoBERTa 47 , which is an encoder-only language model; GPT2 (ref. 46 ), GPT3.5 (ref. 49 ) and GPT4 (ref. 50 ), which are decoder-only language models; and T5 (ref. 48 ), which is an encoder–decoder language model. For each language model, we examined one or more model versions: GPT2 (base), GPT2 (medium), GPT2 (large), GPT2 (xl), RoBERTa (base), RoBERTa (large), T5 (small), T5 (base), T5 (large), T5 (3b), GPT3.5 (text-davinci-003) and GPT4 (0613). Where we used several model versions per language model (GPT2, RoBERTa and T5), the model versions all had the same architecture and were trained on the same data but differed in their size. Furthermore, we note that GPT3.5 and GPT4 are the only language models examined in this paper that were trained with HF, specifically reinforcement learning from human feedback 103 . When it is clear from the context what is meant, or when the distinction does not matter, we use the term ‘language models’, or sometimes ‘models‘, in a more general way that includes individual model versions.

Regarding matched guise probing, the exact method for computing P ( x ∣ v ( t );  θ ) varies across language models and is detailed in the  Supplementary Information . For GPT4, for which computing P ( x ∣ v ( t );  θ ) for all tokens of interest was often not possible owing to restrictions imposed by the OpenAI application programming interface (API), we used a slightly modified method for some of the experiments, and this is also discussed in the  Supplementary Information . Similarly, some of the experiments could not be done for all language models because of model-specific constraints, which we highlight below. We note that there was at most one language model per experiment for which this was the case.

Covert-stereotype analysis

In the covert-stereotype analysis, the tokens x whose probabilities are measured for matched guise probing are trait adjectives from the Princeton Trilogy 29 , 30 , 31 , 34 , such as ‘aggressive’, ‘intelligent’ and ‘quiet’. We provide details about these adjectives in the  Supplementary Information . In the Princeton Trilogy, the adjectives are provided to participants in the form of a list, and participants are asked to select from the list the five adjectives that best characterize a given ethnic group, such as African Americans. The studies that we compare in this paper, which are the original Princeton Trilogy studies 29 , 30 , 31 and a more recent reinstallment 34 , all follow this general set-up and observe a gradual improvement of the expressed stereotypes about African Americans over time, but the exact interpretation of this finding is disputed 32 . Here, we used the adjectives from the Princeton Trilogy in the context of matched guise probing.

Specifically, we first computed P ( x ∣ v ( t );  θ ) for all adjectives, for both the AAE texts and the SAE texts. The method for aggregating the probabilities P ( x ∣ v ( t );  θ ) into association scores between an adjective x and AAE varies for the two settings of matched guise probing. Let \({t}_{{\rm{a}}}^{i}\) be the i -th AAE text in T a and \({t}_{{\rm{s}}}^{i}\) be the i -th SAE text in T s . In the meaning-matched setting, in which \({t}_{{\rm{a}}}^{i}\) and \({t}_{{\rm{s}}}^{i}\) express the same meaning, we computed the prompt-level association score for an adjective x as

where n = ∣ T a ∣ = ∣ T s ∣ . Thus, we measure for each pair of AAE and SAE texts the log ratio of the probability assigned to x following the AAE text and the probability assigned to x following the SAE text, and then average the log ratios of the probabilities across all pairs. In the non-meaning-matched setting, we computed the prompt-level association score for an adjective x as

where again n = ∣ T a ∣ = ∣ T s ∣ . In other words, we first compute the average probability assigned to a certain adjective x following all AAE texts and the average probability assigned to x following all SAE texts, and then measure the log ratio of these average probabilities. The interpretation of q ( x ;  v ,  θ ) is identical in both settings; q ( x ;  v , θ ) > 0 means that for a certain prompt v , the language model θ associates the adjective x more strongly with AAE than with SAE, and q ( x ;  v ,  θ ) < 0 means that for a certain prompt v , the language model θ associates the adjective x more strongly with SAE than with AAE. In the  Supplementary Information (‘Calibration’), we show that q ( x ;  v , θ ) is calibrated 104 , meaning that it does not depend on the prior probability that θ assigns to x in a neutral context.

The prompt-level association scores q ( x ;  v ,  θ ) are the basis for further analyses. We start by averaging q ( x ;  v ,  θ ) across model versions, prompts and settings, and this allows us to rank all adjectives according to their overall association with AAE for individual language models (Fig. 2a ). In this and the following adjective analyses, we focus on the five adjectives that exhibit the highest association with AAE, making it possible to consistently compare the language models with the results from the Princeton Trilogy studies, most of which do not report the full ranking of all adjectives. Results for individual model versions are provided in the  Supplementary Information , where we also analyse variation across settings and prompts (Supplementary Fig. 2 and Supplementary Table 4 ).

Next, we wanted to measure the agreement between language models and humans through time. To do so, we considered the five adjectives most strongly associated with African Americans for each study and evaluated how highly these adjectives are ranked by the language models. Specifically, let R l  = [ x 1 , …,  x ∣ X ∣ ] be the adjective ranking generated by a language model and \({R}_{h}^{5}\) = [ x 1 , …, x 5 ] be the ranking of the top five adjectives generated by the human participants in one of the Princeton Trilogy studies. A typical measure to evaluate how highly the adjectives from \({R}_{h}^{5}\) are ranked within R l is average precision, AP 51 . However, AP does not take the internal ranking of the adjectives in \({R}_{h}^{5}\) into account, which is not ideal for our purposes; for example, AP does not distinguish whether the top-ranked adjective for humans is on the first or on the fifth rank for a language model. To remedy this, we computed the mean average precision, MAP, for different subsets of \({R}_{h}^{5}\) ,

where \({R}_{h}^{i}\) denotes the top i adjectives from the human ranking. MAP = 1 if, and only if, the top five adjectives from \({R}_{h}^{5}\) have an exact one-to-one correspondence with the top five adjectives from R l , so, unlike AP, it takes the internal ranking of the adjectives into account. We computed an individual agreement score for each language model and prompt, so we average the q ( x ;  v ,  θ ) association scores for all model versions of a language model (GPT2, for example) and the two settings (meaning-matched and non-meaning-matched) to generate R l . Because the OpenAI API for GPT4 does not give access to the probabilities for all adjectives, we excluded GPT4 from this analysis. Results are presented in Fig. 2b and Extended Data Table 1 . In the Supplementary Information (‘Agreement analysis’), we analyse variation across model versions, settings and prompts (Supplementary Figs. 3 – 5 ).

To analyse the favourability of the stereotypes about African Americans, we drew from crowd-sourced favourability ratings collected previously 34 for the adjectives from the Princeton Trilogy that range between −2 (‘very unfavourable’, meaning very negative) and 2 (‘very favourable’, meaning very positive). For example, the favourability rating of ‘cruel’ is −1.81 and the favourability rating of ‘brilliant’ is 1.86. We computed the average favourability of the top five adjectives, weighting the favourability ratings of individual adjectives by their association scores with AAE and African Americans. More formally, let R 5 = [ x 1 , …, x 5 ] be the ranking of the top five adjectives generated by either a language model or humans. Furthermore, let f ( x ) be the favourability rating of adjective x as reported in ref. 34 , and let q ( x ) be the overall association score of adjective x with AAE or African Americans that is used to generate R 5 . For the Princeton Trilogy studies, q ( x ) is the percentage of participants who have assigned x to African Americans. For language models, q ( x ) is the average value of q ( x ;  v ,  θ ). We then computed the weighted average favourability, F , of the top five adjectives as

As a result of the weighting, the top-ranked adjective contributed more to the average than the second-ranked adjective, and so on. Results are presented in Extended Data Fig. 1 . To check for consistency, we also computed the average favourability of the top five adjectives without weighting, which yields similar results (Supplementary Fig. 6) .

