University Place

Peer Influences on Crime and Deviance

[Carolyn Field, Associate Professor, Department of Sociology, Edgewood College]

Hello, thank you all for coming.

My name is Carolyn Field, and I am an associate professor of sociology at Edgewood College and the Criminal Justice Program director.

I’m also the, sort of, faculty advisor for the Criminal Justice Association, which is our Criminal Justice club for any students interested in criminal justice or just interested in crime and justice in general.

So, thank you all for coming. Tonight’s event was sponsored by the College of Arts and Sciences, and also the Social Science Department where the Criminal Justice major is housed in that department.

We have a vibrant department in Social Science with many majors to choose from: Criminal justice, Political Science, Economics, Sociology. We also have human services. So, it’s a great multidisciplinary department, and I’m so happy that you are here tonight to support this department event.

Stick around afterwards, buy a Criminal Justice T-shirt, get some free cookies that were baked by the Criminal Justice students and myself, and we’re first gonna have a little introduction of the speaker by the president of the Criminal Justice Club, Haley Massey. And so, I will hand off the microphone to her. Thank you all for coming.

[applause]

[Haley Massey, President, Criminal Justice Club, Edgewood College]

Hello, good evening, everybody. My name is Haley Massey.

As Carolyn said, I’m the president of the Criminal Justice Association here at Edgewood. In the audience with me tonight, I have my fellow officers. Sammy Gander, our historian. Joe Disparte, our vice president. And Austin Schumacher, our secretary.

We’re pleased to welcome Dr. Cesar Rebellon.

Did I say it right? Okay.

To present his research.

Dr. Rebellon, I’m gonna say it a bunch of times, too, grew up mostly in Florida and received his undergraduate degree in psychology and sociology at Rice University.

He went on to receive his PhD in sociology at Emory University in Georgia, and he’s now a professor of sociology at the University of New Hampshire.

His research on juvenile crime concerns the causes and correlates of juvenile delinquency, with a particular focus on peer influences on crime and deviance.

His research has been funded by the National Science Foundation and has been published extensively in the most highly respected academic journals in his field. Those journals include Criminology, Deviant Behavior, Justice Quarterly, the Journal of Criminal Justice, and the Journal of Quantitative Criminology. Please help me give a warm welcome to Dr. Cesar Rebellon.

[applause]

[Dr. Cesar Rebellon, Associate Professor, Department or Sociology, University of New Hampshire]

Thank you, Haley. Can you guys hear me?

And thank you also, Carolyn. It’s wonderful to be here. I appreciate you guys coming out tonight. And what I’m gonna talk about tonight is this issue of whether delinquents jump like lemmings? Peer influences on juvenile crime and deviance.

The short version of the talk is basically as follows: That there is a major theoretical perspective in criminology and sociology, sometimes referred to as differential association theory, sometimes referred to as social learning theory, and these perspectives basically say that individuals engage in juvenile crime and delinquency primarily because they’re imitating the behavior of the peers with whom they associate.

So, what we’re gonna talk about tonight is some of my own work in this area, but some other people’s work as well, looking at whether it’s the fact that this major prediction of differential association theory and social learning theory is in fact accurate.

We’re also gonna look at the degree to which it’s accurate.

How large is this relationship, this supposed relationship, between peer behavior and one’s own behavior? But then we’re also gonna talk a little bit about other ways, other than imitation and modeling, in which peers may influence an individual’s delinquent behavior.

So, I’ll start out by giving you guys a quick overview of the topics that we’ll cover.

Number one, I want to set the stage briefly by talking a little bit about juvenile crime in the United States. So, how much of a problem do we have with juvenile crime in comparison to, let’s say, other countries? And so, I’ll talk about that very briefly. And Ill talk also about how we study juvenile crime and delinquency from a sociological perspective.

So, I am primarily a sociologist who studies crime, deviance, and social psychology. So well talk a little bit about how what I study falls within a sociological framework.

Well then talk about the major predictions of differential association theory and social learning theory, two related theories and theories that continue to be very important in the existing literature.

Well then talk a little bit about traditional approaches to testing, empirically, with some form of data whether the predictions of these perspectives actually hold water. Whether theyre accurate or not.

But then well talk about some alternative interpretations of existing evidence on which differential association and social learning theorists have really tended to hang their hat. Even to the degree that the relationship between personal and peer delinquency may have been exaggerated in some of the existing literature, traditionally. It still may be the case that, as I said before, peers are very, very relevant to juvenile crime and delinquency but that they’re relevant in ways that go beyond just imitation, modeling, differential association. So, we’ll talk a little bit about that.

And then we’ll take some questions and have a little bit of a discussion, as much or as little as you guys would like.

So, that’s basically the overview, the plan for the rest of the evening. Okay.

I’ll start out by talking a little about juvenile crime in the United States. And I’ll begin by saying that we definitely had a lot of media coverage in the United States in the 1980s and ’90s, the late ’80s and early ’90s in particular, about juvenile crime. This was a major issue that was talked about in the news. This was a major social issue that not just academics, but other people as well were talking quite a bit about. There were a whole bunch of high profile cases that took place. We really had this issue in the late ’80s and early ’90s. And it was the case that, in fact, it wasn’t just something hyped by the media. There did seem to have been, according to a number of data sources, a pretty strong surge in juvenile violence specifically. Okay?

Since the early ’90s, probably around 19, maybe 93 or four, what ended up happening is that there was this very precipitous, very steep decline in a bunch of different types of crime nationwide, but particularly in juvenile violence.

However, I will say that the issue of juvenile crime and violence, in particular, remains very, very relevant. So, the overall rate of murder in the United States right now is probably about 4.5 per 100,000 people who live in the United States. It does vary from one geographic region to another within the United States, but the overall average in recent years has been about 4.5 murders per 100,000 people. Disproportionately, however, those are gonna be committed by juveniles and young adults more so than people of other age categories or other age groups.

And I’ll also say another reason that this issue of juvenile crime remains very, very relevant in the United States, particularly juvenile violence, is that even after this very steep decline in the murder rate throughout the United States, from a high, you know, over the course of the past several decades, the high has probably been upwards of about 10 per 100,000 people and we’re now down to about 4.5. So, it’s been cut in half. But even with that big precipitous drop in, particularly, juvenile violence, the United States, as it turns out, still has a pretty high rate of murder in comparison to other industrialized nations.

It’s maybe about average for all nations, if you look throughout the world, but among wealthy industrialized nations, the murder rate in the United States is still more than two or three times that of most of other wealthy industrialized nations. So, this remains an issue.

I’ll give you a couple of quick case studies. Some of you have maybe heard because these things happened. I’ve selected these specifically because they happened in your state. Okay?

One is the Slender Man stabbing. How many of you, just by a show of hands, are familiar with this case, with what happened? I imagine probably most of you. My understanding, the short version of my understanding, is that this involved, actually, juvenile females who stabbed someone else, another young person, many, many, many times, and then that individual ended up crawling out of, I think it might have been the woods, to a road and got, basically, saved by somebody. And and, basically, again, saved. Whereas they may well have died had they not been found.

Well, while this may not be the typical act of delinquency that takes place in the United States, it’s still something that did happen. Presumably it’s called the Slender Man stabbing because it was somehow influenced by this online meme, this Slender Man meme, that I’m only vaguely familiar with. But it brings up the issue of the degree to which people’s behavior, in this case a brutal stabbing, committed by young people, was influenced by the media or by someone else with whom they associated.

It gets at this issue that differential association and social learning theory deal with. About whether it’s the case that people’s behavior really is influenced by the behavior of their peers.

You might recognize these names. Steven Avery and Brenden Dassey. I’m not sure if I’m pronouncing that correctly. But who are these individuals? Okay?

These are individuals who are referenced in what is a documentary called Making a Murderer. How many of you are familiar? Again, most of you.

My understanding, and I haven’t watched it, but I’ve certainly heard quite a bit about it, and actually, it’s in my queue there to watch at some point when I have a little bit more time, but my understanding of this case was that these two individuals were put behind bars, but then years later, were exonerated with new evidence that emerged.

And one of the individuals, the younger one. So, one was presumably the uncle of the other, who is the nephew. And presumably the nephew confessed, if my understanding is correct. And, again, this brings the point to light. Is it the case that either, if they did commit this crime, which presumably they may not have, but if they did, did one influence the other? Did one do this thing, if it happened, because of the influence of the other, in this case maybe a family member? Or, if they didn’t do the crime, was it the case that the one confessed because he was influenced by other individuals, in this case police?

Both of these cases have taken place near where you guys live here in your state. And I think even though, again, they’re not your typical act of juvenile crime, they nonetheless are things that presumably happened.

Even if there wasn’t a crime in the second one, the case certainly took place.

It’s drawing on real events.

And I think both of these illustrate the ways in which people may be influenced, potentially, by the behavior of other people. And that’s largely what we’re gonna deal with today.

Now, to understand differential association and social learning theory, I need to explain to you very briefly, for those of you who are not sociology majors, what sociology actually is as a discipline. How it is that we study something like juvenile crime from a sociological perspective.

And in large part the discipline of sociology has to do with this concept of stratification. Stratification just means that people fall into different categories. There are different groups of people. People fall along various continuums. So, some people are wealthier. Some people are poorer. Some people are more educated. Some people are less educated. There are all these ways in which, in any given society, people are stratified along lines like wealth, education, et cetera. Okay?

Getting ahead of ourselves a little bit.

But there are any number of ways in which people can be stratified. I’ve just listed a few examples. Again, wealth, education, religion. Some people are more religious than others. Some people are not religious at all. Some people who are very religious have one particular faith, while others have a different faith. People, again, fall into different categories. They’re stratified, and sociology deals with stratification.

And, in particular, I tend to think of it as a discipline that deals with two aspects of stratification. One, is what are those things that cause people to be stratified? What are those things that might lead some people to be more educated or wealthy, for example? But, two, what are the things that result from stratification? Okay?

What does stratification itself cause?

As a criminologist, who is primarily a sociologist, and looks at crime through a sociological lens, primarily, I tend to think of criminology as a discipline that looks crime as potentially both a cause of stratification and as a consequence of stratification.

So, if you engage in crime, that might cause you to be stratified into certain institutions that we might call jails or prisons.

It might also be the case that crime is the result of stratification. That if you have a different level of wealth from somebody else, or a different level of education, that might influence your tendency to engage in criminal behavior.

In this particular talk, I’ll be focusing on the ways in which one particular thing may influence somebody’s crime, and that particular thing is the crime and deviance or delinquency of their friends and their peers.

So, let’s talk a little bit about differential association and social learning theory. All right.

The guiding question here, from a sociological perspective, is gonna be the following: Why are some people more criminal? Why is it that crime is stratified, particularly when we look at youth? Young people. Adolescents and young adults. Why is it that some of those individuals engage in more violence, drug use, property crime than others? Why is that behavior stratified?

Well, Edwin Sutherlands differential association theory, a very big theory, very important theory within the criminological literature, has a particular answer to that question. Okay?