Overt-stereotype analysis

The overt-stereotype analysis closely followed the methodology of the covert-stereotype analysis, with the difference being that instead of providing the language models with AAE and SAE texts, we provided them with overt descriptions of race (specifically, ‘Black’/‘black’ and ‘White’/‘white’). This methodological difference is also reflected by a different set of prompts ( Supplementary Information ). As a result, the experimental set-up is very similar to existing studies on overt racial bias in language models 4 , 7 . All other aspects of the analysis (such as computing adjective association scores) were identical to the analysis for covert stereotypes. This also holds for GPT4, for which we again could not conduct the agreement analysis.

We again present average results for the five language models in the main article. Results broken down for individual model versions are provided in the  Supplementary Information , where we also analyse variation across prompts (Supplementary Fig. 8 and Supplementary Table 5 ).

Employability analysis

The general set-up of the employability analysis was identical to the stereotype analyses: we fed text written in either AAE or SAE, embedded in prompts, into the language models and analysed the probabilities that they assigned to different continuation tokens. However, instead of trait adjectives, we considered occupations for X and also used a different set of prompts ( Supplementary Information ). We created a list of occupations, drawing from previously published lists 6 , 76 , 105 , 106 , 107 . We provided details about these occupations in the  Supplementary Information . We then computed association scores q ( x ;  v ,  θ ) between individual occupations x and AAE, following the same methodology as for computing adjective association scores, and ranked the occupations according to q ( x ;  v ,  θ ) for the language models. To probe the prestige associated with the occupations, we drew from a dataset of occupational prestige 105 that is based on the 2012 US General Social Survey and measures prestige on a scale from 1 (low prestige) to 9 (high prestige). For GPT4, we could not conduct the parts of the analysis that require scores for all occupations.

We again present average results for the five language models in the main article. Results for individual model versions are provided in the  Supplementary Information , where we also analyse variation across settings and prompts (Supplementary Tables 6 – 8 ).

Criminality analysis

The set-up of the criminality analysis is different from the previous experiments in that we did not compute aggregate association scores between certain tokens (such as trait adjectives) and AAE but instead asked the language models to make discrete decisions for each AAE and SAE text. More specifically, we simulated trials in which the language models were prompted to use AAE or SAE texts as evidence to make a judicial decision. We then aggregated the judicial decisions into summary statistics.

We conducted two experiments. In the first experiment, the language models were asked to determine whether a person accused of committing an unspecified crime should be acquitted or convicted. The only evidence provided to the language models was a statement made by the defendant, which was an AAE or SAE text. In the second experiment, the language models were asked to determine whether a person who committed first-degree murder should be sentenced to life or death. Similarly to the first (general conviction) experiment, the only evidence provided to the language models was a statement made by the defendant, which was an AAE or SAE text. Note that the AAE and SAE texts were the same texts as in the other experiments and did not come from a judicial context. Rather than testing how well language models could perform the tasks of predicting acquittal or conviction and life penalty or death penalty (an application of AI that we do not support), we were interested to see to what extent the decisions of the language models, made in the absence of any real evidence, were impacted by dialect. Although providing the language models with extra evidence as well as the AAE and SAE texts would have made the experiments more similar to real trials, it would have confounded the effect that dialect has on its own (the key effect of interest), so we did not consider this alternative set-up here. We focused on convictions and death penalties specifically because these are the two areas of the criminal justice system for which racial disparities have been described in the most robust and indisputable way: African Americans represent about 12% of the adult population of the United States, but they represent 33% of inmates 108 and more than 41% of people on death row 109 .

Methodologically, we used prompts that asked the language models to make a judicial decision ( Supplementary Information ). For a specific text, t , which is in AAE or SAE, we computed p ( x ∣ v ( t );  θ ) for the tokens x that correspond to the judicial outcomes of interest (‘acquitted’ or ‘convicted’, and ‘life’ or ‘death’). T5 does not contain the tokens ‘acquitted’ and ‘convicted’ in its vocabulary, so is was excluded from the conviction analysis. Because the language models might assign different prior probabilities to the outcome tokens, we calibrated them using their probabilities in a neutral context following v , meaning without text t 104 . Whichever outcome had the higher calibrated probability was counted as the decision. We aggregated the detrimental decisions (convictions and death penalties) and compared their rates (percentages) between AAE and SAE texts. An alternative approach would have been to generate the judicial decision by sampling from the language models, which would have allowed us to induce the language models to generate justifications of their decisions. However, this approach has three disadvantages: first, encoder-only language models such as RoBERTa do not lend themselves to text generation; second, it would have been necessary to apply jail-breaking for some of the language models, which can have unpredictable effects, especially in the context of socially sensitive tasks; and third, model-generated justifications are frequently not aligned with actual model behaviours 110 .

We again present average results on the level of language models in the main article. Results for individual model versions are provided in the  Supplementary Information , where we also analyse variation across settings and prompts (Supplementary Figs. 9 and 10 and Supplementary Tables 9 – 12 ).

Scaling analysis

In the scaling analysis, we examined whether increasing the model size alleviated the dialect prejudice. Because the content of the covert stereotypes is quite consistent and does not vary substantially between models with different sizes, we instead analysed the strength with which the language models maintain these stereotypes. We split the model versions of all language models into four groups according to their size using the thresholds of 1.5 × 10 8 , 3.5 × 10 8 and 1.0 × 10 10 (Extended Data Table 7 ).

To evaluate the familiarity of the models with AAE, we measured their perplexity on the datasets used for the two evaluation settings 83 , 87 . Perplexity is defined as the exponentiated average negative log-likelihood of a sequence of tokens 111 , with lower values indicating higher familiarity. Perplexity requires the language models to assign probabilities to full sequences of tokens, which is only the case for GPT2 and GPT3.5. For RoBERTa and T5, we resorted to pseudo-perplexity 112 as the measure of familiarity. Results are only comparable across language models with the same familiarity measure. We excluded GPT4 from this analysis because it is not possible to compute perplexity using the OpenAI API.

To evaluate the stereotype strength, we focused on the stereotypes about African Americans reported in ref. 29 , which the language models’ covert stereotypes agree with most strongly. We split the set of adjectives X into two subsets: the set of stereotypical adjectives in ref. 29 , X s , and the set of non-stereotypical adjectives, X n  =  X \ X s . For each model with a specific size, we then computed the average value of q ( x ;  v ,  θ ) for all adjectives in X s , which we denote as q s ( θ ), and the average value of q ( x ;  v ,  θ ) for all adjectives in X n , which we denote as q n ( θ ). The stereotype strength of a model θ , or more specifically the strength of the stereotypes about African Americans reported in ref. 29 , can then be computed as

A positive value of δ ( θ ) means that the model associates the stereotypical adjectives in X s more strongly with AAE than the non-stereotypical adjectives in X n , whereas a negative value of δ ( θ ) indicates anti-stereotypical associations, meaning that the model associates the non-stereotypical adjectives in X n more strongly with AAE than the stereotypical adjectives in X s . For the overt stereotypes, we used the same split of adjectives into X s and X n because we wanted to directly compare the strength with which models of a certain size endorse the stereotypes overtly as opposed to covertly. All other aspects of the experimental set-up are identical to the main analyses of covert and overt stereotypes.