Edwin Sutherland was a very famous criminologist who wrote one of the first criminology textbooks. And within the first ten pages of that textbook, one of the things that he did was he laid out a very, sort of, simple, straightforward, elegant, parsimonious theory of why some people are more criminal than other people.

Well, why, in his view, are some people more criminal than others? Well, he said, number one, because people are stratified in terms of the associates with whom they hand out, with the people with whom they hand out. Peoples social circles differ. Okay?

But beyond that, he says that some peoples social circles are predominantly law abiding. So, they tend to follow the letter of the law. Not completely, but predominantly. He says other individuals have social circles that consist of people who are predominantly criminal. It doesnt mean that everything they do is criminal, but it means that predominantly they embrace criminal values. Okay? They engage in a criminal lifestyle.

Sutherland suggests that criminal associates promote crime. Pretty straightforward. He says, specifically, they model criminal behavior, which is to say that they set a certain example, that they behave in a certain way, and that others can observe that behavior and imitate it. Okay? But he also says that criminal associates promote crime by encouraging attitudes that are favorable to criminal behavior. That if youre hanging out with criminals or deviant people, that their attitudes towards substance use or violence are gonna be positive attitudes towards those types of behaviors, and that those attitudes will rub off on you, and that when those positive attitudes about those behaviors rub off on you, that thats gonna make you more prone than maybe other people to engage in violence, drug use, any number of potentially deviant behaviors.

Overall, Sutherlands premise is the following: That a persons crime, relative to other people, is a function of his or her relative exposure to law-abiding versus criminal associates. If the balance of your associates are law-abiding, the majority of them are law-abiding, then youll tend to be that way too. But, if the majority of your associates in your social circles are deviant or criminal, Sutherlands theory say that you too will have a disproportionate likelihood of engaging things like violence, drug use, any number of deviant behaviors.

So, thats Edwin Sutherlands theory of differential association.

But in 1966, a couple of further scholars, Burgess and Akers, extended Sutherlands theory. They added to his theory by drawing on the discipline of psychology. They did agree that differential association, hanging out with people who are more or less deviant, does contribute to the stratification of crime. They agreed with Sutherland that if you hang out with deviant others, thats gonna make you more likely to be deviant yourself. Okay?

However, they also suggested that some people are more criminal than others because they have been differentially reinforced for their criminal behavior. Those reinforcements are basically rewards. They basically argue that, look, in addition to differential association, promoting criminal behavior, positive rewards positive reinforcements for criminal behavior will promote more criminal behavior. The drew a lot from the psychological literature from people like Watson and Skinner and others from the quote, unquote behavioral school. And they basically put these two ideas of differential association and reinforcement together to suggest that these two things are the primary reasons that some people are more deviant than others.

More recently, in 1998, Ronald Akers wrote a book about social learning theory in which he basically took the ideas that were laid out by Sutherland, combined those with the ideas that he and Burgess laid out in their 1966 article, and Akers basically writes the following. Its an entire book, but the gist is as follows: That crime is learned in social interaction, but that learning occurs through four major mechanisms.

One, he says, is differential association, as described by Sutherland. A second, is this concept of imitation, which is simply to say that if youre differentially associating with more criminal people, youre gonna tend to imitate their behavior. So, for example, little kids, from a very young age, its amazing how much they will imitate what they see around them. You make a funny face, and they can imitate that. Its amazing how much they can imitate behavior. Right?

So, Acker says that, yes, differential association is gonna be one major mechanism, but part of the reason differential association might promote crime is because when you hang out with deviant people, youre probably gonna imitate their behavior. Sometimes even if you dont know youre doing it.

Number three, he says that some people have attitudes that are more favorable to crime, other people have attitudes that say crime is wrong. That the people who have more pro-crime attitudes are more likely to engage in criminal behavior.

And, lastly, he draws on this idea of reinforcement, again going back to his article with Burgess, and says that individuals who are rewarded for criminal behavior are gonna continue committing more and more of that behavior. And thats kind of become the dominant framework of social learning theory, that is sort of this umbrella theory now, including his prior work with Burgess, but also including the contributions of people like Sutherland. All right.

Well, in sociology, its one thing to come up with a theory that you think is brilliant and explains a certain type of behavior and the stratification of that behavior, but its another thing to come up with a theory that actually ends up being supported by the empirical evidence. So, again, I look at sociology as a discipline that concerns stratification, the causes and consequences of social stratification, but I also look at it as a discipline that studies, scientifically as much as possible, the causes and consequences of social stratification. To study something scientifically, we need to subject predictions to empirical tests. So, a number of scholars for many, many years have done that with differential association theory and social learning theory.

And heres the traditional way in which theyve gone about testing whether the theory is correct or not. In large part, theyve used self-report methodology. I use this in my own research. Self-report methodology means that I give you guys a survey, sometimes it might be anonymous, sometimes not, but I ask you about your criminal behaviors. And I ask you to disclose on that survey how much youve engaged in any number of different criminal behaviors. But Ill also, in that self-report survey, measure any number of other things so that I can try to get a sense of what factors in the social environment predict how much somebody will or will not be delinquent.

And when testing differential association theory and social learning theory, the major variable, the major thing that sociologists and criminologists have measured has been a respondent’s perception of how much his or her friends get into trouble.

So, a typical self-report survey that aims to test differential association theory or social learning theory will ask individuals about their own behavior. So, imagine filling out a survey in which I ask you about your own deviant behaviors, but then I also include a bunch of questions about how much you think that your friends engage in all those same behaviors.

And usually that takes the form of questions like, what proportion of your friends have engaged in the use of this particular drug? What proportion of your friends have engaged in fighting in the past six months or year? What proportion of your friends have engaged in any number of other deviant behaviors?

Well, using those data, typically, tests of differential association theory and social learning theory will look at the correlation, the co-variance over time, of personal delinquency versus peer delinquency. Personal delinquency, again, is measured by asking respondents about their own behavior and putting together a scale that reflects their overall delinquency score, and then looking at whether those individuals with high scores for their own delinquent behavior also have high scores for the behavior of their peers, the delinquent behavior of their peers.

And this is typically what you will find. Imagine that each of these little dots here is a different person. Each of these little dots has one score for peer delinquency on the x-axis here, the horizontal axis here, but then each of these little dots also has a score for personal delinquency on the y-axis, the vertical axis here. Differential association theory’s major prediction is pretty straightforward. It says that those individuals who have higher scores for peer delinquency should also have higher scores for their own personal delinquency. Lo and behold, that’s what we find time and time again.

We find it in every single sample that you could possibly imagine. In representative samples, in non-representative samples, among whites, among Blacks, among Latinos, Hispanics, Asian people, any number of different groups you’ll find the same thing. The result is very, very consistent. But aside from being very consistent, the correlation between peer delinquency and personal delinquency is extremely strong. Extremely strong. Okay?

You might find a correlation between these two things typically of maybe 0.3, 0.4, 0.5. So, for those of you who haven’t had maybe a stats and methods class, correlation will range from negative one to one. And in social science, we’re not biologists, we can’t measure things as precisely as people in a discipline like biology can. You might see correlations of 0.9 in the biological sciences sometimes. That’s very rare in social science. If you see a correlation of .3, that’s extremely strong. The correlation between peer delinquency and personal delinquency exceeds that regularly. It is incredibly strong within social science. That seems some powerful support for this idea coming out of differential association theory.

How do differential association and learning theorists interpret this correlation? Well, you can imagine. They say, uh-huh, strong, strong correlation. In fact, they’re kind of fond of saying to other theorists that in some sense our correlation is way stronger than your correlation. Therefore, our theory is correct, and your theory is maybe not as correct. All right?

Well, there are some alternative interpretations of that really massively strong correlation. And we’ll talk about two alternative interpretations.

Alternative number one, I’m gonna label that the measurement critique. Well, the measurement critique, what do I mean by that? I mean that respondent reports of their personal behavior and of their friend’s behavior, their peer’s behavior, may be inaccurate. Maybe there’s some way in which one or both of those measures coming from respondents are just not accurate. All right?

Well, that can take a couple of different forms. It might be, number one, that respondents want to conceal their deviant behavior. They don’t want to write down on that survey, even if it’s anonymous. Maybe they’re scared to write down that they’ve done whatever behaviors. So, they want to conceal that. That’s plausible. In some cases, it might be that they want to hype their own delinquency and make up how much of a badass they are by telling you that, yeah, I engaged in a bunch of delinquent behavior, look at me. So that’s also very plausible. And we have some evidence that that tends to happen in some cases, too. But beyond that, a respondent’s perceptions of his or her friends may be inaccurate. Right?

You think you’re close friends with these 10 people, and maybe you are, but are you really certain of exactly how much deviant and criminal behavior they’re engaging in? Maybe you’re not. Okay?

Well, there’s research about both of these possibilities in the criminological literature. And that research suggests that people’s reports of their own behavior actually tend to be surprisingly accurate. Okay? They’re not perfectly accurate. People do lie. People misremember. People do all these things. But people’s estimates of their own criminal and deviant behavior are surprisingly accurate. And how do we know that?

Well, there are various ways that we can go about checking the accuracy of people’s self-reports concerning their own behavior. One, is that if they say they haven’t engaged in this behavior or that behavior, we can go and do record checks. You know, with the local authorities, if we go through the proper channels, through the IRB and so forth, we can then check their record. And if we find that they have been picked up for any number of different behaviors all these times but they didn’t self-disclose that, then we know that they’re lying. And that does happen.

But it’s surprising how much they actually disclose those behaviors. How much it’s the case that they disclose a particular behavior and then we go and check their record and find that, yep, they’re being honest about their drug use, their violence, any number of different things.

We have other ways that we can try to figure out whether they’re telling the truth, and one of those is lie detectors. People can beat lie detectors. That’s true. People can beat them, but a lot of people don’t. And it’s just one more way that we can go about trying to validate people’s self-reports of their own behavior. And when we’ve used lie detectors, we have found that, yeah, lo and behold, people are remarkably honest, surprisingly honest, not in all cases but in most cases, about their substance use, about their violent behaviors and so forth.

We can do drug tests. We can see if somebody says, no, I’ve never used such and such chemical, and then we can see whether it’s the case that they passed the drug test.

These are examples of the different types of ways that criminologists have gone about trying to validate self-report measures of personal behavior. And there’s a whole literature about that, but the short version for our purposes is, that people are remarkably honest, not 100%, but you’d be surprised how honest people in these self-report surveys are, suggesting to we as researchers, to us as researchers, that maybe problems with peoples self-reports of their own behavior cant explain whats going on here. Okay? But something else might.

Theres research also that looks at the accuracy of peoples perceptions concerning their friends. And that research suggests greater inaccuracy in respondents perceptions. That is to say we have some reason to believe that when people report about their friends, whether deliberately or probably unconsciously without meaning to, they just dont actually know how much their friends are engaging in a bunch of different delinquent behaviors. Okay?

How do we know that?