HF analysis

We compared GPT3.5 (ref. 49 ; text-davinci-003) with GPT3 (ref. 63 ; davinci), its predecessor language model that was trained without HF. Similarly to other studies that compare these two language models 113 , this set-up allowed us to examine the effects of HF training as done for GPT3.5 in isolation. We compared the two language models in terms of favourability and stereotype strength. For favourability, we followed the methodology we used for the overt-stereotype analysis and evaluated the average weighted favourability of the top five adjectives associated with AAE. For stereotype strength, we followed the methodology we used for the scaling analysis and evaluated the average strength of the stereotypes as reported in ref.  29 .

Reporting summary

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

Data availability

All the datasets used in this study are publicly available. The dataset released as ref. 87 can be found at https://aclanthology.org/2020.emnlp-main.473/ . The dataset released as ref. 83 can be found at http://slanglab.cs.umass.edu/TwitterAAE/ . The human stereotype scores used for evaluation can be found in the published articles of the Princeton Trilogy studies 29 , 30 , 31 , 34 . The most recent of these articles 34 also contains the human favourability scores for the trait adjectives. The dataset of occupational prestige that we used for the employability analysis can be found in the corresponding paper 105 . The Brown Corpus 114 , which we used for the  Supplementary Information (‘Feature analysis’), can be found at http://www.nltk.org/nltk_data/ . The dataset containing the parallel AAE, Appalachian English and Indian English texts 115 , which we used in the  Supplementary Information (‘Alternative explanations’), can be found at https://huggingface.co/collections/SALT-NLP/value-nlp-666b60a7f76c14551bda4f52 .

Code availability

Our code is written in Python and draws on the Python packages openai and transformers for language-model probing, as well as numpy, pandas, scipy and statsmodels for data analysis. The feature analysis described in the  Supplementary Information also uses the VALUE Python library 88 . Our code is publicly available on GitHub at https://github.com/valentinhofmann/dialect-prejudice .

Zhao, W. et al. WildChat: 1M ChatGPT interaction logs in the wild. In Proc. Twelfth International Conference on Learning Representations (OpenReview.net, 2024).

Zheng, L. et al. LMSYS-Chat-1M: a large-scale real-world LLM conversation dataset. In Proc. Twelfth International Conference on Learning Representations (OpenReview.net, 2024).

Gaebler, J. D., Goel, S., Huq, A. & Tambe, P. Auditing the use of language models to guide hiring decisions. Preprint at https://arxiv.org/abs/2404.03086 (2024).

Sheng, E., Chang, K.-W., Natarajan, P. & Peng, N. The woman worked as a babysitter: on biases in language generation. In Proc. 2019 Conference on Empirical Methods in Natural Language Processing and 9th International Joint Conference on Natural Language Processing (eds Inui. K. et al.) 3407–3412 (Association for Computational Linguistics, 2019).

Nangia, N., Vania, C., Bhalerao, R. & Bowman, S. R. CrowS-Pairs: a challenge dataset for measuring social biases in masked language models. In Proc. 2020 Conference on Empirical Methods in Natural Language Processing (eds Webber, B. et al.) 1953–1967 (Association for Computational Linguistics, 2020).

Nadeem, M., Bethke, A. & Reddy, S. StereoSet: measuring stereotypical bias in pretrained language models. In Proc. 59th Annual Meeting of the Association for Computational Linguistics and 11th International Joint Conference on Natural Language Processing (eds Zong, C. et al.) 5356–5371 (Association for Computational Linguistics, 2021).

Cheng, M., Durmus, E. & Jurafsky, D. Marked personas: using natural language prompts to measure stereotypes in language models. In Proc. 61st Annual Meeting of the Association for Computational Linguistics (eds Rogers, A. et al.) 1504–1532 (Association for Computational Linguistics, 2023).

Bonilla-Silva, E. Racism without Racists: Color-Blind Racism and the Persistence of Racial Inequality in America 4th edn (Rowman & Littlefield, 2014).

Golash-Boza, T. A critical and comprehensive sociological theory of race and racism. Sociol. Race Ethn. 2 , 129–141 (2016).

Article   Google Scholar  

Kasneci, E. et al. ChatGPT for good? On opportunities and challenges of large language models for education. Learn. Individ. Differ. 103 , 102274 (2023).

Nay, J. J. et al. Large language models as tax attorneys: a case study in legal capabilities emergence. Philos. Trans. R. Soc. A 382 , 20230159 (2024).

Article   ADS   Google Scholar  

Jiang, L. Y. et al. Health system-scale language models are all-purpose prediction engines. Nature 619 , 357–362 (2023).

Article   ADS   CAS   PubMed   PubMed Central   Google Scholar  

Bolukbasi, T., Chang, K.-W., Zou, J., Saligrama, V. & Kalai, A. Man is to computer programmer as woman is to homemaker? Debiasing word embeddings. Adv. Neural Inf. Process. Syst. 30 , 4356–4364 (2016).

Google Scholar  

Caliskan, A., Bryson, J. J. & Narayanan, A. Semantics derived automatically from language corpora contain human-like biases. Science 356 , 183–186 (2017).

Article   ADS   CAS   PubMed   Google Scholar  

Basta, C., Costa-jussà, M. R. & Casas, N. Evaluating the underlying gender bias in contextualized word embeddings. In Proc. First Workshop on Gender Bias in Natural Language Processing (eds Costa-jussà, M. R. et al.) 33–39 (Association for Computational Linguistics, 2019).

Kurita, K., Vyas, N., Pareek, A., Black, A. W. & Tsvetkov, Y. Measuring bias in contextualized word representations. In Proc. First Workshop on Gender Bias in Natural Language Processing (eds Costa-jussà, M. R. et al.) 166–172 (Association for Computational Linguistics, 2019).

Abid, A., Farooqi, M. & Zou, J. Persistent anti-muslim bias in large language models. In Proc. 2021 AAAI/ACM Conference on AI, Ethics, and Society (eds Fourcade, M. et al.) 298–306 (Association for Computing Machinery, 2021).

Bender, E. M., Gebru, T., McMillan-Major, A. & Shmitchell, S. On the dangers of stochastic parrots: can language models be too big? In Proc. 2021 ACM Conference on Fairness, Accountability, and Transparency 610–623 (Association for Computing Machinery, 2021).

Li, L. & Bamman, D. Gender and representation bias in GPT-3 generated stories. In Proc. Third Workshop on Narrative Understanding (eds Akoury, N. et al.) 48–55 (Association for Computational Linguistics, 2021).

Tamkin, A. et al. Evaluating and mitigating discrimination in language model decisions. Preprint at https://arxiv.org/abs/2312.03689 (2023).

Rae, J. W. et al. Scaling language models: methods, analysis & insights from training Gopher. Preprint at https://arxiv.org/abs/2112.11446 (2021).