Well, some studies have moved beyond this initial methodology, the traditional methodology, of measuring peer behavior based upon a respondents perception. Some studies have actually gone ahead and asked the peers themselves without relying on what Person A says about his friends or her friends, and instead going directly to the friends themselves and asking them how much theyve engaged in these behaviors to get what we think is a more accurate depiction of peer delinquency.

These measures, this alternative strategy for measuring peer delinquency, allows us to reexamine the correlation between personal and peer delinquency to see if the correlation continues to be that strong once we measure peer behavior more accurately. And here’s an example of something that people have found.

Okay, so a guy named Frank Weerman and a colleague, Smeenk, in 2005 published an article in the journal Criminology that answered the question Does the correlation between personal and peer delinquency diminish when you use direct peer measures, rather than basing your measures of peers on a primary respondent’s perception that might be inaccurate? And here’s basically what they found.

So, what we’re looking at here is the correlation between personal and peer delinquency. This correlation that tends to be so large in the criminological literature. Well, we’re gonna see whether that correlation remains that large once we use what we think are maybe more accurate measures of peer behavior.

So, in the light blue here will be correlations that are calculated by looking at the individual’s own self-reports, his or her behavior, and looking at the correlation between that and his or her perception of his or her peers. In the darker blue, we’re gonna look at the same correlation, but we’re gonna measure peer delinquency differently. We’re gonna actually measure it from the peers themselves, and we’ll see what the correlation looks like then. And here’s what happens.

In this particular study, when they looked at peer delinquency in terms of whether peers had engaged in any amount of delinquent behavior, and they said, all right, well, what proportion of your friends have engaged in any amount of delinquent behavior? They wrote down those measures. They looked at this correlation in question using the traditional method, in the lighter blue, and using the new method, direct peer measures. And the correlation diminished quite a lot. The correlation was no longer so strong.

But then they said, all right, well, what if we focus on severe delinquency? What if we focus on those who are repeat offenders? Maybe that will change our result. And it didn’t change the result. The same thing happened. This huge correlation of over 0.4 diminished to less than 0.2. It was more than cut in half.

Well, what if we instead of focusing on friends generally, focus on people’s perception of their best friends. Maybe they know their best friends really well, and they’ll be more accurate in their portrayals of their best friend’s delinquency. Well, when we measured things that way, or when Weerman and Smeenk did, then they found the same result. It didn’t make a difference.

Then they said, one last check. Let’s measure peer behavior in terms of severe delinquency, repeat offenders, and we’ll look at best friends and see what happens there. The same thing. Across every one of these analyses they basically found that if you measure peer behavior based on information that comes directly from the peers themselves, this correlation between personal and peer behavior is no longer nearly as strong.

It’s much, much weaker. It’s still there, but it’s much, much weaker.

The question then becomes: why?

Well, I’ve done some work with a colleague named Kathy Modecki. This was published in 2014 in Journal of Quantitative Criminology. And, among other things in this article, we were looking at whether respondents underestimate the differences between their behavior and that of their friends. Okay?

So, we had a number of different items. What we did for this particular study is we asked individuals at the University of New Hampshire to come into the lab. And we asked them to fill out a survey. And in this survey we asked them about their own behavior, and we said, on this scale from one to nine, how often in the past six months have you pushed somebody? And they filled out that question. And then we said, in addition to coming to the lab, we would like for, not just you to show up, but you want you to bring a friend of yours to the lab. And on the surveys that we handed to these primary respondents, we not only asked them about their own behavior, but we also asked them about their perception of how often they believed that this friend of theirs, that they brought with them, pushed people in the past six months.

And that was a perceptual measure. Thats their belief about how much their friend has engaged in this behavior. But in the meantime, we took that friend, put them in a separate room, and gave that friend their own survey, and asked them their own estimate of how much they, themselves responded that they had engaged in pushing somebody.

So, now we had two different measures of this friends behavior. The one that comes from a perception from the primary respondent, and the one that comes from the actual peer, him or herself, which presumably, given the relative accuracy of self-reports, should be more of a gold standard measure of that individuals behavior.

What we then did is we looked at the difference between the personal behavior and the peer behavior. We just subtracted one from the other. We looked at how different those two things were. But we did that in two different ways. We looked at the difference between the respondent’s behavior and the peer’s behavior, when both of those measures came from that primary respondent. Okay.

That’s in light blue here.

And then we looked at the difference between person A and person B when they each use their own self-reports.

And here’s what we found. When it came to pushing, that light blue bar right there at the left, basically suggests that individuals severely underestimate the difference between themselves and their friend. They presumably know this person very well.

But if you go by what the friend actually says, you calculate the difference between these two people using the friend’s own self-reported information, the difference between person A and person B is actually much, much bigger in that behavior than the one individual apparently realizes.

We did the same thing with other items to see if we would get the same result. We asked: How much in the past six months have you hit other people? How much in the past six months do you perceive that your friend has hit other people? And then, friend: How much in the past six months have you actually done this behavior? Same results.

The primary respondent, again, underestimated the difference between their hitting behavior and their friend’s hitting behavior. The real difference between the two was much bigger, as indicated by the darker bar here. Okay?

Well, what if we ask the question: How often in the past six months have you kicked somebody? Or: Have you used force to get something from somebody? That’s basically robbery. Same result.

To a slightly lesser degree, when we looked at this kicked item, but across all four of these, individuals tended to underestimate the difference between their behavior and that of their friend.

This reflects something that has been called, maybe projection effect or a false-consensus effect. So, false consensus means people are subject to this, myself included. For better or for worse, people do, to different degrees, have a tendency to assume greater similarity of other people, okay, to themselves than really is warranted.