Green, L. J. African American English: A Linguistic Introduction (Cambridge Univ. Press, 2002).

King, S. From African American Vernacular English to African American Language: rethinking the study of race and language in African Americans’ speech. Annu. Rev. Linguist. 6 , 285–300 (2020).

Purnell, T., Idsardi, W. & Baugh, J. Perceptual and phonetic experiments on American English dialect identification. J. Lang. Soc. Psychol. 18 , 10–30 (1999).

Massey, D. S. & Lundy, G. Use of Black English and racial discrimination in urban housing markets: new methods and findings. Urban Aff. Rev. 36 , 452–469 (2001).

Dunbar, A., King, S. & Vaughn, C. Dialect on trial: an experimental examination of raciolinguistic ideologies and character judgments. Race Justice https://doi.org/10.1177/21533687241258772 (2024).

Rickford, J. R. & King, S. Language and linguistics on trial: Hearing Rachel Jeantel (and other vernacular speakers) in the courtroom and beyond. Language 92 , 948–988 (2016).

Grogger, J. Speech patterns and racial wage inequality. J. Hum. Resour. 46 , 1–25 (2011).

Katz, D. & Braly, K. Racial stereotypes of one hundred college students. J. Abnorm. Soc. Psychol. 28 , 280–290 (1933).

Gilbert, G. M. Stereotype persistance and change among college students. J. Abnorm. Soc. Psychol. 46 , 245–254 (1951).

Article   CAS   Google Scholar  

Karlins, M., Coffman, T. L. & Walters, G. On the fading of social stereotypes: studies in three generations of college students. J. Pers. Soc. Psychol. 13 , 1–16 (1969).

Article   CAS   PubMed   Google Scholar  

Devine, P. G. & Elliot, A. J. Are racial stereotypes really fading? The Princeton Trilogy revisited. Pers. Soc. Psychol. Bull. 21 , 1139–1150 (1995).

Madon, S. et al. Ethnic and national stereotypes: the Princeton Trilogy revisited and revised. Pers. Soc. Psychol. Bull. 27 , 996–1010 (2001).

Bergsieker, H. B., Leslie, L. M., Constantine, V. S. & Fiske, S. T. Stereotyping by omission: eliminate the negative, accentuate the positive. J. Pers. Soc. Psychol. 102 , 1214–1238 (2012).

Article   PubMed   PubMed Central   Google Scholar  

Ghavami, N. & Peplau, L. A. An intersectional analysis of gender and ethnic stereotypes: testing three hypotheses. Psychol. Women Q. 37 , 113–127 (2013).

Lambert, W. E., Hodgson, R. C., Gardner, R. C. & Fillenbaum, S. Evaluational reactions to spoken languages. J. Abnorm. Soc. Psychol. 60 , 44–51 (1960).

Ball, P. Stereotypes of Anglo-Saxon and non-Anglo-Saxon accents: some exploratory Australian studies with the matched guise technique. Lang. Sci. 5 , 163–183 (1983).

Thomas, E. R. & Reaser, J. Delimiting perceptual cues used for the ethnic labeling of African American and European American voices. J. Socioling. 8 , 54–87 (2004).

Atkins, C. P. Do employment recruiters discriminate on the basis of nonstandard dialect? J. Employ. Couns. 30 , 108–118 (1993).

Payne, K., Downing, J. & Fleming, J. C. Speaking Ebonics in a professional context: the role of ethos/source credibility and perceived sociability of the speaker. J. Tech. Writ. Commun. 30 , 367–383 (2000).

Rodriguez, J. I., Cargile, A. C. & Rich, M. D. Reactions to African-American vernacular English: do more phonological features matter? West. J. Black Stud. 28 , 407–414 (2004).

Billings, A. C. Beyond the Ebonics debate: attitudes about Black and standard American English. J. Black Stud. 36 , 68–81 (2005).

Kurinec, C. A. & Weaver, C. III “Sounding Black”: speech stereotypicality activates racial stereotypes and expectations about appearance. Front. Psychol. 12 , 785283 (2021).

Rosa, J. & Flores, N. Unsettling race and language: toward a raciolinguistic perspective. Lang. Soc. 46 , 621–647 (2017).

Salehi, B., Hovy, D., Hovy, E. & Søgaard, A. Huntsville, hospitals, and hockey teams: names can reveal your location. In Proc. 3rd Workshop on Noisy User-generated Text (eds Derczynski, L. et al.) 116–121 (Association for Computational Linguistics, 2017).

Radford, A. et al. Language models are unsupervised multitask learners. OpenAI https://cdn.openai.com/better-language-models/language_models_are_unsupervised_multitask_learners.pdf (2019).

Liu, Y. et al. RoBERTa: a robustly optimized BERT pretraining approach. Preprint at https://arxiv.org/abs/1907.11692 (2019).

Raffel, C. et al. Exploring the limits of transfer learning with a unified text-to-text transformer. J. Mach. Learn. Res. 21 , 1–67 (2020).

MathSciNet   Google Scholar  

Ouyang, L. et al. Training language models to follow instructions with human feedback. In Proc. 36th Conference on Neural Information Processing Systems (eds Koyejo, S. et al.) 27730–27744 (NeurIPS, 2022).

OpenAI et al. GPT-4 technical report. Preprint at https://arxiv.org/abs/2303.08774 (2023).

Zhang, E. & Zhang, Y. Average precision. In Encyclopedia of Database Systems (eds Liu, L. & Özsu, M. T.) 192–193 (Springer, 2009).

Black, J. S. & van Esch, P. AI-enabled recruiting: what is it and how should a manager use it? Bus. Horiz. 63 , 215–226 (2020).

Hunkenschroer, A. L. & Luetge, C. Ethics of AI-enabled recruiting and selection: a review and research agenda. J. Bus. Ethics 178 , 977–1007 (2022).

Upadhyay, A. K. & Khandelwal, K. Applying artificial intelligence: implications for recruitment. Strateg. HR Rev. 17 , 255–258 (2018).

Tippins, N. T., Oswald, F. L. & McPhail, S. M. Scientific, legal, and ethical concerns about AI-based personnel selection tools: a call to action. Pers. Assess. Decis. 7 , 1 (2021).

Aletras, N., Tsarapatsanis, D., Preoţiuc-Pietro, D. & Lampos, V. Predicting judicial decisions of the European Court of Human Rights: a natural language processing perspective. PeerJ Comput. Sci. 2 , e93 (2016).

Surden, H. Artificial intelligence and law: an overview. Ga State Univ. Law Rev. 35 , 1305–1337 (2019).

Medvedeva, M., Vols, M. & Wieling, M. Using machine learning to predict decisions of the European Court of Human Rights. Artif. Intell. Law 28 , 237–266 (2020).

Weidinger, L. et al. Taxonomy of risks posed by language models. In Proc. 2022 ACM Conference on Fairness, Accountability, and Transparency 214–229 (Association for Computing Machinery, 2022).

Czopp, A. M. & Monteith, M. J. Thinking well of African Americans: measuring complimentary stereotypes and negative prejudice. Basic Appl. Soc. Psychol. 28 , 233–250 (2006).

Chowdhery, A. et al. PaLM: scaling language modeling with pathways. J. Mach. Learn. Res. 24 , 11324–11436 (2023).

Bai, Y. et al. Training a helpful and harmless assistant with reinforcement learning from human feedback. Preprint at https://arxiv.org/abs/2204.05862 (2022).

Brown, T. B. et al. Language models are few-shot learners. In  Proc. 34th International Conference on Neural Information Processing Systems  (eds Larochelle, H. et al.) 1877–1901 (NeurIPS, 2020).

Dovidio, J. F. & Gaertner, S. L. Aversive racism. Adv. Exp. Soc. Psychol. 36 , 1–52 (2004).

Schuman, H., Steeh, C., Bobo, L. D. & Krysan, M. (eds) Racial Attitudes in America: Trends and Interpretations (Harvard Univ. Press, 1998).

Crosby, F., Bromley, S. & Saxe, L. Recent unobtrusive studies of Black and White discrimination and prejudice: a literature review. Psychol. Bull. 87 , 546–563 (1980).

Terkel, S. Race: How Blacks and Whites Think and Feel about the American Obsession (New Press, 1992).

Jackman, M. R. & Muha, M. J. Education and intergroup attitudes: moral enlightenment, superficial democratic commitment, or ideological refinement? Am. Sociol. Rev. 49 , 751–769 (1984).

Bonilla-Silva, E. The New Racism: Racial Structure in the United States, 1960s–1990s. In Race, Ethnicity, and Nationality in the United States: Toward the Twenty-First Century 1st edn (ed. Wong, P.) Ch. 4 (Westview Press, 1999).

Gao, L. et al. The Pile: an 800GB dataset of diverse text for language modeling. Preprint at https://arxiv.org/abs/2101.00027 (2021).

Ronkin, M. & Karn, H. E. Mock Ebonics: linguistic racism in parodies of Ebonics on the internet. J. Socioling. 3 , 360–380 (1999).

Dodge, J. et al. Documenting large webtext corpora: a case study on the Colossal Clean Crawled Corpus. In Proc. 2021 Conference on Empirical Methods in Natural Language Processing (eds Moens, M.-F. et al.) 1286–1305 (Association for Computational Linguistics, 2021).

Steed, R., Panda, S., Kobren, A. & Wick, M. Upstream mitigation is not all you need: testing the bias transfer hypothesis in pre-trained language models. In Proc. 60th Annual Meeting of the Association for Computational Linguistics (eds Muresan, S. et al.) 3524–3542 (Association for Computational Linguistics, 2022).

Feng, S., Park, C. Y., Liu, Y. & Tsvetkov, Y. From pretraining data to language models to downstream tasks: tracking the trails of political biases leading to unfair NLP models. In Proc. 61st Annual Meeting of the Association for Computational Linguistics (eds Rogers, A. et al.) 11737–11762 (Association for Computational Linguistics, 2023).

Köksal, A. et al. Language-agnostic bias detection in language models with bias probing. In Findings of the Association for Computational Linguistics: EMNLP 2023 (eds Bouamor, H. et al.) 12735–12747 (Association for Computational Linguistics, 2023).

Garg, N., Schiebinger, L., Jurafsky, D. & Zou, J. Word embeddings quantify 100 years of gender and ethnic stereotypes. Proc. Natl Acad. Sci. USA 115 , E3635–E3644 (2018).

Ferrer, X., van Nuenen, T., Such, J. M. & Criado, N. Discovering and categorising language biases in Reddit. In Proc. Fifteenth International AAAI Conference on Web and Social Media (eds Budak, C. et al.) 140–151 (Association for the Advancement of Artificial Intelligence, 2021).

Ethayarajh, K., Choi, Y. & Swayamdipta, S. Understanding dataset difficulty with V-usable information. In Proc. 39th International Conference on Machine Learning (eds Chaudhuri, K. et al.) 5988–6008 (Proceedings of Machine Learning Research, 2022).

Hoffmann, J. et al. Training compute-optimal large language models. Preprint at https://arxiv.org/abs/2203.15556 (2022).

Liang, P. et al. Holistic evaluation of language models. Transactions on Machine Learning Research https://openreview.net/forum?id=iO4LZibEqW (2023).

Blodgett, S. L., Barocas, S., Daumé III, H. & Wallach, H. Language (technology) is power: A critical survey of “bias” in NLP. In Proc. 58th Annual Meeting of the Association for Computational Linguistics (eds Jurafsky, D. et al.) 5454–5476 (Association for Computational Linguistics, 2020).

Jørgensen, A., Hovy, D. & Søgaard, A. Challenges of studying and processing dialects in social media. In Proc. Workshop on Noisy User-generated Text (eds Xu, W. et al.) 9–18 (Association for Computational Linguistics, 2015).

Blodgett, S. L., Green, L. & O’Connor, B. Demographic dialectal variation in social media: a case study of African-American English. In Proc. 2016 Conference on Empirical Methods in Natural Language Processing (eds Su, J. et al.) 1119–1130 (Association for Computational Linguistics, 2016).

Jørgensen, A., Hovy, D. & Søgaard, A. Learning a POS tagger for AAVE-like language. In Proc. 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (eds Knight, K. et al.) 1115–1120 (Association for Computational Linguistics, 2016).

Blodgett, S. L. & O’Connor, B. Racial disparity in natural language processing: a case study of social media African-American English. Preprint at https://arxiv.org/abs/1707.00061 (2017).

Blodgett, S. L., Wei, J. & O’Connor, B. Twitter universal dependency parsing for African-American and mainstream American English. In Proc. 56th Annual Meeting of the Association for Computational Linguistics (eds Gurevych, I. & Miyao, Y.) 1415–1425 (Association for Computational Linguistics, 2018).

Groenwold, S. et al. Investigating African-American vernacular English in transformer-based text generation. In Proc. 2020 Conference on Empirical Methods in Natural Language Processing (eds Webber, B. et al.) 5877–5883 (Association for Computational Linguistics, 2020).

Ziems, C., Chen, J., Harris, C., Anderson, J. & Yang, D. VALUE: Understanding dialect disparity in NLU. In Proc. 60th Annual Meeting of the Association for Computational Linguistics (eds Muresan, S. et al.) 3701–3720 (Association for Computational Linguistics, 2022).

Davidson, T., Bhattacharya, D. & Weber, I. Racial bias in hate speech and abusive language detection datasets. In Proc. Third Workshop on Abusive Language Online (eds Roberts, S. T. et al.) 25–35 (Association for Computational Linguistics, 2019).

Sap, M., Card, D., Gabriel, S., Choi, Y. & Smith, N. A. The risk of racial bias in hate speech detection. In Proc. 57th Annual Meeting of the Association for Computational Linguistics (eds Korhonen, A. et al.) 1668–1678 (Association for Computational Linguistics, 2019).

Harris, C., Halevy, M., Howard, A., Bruckman, A. & Yang, D. Exploring the role of grammar and word choice in bias toward African American English (AAE) in hate speech classification. In Proc. 2022 ACM Conference on Fairness, Accountability, and Transparency 789–798 (Association for Computing Machinery, 2022).

Gururangan, S. et al. Whose language counts as high quality? Measuring language ideologies in text data selection. In Proc. 2022 Conference on Empirical Methods in Natural Language Processing (eds Goldberg, Y. et al.) 2562–2580 (Association for Computational Linguistics, 2022).