People tend to assume that others are more like them than others actually are. This is one example of that applied to the criminological literature. Okay?

When we combined all four of those items, I’ll mention, for an overall scale, we found the same thing. Again, this light blue bar basically means that individuals are underestimating how different their friend is from them in terms of forceful, aggressive, violent behavior. That might be one of the reasons, then, that the correlation between personal and peer delinquency is much, much weaker when you use data coming directly from friends themselves rather than just using data from one individual who reports about their behavior but then gives a perception that clearly is faulty of their friend’s behavior. Right?

All of that is to say that, look, we still find this correlation between personal and peer behavior exists, but when you actually measure peer behavior in a more appropriate way that’s probably more objectively correct, the magnitude, the size of that correlation where the learning theorists are saying our correlation is way bigger than your correlation, that isn’t so true, as it turns out, if you use better measures of peer behavior that come from the peers themselves. All right.

There’s another critique, potentially, of learning theory and of the research that presumably supports learning theory, and that is, even for that correlation that remains, the weaker correlation that still remains between personal and peer behavior when you’ve measured peer behavior more correctly, maybe instead of peer delinquency causing personal delinquency, as indicated by this little graphic here, maybe what’s going on is the reverse of that. Maybe personal delinquency influences one’s choice of peers.

Why would that happen? Well, if you’re a hardcore, you know, gang-banger, whatever you want to call it, or if you are a heavy substance user, or if you’re just, you know, whatever type of a criminal, imagine, are you gonna want to hang out with people who are criticizing your behavior day in and day out, who don’t do the same stuff, who don’t find that stuff fun? Or are you gonna want to hang out with people who are like you, and think all that same stuff is fun, and engage in that stuff, and can provide you with opportunities to use drugs or to engage in violence or to do whatever other thing? You’re probably, one could argue, gonna want to hang out with people who are like you.

So, maybe your own delinquent behavior leads you to self-select yourself into social groups that are similar to you. Maybe your own behavior is affecting who you associate with rather than the reverse. And maybe that reverse correlation, or that reverse causality, is what accounts for the remaining small correlation between personal and peer behavior. That’s a second criticism. Okay?

Well, as it turns out, when I read the Weerman article from 2005 that I talked previously about, when I read that article I I thought, wow, what an amazing study. This guy, he’s in the Netherlands and he’s collected these amazing data in the Netherlands and I wish I could work with this guy because I don’t have access to data like what he’s got. I mean he had this amazing network data set where basically what he did was he had a sample of more than a thousand young people, and he asked them about their own deviant behavior but he asked them to name ten of their best friends, closest friends, up to ten.

But those 10 people were also in the data set. They were also respondents. They weren’t just friends of respondent number one. They were respondents three and 17 and 242. They were also in the data set. So, he could use all these data. Once he had the list of each person’s friends, he could then actually create a more objective measure of friend’s delinquency than you can do if all you’ve got is the primary respondent’s perception of friend’s behavior. Right? Amazing data set.

And so, I wrote this guy I didn’t know him but I wrote him an email. I found his email address. Hey, you don’t know me. I’m a criminologist. I was, you know, a younger criminologist at the time, whatever, and I said, you know, but I’m interested in some of these issues. I’ve published, you know, this article. I gave him an example of something that I’d done before. And I said, here’s something I’m working on now, but you have these amazing data and just, you know, the odd chance that you’d ever want to collaborate. And I didn’t think he would, but I figured it doesn’t hurt to just ask.

Lo and behold, pretty quickly he emailed me right back and he said, oh, well, interestingly enough, I’ve got these two other colleagues that are also young criminologists who are interested in these issues, and we were thinking about some ideas similar to the ideas that you had. And the most amazing thing happened. He just asked me to collaborate with them or took up my offer to collaborate with him. Either way, same result.

So, I end up with this amazing collaboration with a number of different people that I didn’t really previously know that well anyway. And here’s what we did in this particular study that came out in 2014 in the journal Criminology.

We used Frank, so, Weerman, he’s the fourth author here, is is Frank. We used his data, his network data. And with his data we created a measure of respondent’s behavior, the primary respondent’s deviant behavior. Their delinquency.

We also created a measure of their perception of deviant and delinquent behavior among the friends that they nominated. What was their belief about how delinquent their friends were? Okay?

But because we had data from the friends as well in this network study that Frank had put together, we could create a measure of the peer’s own self-reports of their own behavior, the more objective measure of peer behavior.

But not only were we able to create all three of these different measures, we also have these measures from the same people over multiple time periods.

So, this second set of circles here, are the same three measures at a later point in time.

And with these data we could do some really cool stuff. I hope you’ll find it even, you know, half as cool as I find it, anyway.

Number one, we could estimate how much prior behavior was associated with future behavior. It turns out there’s this adage that some of you may have heard that the best predictor of future behavior is what? Prior behavior.

Well, we wanted to control for that. And given that we had multiple waves of data from the same people in what’s called a panel study, we could estimate. How much does somebody’s prior behavior seem to be associated with their future behavior? And the relationship there is pretty strong, and criminologists know this and that’s one of the reasons that we wanted to control for this. So, we did that.

But similarly, we wanted to look at how much respondent perception of peer behavior at time one was associated with respondent perception of peer behavior later. Right?

And similarly, how much does the direct measure of peer behavior, how much is it associated with the later version of that measure, at a later point in time? All of those are quote, unquote, statistically significant, for those of you who have had stats.