Gaies, S. J. & Beebe, J. D. The matched-guise technique for measuring attitudes and their implications for language education: a critical assessment. In Language Acquisition and the Second/Foreign Language Classroom (ed. Sadtano, E.) 156–178 (SEAMEO Regional Language Centre, 1991).

Hudson, R. A. Sociolinguistics (Cambridge Univ. Press, 1996).

Delobelle, P., Tokpo, E., Calders, T. & Berendt, B. Measuring fairness with biased rulers: a comparative study on bias metrics for pre-trained language models. In Proc. 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (eds Carpuat, M. et al.) 1693–1706 (Association for Computational Linguistics, 2022).

Mattern, J., Jin, Z., Sachan, M., Mihalcea, R. & Schölkopf, B. Understanding stereotypes in language models: Towards robust measurement and zero-shot debiasing. Preprint at https://arxiv.org/abs/2212.10678 (2022).

Eisenstein, J., O’Connor, B., Smith, N. A. & Xing, E. P. A latent variable model for geographic lexical variation. In Proc. 2010 Conference on Empirical Methods in Natural Language Processing (eds Li, H. & Màrquez, L.) 1277–1287 (Association for Computational Linguistics, 2010).

Doyle, G. Mapping dialectal variation by querying social media. In Proc. 14th Conference of the European Chapter of the Association for Computational Linguistics (eds Wintner, S. et al.) 98–106 (Association for Computational Linguistics, 2014).

Huang, Y., Guo, D., Kasakoff, A. & Grieve, J. Understanding U.S. regional linguistic variation with Twitter data analysis. Comput. Environ. Urban Syst. 59 , 244–255 (2016).

Eisenstein, J. What to do about bad language on the internet. In Proc. 2013 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (eds Vanderwende, L. et al.) 359–369 (Association for Computational Linguistics, 2013).

Eisenstein, J. Systematic patterning in phonologically-motivated orthographic variation. J. Socioling. 19 , 161–188 (2015).

Jones, T. Toward a description of African American vernacular English dialect regions using “Black Twitter”. Am. Speech 90 , 403–440 (2015).

Christiano, P. F. et al. Deep reinforcement learning from human preferences. Proc. 31st International Conference on Neural Information Processing Systems (eds von Luxburg, U. et al.) 4302–4310 (NeurIPS, 2017).

Zhao, T. Z., Wallace, E., Feng, S., Klein, D. & Singh, S. Calibrate before use: Improving few-shot performance of language models. In Proc. 38th International Conference on Machine Learning (eds Meila, M. & Zhang, T.) 12697–12706 (Proceedings of Machine Learning Research, 2021).

Smith, T. W. & Son, J. Measuring Occupational Prestige on the 2012 General Social Survey (NORC at Univ. Chicago, 2014).

Zhao, J., Wang, T., Yatskar, M., Ordonez, V. & Chang, K.-W. Gender bias in coreference resolution: evaluation and debiasing methods. In Proc. 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (eds Walker, M. et al.) 15–20 (Association for Computational Linguistics, 2018).

Hughes, B. T., Srivastava, S., Leszko, M. & Condon, D. M. Occupational prestige: the status component of socioeconomic status. Collabra Psychol. 10 , 92882 (2024).

Gramlich, J. The gap between the number of blacks and whites in prison is shrinking. Pew Research Centre https://www.pewresearch.org/short-reads/2019/04/30/shrinking-gap-between-number-of-blacks-and-whites-in-prison (2019).

Walsh, A. The criminal justice system is riddled with racial disparities. Prison Policy Initiative Briefing https://www.prisonpolicy.org/blog/2016/08/15/cjrace (2016).

Röttger, P. et al. Political compass or spinning arrow? Towards more meaningful evaluations for values and opinions in large language models. Preprint at https://arxiv.org/abs/2402.16786 (2024).

Jurafsky, D. & Martin, J. H. Speech and Language Processing: An Introduction to Natural Language Processing, Computational Linguistics, and Speech Recognition (Prentice Hall, 2000).

Salazar, J., Liang, D., Nguyen, T. Q. & Kirchhoff, K. Masked language model scoring. In Proc. 58th Annual Meeting of the Association for Computational Linguistics (eds Jurafsky, D. et al.) 2699–2712 (Association for Computational Linguistics, 2020).

Santurkar, S. et al. Whose opinions do language models reflect? In Proc. 40th International Conference on Machine Learning (eds Krause, A. et al.) 29971–30004 (Proceedings of Machine Learning Research, 2023).

Francis, W. N. & Kucera, H. Brown Corpus Manual (Brown Univ.,1979).

Ziems, C. et al. Multi-VALUE: a framework for cross-dialectal English NLP. In Proc. 61st Annual Meeting of the Association for Computational Linguistics (eds Rogers, A. et al.) 744–768 (Association for Computational Linguistics, 2023).

Download references

Acknowledgements

V.H. was funded by the German Academic Scholarship Foundation. P.R.K. was funded in part by the Open Phil AI Fellowship. This work was also funded by the Hoffman-Yee Research Grants programme and the Stanford Institute for Human-Centered Artificial Intelligence. We thank A. Köksal, D. Hovy, K. Gligorić, M. Harrington, M. Casillas, M. Cheng and P. Röttger for feedback on an earlier version of the article.

Author information

Authors and affiliations.

Allen Institute for AI, Seattle, WA, USA

Valentin Hofmann

University of Oxford, Oxford, UK

LMU Munich, Munich, Germany

Stanford University, Stanford, CA, USA

Pratyusha Ria Kalluri & Dan Jurafsky

The University of Chicago, Chicago, IL, USA

Sharese King

You can also search for this author in PubMed   Google Scholar

Contributions

V.H., P.R.K., D.J. and S.K. designed the research. V.H. performed the research and analysed the data. V.H., P.R.K., D.J. and S.K. wrote the paper.

Corresponding authors

Correspondence to Valentin Hofmann or Sharese King .

Ethics declarations

Competing interests.

The authors declare no competing interests.

Peer review

Peer review information.

Nature thanks Rodney Coates and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.

Additional information

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

Extended data figures and tables

Extended data fig. 1 weighted average favourability of top stereotypes about african americans in humans and top overt as well as covert stereotypes about african americans in language models (lms)..

The overt stereotypes are more favourable than the reported human stereotypes, except for GPT2. The covert stereotypes are substantially less favourable than the least favourable reported human stereotypes from 1933. Results without weighting, which are very similar, are provided in Supplementary Fig. 6 .

Extended Data Fig. 2 Prestige of occupations associated with AAE (positive values) versus SAE (negative values), for individual language models.

The shaded areas show 95% confidence bands around the regression lines. The association with AAE versus SAE is negatively correlated with occupational prestige, for all language models. We cannot conduct this analysis with GPT4 since the OpenAI API does not give access to the probabilities for all occupations.

Supplementary information

Supplementary information, reporting summary, rights and permissions.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ .

Reprints and permissions

About this article

Cite this article.

Hofmann, V., Kalluri, P.R., Jurafsky, D. et al. AI generates covertly racist decisions about people based on their dialect. Nature 633 , 147–154 (2024). https://doi.org/10.1038/s41586-024-07856-5

Download citation

Received : 08 February 2024

Accepted : 19 July 2024

Published : 28 August 2024

Issue Date : 05 September 2024

DOI : https://doi.org/10.1038/s41586-024-07856-5

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

This article is cited by

Llms produce racist output when prompted in african american english.