But then we could do some really cool other things. In this model that we created, we could estimate the association of peer’s actual behavior down here, this is what the peers themselves report, with primary respondent’s perception. We could look at whether their perceptions of their friends were accurate or not, in a more sophisticated way than in the prior study.

And we found that there was surprisingly, it still boggles my mind a little bit. There was surprisingly little relationship there. There was some, but not as strong a relationship nearly as you would think. In some cases, not statistically significant or barely so.

Well, in addition to that, we found that there was a relatively weak relationship between respondent perceptions of peer behavior and respondent behavior once we controlled for all of these other factors in a longitudinal study across time. And we said, all right, well, maybe there’s some way in which peer’s actual behavior influences the respondent’s behavior. Maybe friends influence people, but they don’t do so through conscious awareness, through perception, they do so unconsciously. Anything’s possible. Let’s look at that simultaneously. We found very little relationship there.

But the big thing that we did find that accounts for the majority of the association between personal and peer behavior was this. We found a really strong association between a respondent’s behavior and his or her perception of friends; which is to say we found very, very strong evidence of projection. Have you guys heard of projection? So, you know, you’re accusing me of all the things that you do. You’re projecting your stuff onto me. We actually found something that very strongly supported that when we used this longitudinal approach and Weerman’s data. Okay?

All right. So, we’re not finding a lot of support. Now, once once we’ve addressed both of these two critiques, okay, the measurement critique and the temporal ordering critique, we’re no longer finding that much support for the idea that peers model behavior that then, you know, others imitate. We’re not finding that much support for this key idea of differential association and social learning theory anymore.

Well, does that mean that peers are not relevant?

No. There are some experimental studies, and I’ll try to be brief in the interest of time, but we can talk more about them afterwards if you guys are interested. There are a limited number of experimental studies. They’re difficult to do for a host of reasons. But there are some experimental studies that try and set up scenarios in the lab where individuals are presented with a deviant model. A confederate, which is to say in social science, that as many of you may know, means somebody who’s in on the experiment, plays the role of a participant in the experiment, unbeknownst to the actual participant. And that confederate will cheat somehow on some task.

And then, in other situations, other people will be assigned, randomly, to scenarios in which that confederate does not cheat. And lo and behold, we find that there is a small relationship between peer behavior and personal behavior. That is to say that if somebody sees another individual cheating on some experimental task in the lab, they do become a little bit more likely to cheat themselves. So there still remains a little bit of support for this idea that delinquency may happen, in this case a relatively minor form of delinquency, because people are imitating the behavior of those around them. But it’s a very small relationship. There’s not much going on there. Not nearly as much as learning theorists used to tout when they said look at this massive correlation and your very small correlation, our theory is better than yours. Okay?

But there are other ways in which peers might matter.

Peers can have influence on behavior, as it turns out, if you think about it, without modeling that behavior, without actually doing the behavior themselves. Well, how?

Well, any number of ways, but here’s one example. So, now we’ll go back to a study from 2005 that I did not do, but I think is a genius study. It’s absolutely brilliant, I mean, in my opinion. This was a study by Gardner and Steinberg, 2005, and what they did is they brought individuals into a lab. And they said we’re gonna have you play a video game. And the video game is called Chicken. And what you have to do in this video game, you just control one thing. You are in control of a car, and you can make the car go or you can make the car stop. That’s all you can do.

You can make it go. you can make it stop. Here are the buttons to do that.

Now, if you make the car go, the farther you get, the more points you will accumulate, and you’ll get some reward at the end of the experiment depending on how many points you’ve accumulated. Right?

So, people want to accumulate points. And they get more points the farther they drive the car. However, here’s the catch. There is this stop sign over here, and at some point these individuals, who are gonna go through about ten or fifteen trials, I think it was maybe fifteen. These individuals are told that in each trial at some point, unbeknownst to you when, the yellow light is gonna appear. And when that yellow light appears, you don’t know how long it’s gonna be before red appears, but once the red light appears, this wall is gonna come onto the screen and you’re gonna crash into the wall and you lose whatever points you accumulated in that trial.

What they’ve done is they’ve created a measure of how much people are willing to take some risk. And they measured that risk in terms of how long it was after the yellow light that individuals kept on going. Some individuals, you may have friends who are, you know, like: Yeah, just go! And they don’t even care. And they’re just and then you have some who are more conservative and they’re like: Oh, no, I have my points; this is how many points I’ve accumulated, and I don’t want to take the risk so I stop immediately, and I’m gonna bank those points. And they’re less risk-taking.

But here’s what was really cool about this experiment. The researchers included multiple different conditions. Two primary conditions. For some people, they were put through this experiment with a group of peers watching them. The peers didn’t take any risks. They didn’t model any behaviors. They didn’t do anything. They were just there observing.

So, in this version of the experiment, these participants knew they were being watched by people of their own age. In the other version, people just did this experiment by themselves. There was nobody watching them, and they did it solo with no audience.

There were three different age groups that were subjected to these two different conditions. There were adults, young adults, and adolescents.

What Gardner and Steinberg then did is they basically measured risky driving in each of these two conditions across each of these three groups. And here is what they found. Among the adults, adults in both conditions, sole participant and peer group observing, were both pretty conservative. In comparison to the overall average of risk-taking here at zero, the adults tended to be risk averse. Go figure.

[laughter]

There was a minor difference between the adults who participated by themselves and those who had peers observing them. The ones who had peers observing them were a little less risk-averse or a little more risk-taking, but they were still pretty risk-averse. And that was a minor difference.