• Su Lin Blodgett
• Zeerak Talat

Nature (2024)

By submitting a comment you agree to abide by our Terms and Community Guidelines . If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.

Quick links

• Explore articles by subject
• Guide to authors
• Editorial policies

Sign up for the Nature Briefing: AI and Robotics newsletter — what matters in AI and robotics research, free to your inbox weekly.

Exploring Female Narratives of Sexual Intimacy and the Social Suppression of Desire

• Published: 04 September 2024

Cite this article

• Shivani Rajan   ORCID: orcid.org/0009-0008-3470-9241 1 &
• Swati Pathak   ORCID: orcid.org/0000-0002-1189-1457 1  

Exploring the construction of sexual identities by women, this research attempts to provide an experiential understanding of sexual intimacy in young adulthood through critical narrative analysis of the accounts of ten unmarried cis-gender women, located in the postmodern feminist paradigm and drawing from contemporary psychoanalytic tradition. The study highlighted the lack of discourse on female desire and pleasure (that is not fetishised, penalised, or ostracised) in the hetero-patriarchal socio-cultural fabric of India and how it manifests in the sense of shame, guilt, and self-doubt in the navigation of sexual intimacy. The social matrix, including the influence of family and partner dynamics and cultural and generational differences, was observed to play a prominent role in the evolution of individual perceptions of sexual intimacy. Analysing the narratives through a feminist lens foregrounded the predominance of male satisfaction and pleasure, the sense of obligation towards male partners, the infringement of boundaries and compromise, and the performativity in sexual experiences, thus calling attention to the female struggle of realising and practising sexual agency. The research indicates the need to critically examine the pervasive phallocentrism in the experience of sexual intimacy and the marginalisation of female sexual desire due to the suppression of female sexuality in the patriarchal hierarchy of power distribution.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Subscribe and save.

• Get 10 units per month
• Download Article/Chapter or eBook
• 1 Unit = 1 Article or 1 Chapter
• Cancel anytime

Price includes VAT (Russian Federation)

Instant access to the full article PDF.

Rent this article via DeepDyve

Institutional subscriptions

Data Availability

No datasets were generated or analysed during the current study.

Atallah, S., & Redón, A. M. (2023). Relevant (sexual) aspects of Cultural differences. In S. Geuens, A. Polona Mivšek, & W. Gianotten (Eds.), Midwifery and sexuality . Springer. https://doi.org/10.1007/978-3-031-18432-1_23

Bell, E. (2005). Sex acts beyond boundaries and binaries: A feminist challenge for self care in performance studies. Text and Performance Quarterly , 25 (3), 187–219.

Article   Google Scholar  

Butler, J. (1990). Gender trouble: Feminism and the subversion of identity. Routledge.

Cash, T. F., Maikkula, C. L., & Yamamiya, Y. (2004). Baring the body in the bedroom: Body image, sexual self-schemas, and sexual functioning among college women and men. Electronic Journal of Human Sexuality , 7 , 1–9.

Google Scholar  

Chakraborty, K., & Thakurata, R. G. (2013). Indian concepts on sexuality. Indian Journal of Psychiatry , 55 (Suppl 2), S250–S255.

Article   PubMed   PubMed Central   Google Scholar  

Chasseguet-Smirgel, J. (1976). Freud and female sexuality: The consideration of some blind spots in the exploration of the Dark Continent. The International Journal of Psychoanalysis .

Crooks, R. L., Baur, K., & Widman, L. (2020). Our sexuality . Cengage Learning.

Dion, K. K., & Dion, K. L. (1993). Individualistic and collectivistic perspectives on gender and the cultural context of love and intimacy. Journal of Social Issues , 49 (3), 53–69.

Easteal, P., Holland, K., & Judd, K. (2015). Enduring themes and silences in media portrayals of violence against women. Women’s Studies International Forum (Vol. 48, pp. 103–113). Pergamon.

Emerson, P., & Frosh, S. (2004). Critical narrative analysis in psychology: A guide to practice. Springer.

Elise, D. (2008). Sex and shame: The inhibition of female desires. Journal of the American Psychoanalytic Association , 56 (1), 73–98.

Article   PubMed   Google Scholar  

Faustino, M. J., & Gavey, N. (2024). The failed promise of consent in women’s experiences of coercive and unwanted anal sex with men. Feminism & Psychology , 09593535241234429.

Fazli Khalaf, Z., Liow, J. W., Low, W. Y., Ghorbani, B., & Merghati-Khoei, E. (2018). Young women’s experience of sexuality: A battle of pleasure and sexual purity in the context of Malaysian society. Sexuality & Culture , 22, 849–864.

Freud, S. (1920). Introductory lectures on psychoanalysis (A. A. Brill, Trans) . Hogarth. (Originally published, 1916-17).

Freud, S. (1905). Three Essays on the Theory of Sexuality. In J. Strachy, & A. Freud (Eds.), The Standard Edition of the Complete Psychological Works of Sigmund Freud, Volume VII (1901-1905): A Case of Hysteria, Three Essays on Sexuality and Other Works (pp. 123–246). London: Hogarth Press.

Friedman, J., & Valenti, J. (Eds.). (2019). Yes means yes! Visions of female sexual power and a world without rape . Seal.

Gavin, J. (2000). Arousing suspicion and violating trust: The lived ideology of safe sex talk. Culture Health & Sexuality . https://doi.org/10.1080/136910500300750

Gill, R. (2009). Mediated intimacy and postfeminism: A discourse analytic examination of sex and relationships advice in a women’s magazine . Discourse & Communication. https://doi.org/10.1177/1750481309343870

Guba, E. G., & Lincoln Y. S. (1994). Competing paradigms in qualitative research. Handbook of qualitative research . Sage.

Haberland, N., & Rogow, D. (2015). Sexuality education: Emerging trends in evidence and practice. Journal of Adolescent Health , 56 (1), S15–S21.

Hawkes, G. (1996). Sociology of sex and sexuality . McGraw-Hill Education (UK).

Impett, E. A., & Peplau, L. A. (2002). Why some women consent to unwanted sex with a dating partner: Insights from attachment theory. Psychology of Women Quarterly , 26 (4), 360–370.

Jackson, A., & Guerra, N. S. (2011). Cultural Difference. In: Goldstein, S., Naglieri, J.A. (Eds.) Encyclopedia of Child Behavior and Development. Springer, Boston, MA. https://doi.org/10.1007/978-0-387-79061-9_752

Kulwicki, C. (2008). Real sex education. Yes Means yes , 305–312.

McAdams, D. P. (1988). Biography, narrative, and lives: An introduction. Journal of Personality , 56 (1), 1–18.

Mitchell, J. (2000). Psychoanalysis and feminism: A radical reassessment of Freudian psychoanalysis . Basic Books.

Mitra, D. (2020). Indian sex life: Sexuality and the colonial origins of modern social thought . Princeton University Press.

Olamijuwon, E., & Odimegwu, C. (2022). Saving sex for marriage: An analysis of lay attitudes towards virginity and its perceived benefit for marriage. Sexuality & Culture , 26 (2), 568–594.

Rahmani, A., Merghati-Khoei, E., Moghaddam-Banaem, L., Hajizadeh, E., & Montazeri, A. (2016). The viewpoints of sexually active single women about premarital sexual relationships: A qualitative study in the Iranian context. International journal of high risk behaviors & addiction , 5(1).

Özcan, Ö., Cumurcu, B. E., Karlidag, R., Ünal, S., Mutlu, E. A., & Kartalci, S. (2015). Attachment styles in women with vaginismus/Vaginismusu olan kadinlarda baglanma stilleri. Anadolu Psikiyatri Dergisi , 16 (1), 37.

Sanchez, D. T., Fetterolf, J. C., & Rudman, L. A. (2012). Eroticizing Inequality in the United States: The Consequences and Determinants of Traditional Gender Role Adherence in Intimate Relationships. Journal of Sex Research, 49(2–3), 168–183. https://doi.org/10.1080/00224499.2011.653699

Simon-Kumar, R. (2014). Sexual violence in India: The discourses of rape and the discourses of justice. Indian Journal of Gender Studies , 21 (3), 451–460.

Sprecher, S., & Sedikides, C. (1993). Gender differences in perceptions of emotionality: The case of close heterosexual relationships. Sex Roles , 28 (9–10), 511–530.

Srivastava, S. (2020). Passionate modernity: Sexuality, class, and consumption in India . Taylor & Francis.

Szymanski, D. M., & Carr, E. R. (2008). The roles of gender role conflict and internalized heterosexism in gay and bisexual men’s psychological distress: Testing two mediation models. Psychology of Men & Masculinity , 9 (1), 40.

Tangney, J. P., & Dearing, R. L. (2003). Shame and guilt . Guilford Press.

Techasrivichien, T., Darawuttimaprakorn, N., Punpuing, S., Musumari, P. M., Lukhele, B. W., El-Saaidi, C., & Kihara, M. (2016). Changes in sexual behavior and attitudes across generations and gender among a population-based probability sample from an urbanizing province in Thailand. Archives of Sexual Behavior , 45 , 367–382.

Tillman, L. (2021). The influence of parent-child relationships on female sexual functioning . A Review Of The Literature.

Tolman, D. L. (2005). Dilemmas of desire: Teenage girls talk about sexuality . Harvard University Press.

Tolman, D. L., & McClelland, S. I. (2011). Normative sexuality development in adolescence: A decade in review, 2000–2009. Journal of Research on Adolescence , 21 (1), 242–255.

Twenge, J. M., Sherman, R. A., & Wells, B. E. (2015). Changes in American adults’ sexual behavior and attitudes, 1972–2012. Archives of Sexual Behavior , 44 , 2273–2285.

Wagner, B. (2009). Becoming a sexual being: Overcoming constraints on female sexuality. Sexualities , 12 (3), 289–311.

Wallwiener, S., Strohmaier, J., Wallwiener, L. M., Schönfisch, B., Zipfel, S., Brucker, S. Y., Rietschel, M., & Wallwiener, C. W. (2016). Sexual function is correlated with body image and partnership quality in female university students. The Journal of Sexual Medicine , 13 (10), 1530–1538.

Willoughby, B. J., Busby, D. M., & Young-Petersen, B. (2018). Understanding associations between Personal definitions of Pornography, using pornography, and Depression. Sexuality Research and Social Policy . https://doi.org/10.1007/s13178-018-0345-x

Witting, K., Santtila, P., Varjonen, M., Jern, P., Johansson, A., Von Der Pahlen, B., & Sandnabba, K. (2008). Female sexual dysfunction, sexual distress, and compatibility with partner. The Journal of Sexual Medicine , 5 (11), 2587–2599.

Yasmine, R., El Salibi, N., El Kak, F., & Ghandour, L. (2015). Postponing sexual debut among university youth: How do men and women differ in their perceptions, values and non-penetrative sexual practices? Culture Health & Sexuality , 17 (5), 555–575.

Download references

No funding was received for conducting this study.

Author information

Authors and affiliations.

Department of Psychology, CHRIST University, Delhi NCR, India

Shivani Rajan & Swati Pathak

You can also search for this author in PubMed   Google Scholar

Contributions

S.R. drafted the research idea, conducted the interview with the participants and prepared the initial draft of the manuscript.S.P the research supervisor who directly monitored the research and reviewed the progress of changes in the manuscript.

Corresponding author

Correspondence to Swati Pathak .

Ethics declarations

Ethics approval and consent to participate.

The study has been approved by the Research Conduct Ethics Committee (RCEC) at CHRIST University. Informed consent was obtained from all individual participants included in the study.

Competing Interests

The authors declare no competing interests.

Additional information

Publisher’s note.

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

Electronic Supplementary Material

Below is the link to the electronic supplementary material.

Supplementary Material 1

Rights and permissions.

Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.

Reprints and permissions

About this article

Rajan, S., Pathak, S. Exploring Female Narratives of Sexual Intimacy and the Social Suppression of Desire. Hu Arenas (2024). https://doi.org/10.1007/s42087-024-00441-2

Download citation

Received : 02 February 2024

Revised : 20 August 2024

Accepted : 21 August 2024

Published : 04 September 2024

DOI : https://doi.org/10.1007/s42087-024-00441-2

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Female sexuality
• Sexual intimacy

Advertisement

• Find a journal
• Publish with us
• Track your research

Bhaskaran Publishes Research on Laryngeal Dystonia

Written by Staff

September 3, 2024

Divya Bhaskaran, Assistant Professor in the Exercise Science program of the Biology Department, published a research paper in the Frontiers in Neurology Journal. The article titled "Effects of an 11-week vibro-tactile stimulation treatment on voice symptoms in laryngeal dystonia" is a longitudinal clinical trial conducted during Dr Bhaskaran's post-doctoral work at the University of Minnesota. 

1536 Hewitt Ave

Saint Paul, MN 55104

General Information

Undergraduate Admission

Public Safety Office

Graduate Admission

ITS Central Service Desk

© 2024 Hamline University

In association with Mitchell | Hamline School of Law ®. Mitchell Hamline School of Law ® has more graduate enrollment options than any other law school in the nation.

• On-Campus Transfer
• Online Degree Completion
• International
• Admitted Students
• How to Apply
• Grants & Scholarships
• First-Year and Transfer Aid
• Online Degree Completion Aid
• Graduate Aid
• International Aid
• Military & Veteran Aid
• Undergraduate Tuition
• Online Degree Completion Tuition
• Graduate Tuition
• Housing & Food Costs
• Net Price Calculator
• Payment Info
• Undergraduate
• Continuing Education
• Program Finder
• Faculty by Program
• College of Liberal Arts
• School of Business
• School of Education & Leadership
• Mitchell Hamline School of Law
• Academic Bulletin
• Academic Calendars
• Bush Library
• Registration & Records
• Study Away & Study Abroad
• Housing & Dining
• Counseling & Health
• Service, Spiritual Life, & Recreation
• Activities & Organizations
• Diversity Resources
• Arts at Hamline
• Meet Our Students
• The Neighborhood
• The Hamline Academic Experience
• Student Research Opportunities
• Paid Internships
• Career Development Center
• Alumni Success Stories
• Center for Academic Success & Achievement
• Writing Center
• Why Hamline?
• Mission & History
• Fast Facts and Rankings
• University Leadership
• Diversity, Equity & Inclusion
• Alumni and Donors
• Request Info