Among the young adults, things changed a little bit. The young adults who were sole participants, they were actually pretty risk averse. But the young adults who had peers of young adults watching them, not giving them a model of any behavior but just watching them. Okay? They tended to take a little bit more risk, but it was still statistically a relatively minor difference.

Here’s what happens with the adolescents that I find amazing. This is why I love this study. I mean, I interpret it a little bit differently actually than the authors do, but I don’t even care. I just think this is such a brilliant study.

The adolescents who were by themselves were actually pretty reasonable in assessing whether they should just stop and accumulate their points. But the adolescents who knew they were being watched by other adolescents just tended to step on the gas. That’s peer influence without modeling. Right?

Why does that happen? It’s the last question we’ll talk about, okay?

Why is it that audiences may promote risk, particularly among adolescents, young people?

There is some literature that Steinberg and some of Steinberg’s students have published about this issue. And they come to the following conclusion using some really sophisticated psychological measures. They They’ve measured various things in adolescents’ brains, while they, they’ve replicated this experiment many times, by the way. They find the same results, more or less. But in later versions of the experiment, they have hooked people up to sophisticated equipment that I don’t even understand because I’m a sociologist, but I still find it fascinating.

And what they have found in multiple studies is that when adolescents are surrounded by peer audiences, parts of their limbic system really become very, very activated. The parts of the limbic system, basically those that are associated with the anticipation of some sort of rewarding stimulus. It’s as if, when they’re surrounded by an audience, they’re anticipating some sort of reward.

Okay, I believe that because, you know, they know more about those pieces of equipment than I do, and I think they’re brilliant anyway, so I believe them. But, to me, that still begs a further question, why does an audience prime individuals to anticipate reward? That question really, to my mind, doesn’t seem to have been answered in this line of literature.

I would suggest, and I’m now doing some further work. This has not yet been published but is ongoing, and we have some preliminary results and I’ll present those to you before I conclude.

I, along with some other colleagues, think that maybe what’s going on here is that adolescents maybe have already learned in their real world social interactions and their real world experience that if you are an adolescent and you take a certain amount of risk, that might increase your status among peers for a whole host of potential reasons.

A simple example is hazing. I mean, you guys are at least, you know, college-age people. You have some sense of what hazing is. Basically, you have to do something, you know, either entertaining or dumb or risky or something in order to get the approval of some group. Maybe you’re beyond adolescence at that point and more into young adulthood or emerging adulthood, but same principle.

That’s what we think might be going on, and we wanted to test this. And we did so in the following way. Over the past ten years, with a couple of colleagues in particular, Karen Van Gundy and Ellen Cohn, and Rick Trinkner as well, all three of us, actually all four of us, have been involved in the collection of data from two different cohorts of youth who were in sixth and ninth grade in 2006 when we began the study. And we have tried to follow them every six months or so. We have accumulated a bunch of data from as many as we could follow, even past their graduation from high school.

We’ve collected a bunch of variables, a bunch of things from them. One is that we have a measure, a scale, that basically reflects their perception of how popular they are among other people their age. There are a number of questions we include on that scale. Examples are, I’m popular among people my age. People invite me to do things often with them. I am liked when I’m around other people. All these questions that basically get at their status within their peer group.

We also have a measure of risk-seeking that asks individuals how much they like to take risks. How much they sometimes find themselves doing something daring just for the thrill of it. How much any number of other things that are related to their risk-taking tendencies.

So, we have these two different scales. One, a popularity measure. Another, a risk-seeking measure. And we’ve used data from this study, from multiple waves of this study, to look at the following issue.

Even when you control for people’s prior level of popularity, okay? Wherever they started out at the beginning of our study. When they take risks, does that elevate their status among peers? Does it decrease their status among peers? Or does it have no effect?

And here’s what we found. A little bit complicated for the males, the boys in our study. That at relatively lower levels, up through moderate and slightly high levels, the more they were risk-seeking, the more their popularity elevated, the more this potential social reward increased.

But we found something interesting. We found that beyond a certain level, if they were taking too much risk, they were too risk-seeking, it would appear, at least if you go by our measures, they do have some of their own problems which we can talk about any questions that you have, but when they go beyond a certain point, all bets are off, and it’s kind of like, oh, no, uh-uh, their peers find it to be a little too much and their popularity starts to decline a little bit beyond a certain point. But overall, there’s an overall positive relationship between risk-seeking and popularity among the guys.

We broke this out by males versus females for a whole host of reasons ’cause we had some sense there might be some differences here, and it turned out there was a big difference, at least according to our measures. For the females, taking risks didn’t have one influence, like an influence one way or the other. So, a different result.

So, I’ll summarize. I started out by saying I’m gonna approach this issue of peers and delinquency from a sociological perspective. From that perspective, there’s a dominant theory that says you hang out with deviants, that makes you more deviant. There’s a lot of empirical evidence traditionally over many decades suggesting a huge correlation between personal behavior and that of peers. That seems to support this theory in a very strong way. But wait a second. There are a couple of major criticisms that I don’t think had necessarily been dealt with as much as they should have been in prior decades.

When you deal with those two criticisms, the relationship between personal and peer behavior diminishes very, very sharply. It might still be there, but it’s small. Lastly, peers may still matter because there are other ways in which they can influence behavior. Okay?

Gardner and Steinberg’s study suggests that that’s the case. It’s been replicated. The research that I’m doing with Trinkner, Van Gundy, and Cohn suggests, we think, one potential reason.

That audiences have an influence on behavior, at least for males. That when males engage in some amount of risk at that age, that may come with some social reward. Okay?

So, thank you very much. And with that, I will take any questions that you guys have.

[applause]