Replicated findings of an evaluation of a brief intervention designed to prevent high-risk drinking among first-year college students: implications for social norming theory.

College students were randomly assigned to receive either (a) standard alcohol programming (SAP) or (b) SAP plus an intervention designed to change perceptions of alcohol norms (PAN). Effects of the intervention delivered during one class period (i.e., 55 minutes) were assessed using pre- and post-intervention surveys about personal alcohol use and perceptions of other students' drinking. Compared to students in the SAP only condition (N = 472), those receiving the PAN intervention (N = 402) reported significant reductions in perceived peer drinking; but minimal differences in serf-reported drinking. These results replicate the findings from the previous year of implementation of the PAN intervention and raise concerns about the practical and empirical validity of brief social norming interventions designed to reduce drinking among college students.

INTRODUCTION

There has been a dramatic proliferation of interventions that use techniques based on social norming theory (SNT) to attempt to reduce problematic drinking among college students. According to one estimate, at least 20% of colleges in the United States have implemented an intervention based on SNT (Wechsler, Lee, Kuo, & Lee, 2000). However, in a recent comprehensive review of individual-focussed interventions designed to reduce problematic drinking among college students, Latimer and Cronce (2002) reported that conclusions about interventions based on SNT are premature because the literature is severely limited in terms of both the quantity and the quality of published studies. Thus, it seems that there is very little evidence to support the widespread dissemination of social norming programs (Smith, 2003).

Further concern about the practical utility of social norming theory comes from our pilot research (Smith, Stamper, & Gant, in press). This randomized, post-test only study compared the effects of a brief (50 minute) small group intervention designed to change perceptions of drinking norms (PAN) with a standard alcohol presentation (SAP) that focussed on education about consequences of drinking. Compared to SAP, post-intervention measures found that the PAN intervention was associated with lower estimates of the frequency of peer drinking. However, there were no statistically significant differences in three measures of self-reported drinking: quantity, frequency, and rates of binge drinking.

The results of Smith et al. (in press) raise some concerns about the validity of SNT as an intervention model for college student drinking. The critical proposition of SNT interventions is that the effect of the social norming intervention on changes in drinking should be mediated by changes in perceptions. Thus, the proximal target of intervention is perceptions and the distal target is drinking. No other causal mechanisms are postulated by social norming theory, so the effects of a SNT intervention should be completely mediated by changes in perceptions.

Changes in perceptions without corresponding changes in drinking, which was the pattern of results in Smith et al. (in press) and at least one other study (Thombs & Hamilton, 2002), raises questions about the practical utility of changing perceptions. Moreover, studies that find changes in drinking, but not in perceptions (Schroeder & Prentice, 1998) raise questions about the theoretical validity of the proposition that the putative effects of SNT interventions are mediated by changes in perceptions. Thus, there is some preliminary research from randomized studies that suggest that there may be some serious problems with the theoretical validity of interventions based on SNT. However, the preliminary randomized studies suffer from a number of methodological limitations (Berkowitz, 2003). Moreover, the results of the preliminary randomized studies contrast with significant changes in perceptions and drinking seen in several campus campus-wide implementations (Berkowitz, 2003). Unfortunately, these large-scale, campus-wide studies are inconclusive because they lacked important experimental controls. For example, these studies did not have no-treatment comparison groups and failed to measure other interventions, such as changes in administrative policy that may have been implemented simultaneously with the social norming intervention (Smith, 2003).

Taken together, the preliminary studies on SNT have not put SNT theory to a rigorous empirical test. Thus, there continues to be considerable uncertainty about the theoretical and practical utility of interventions based on SNT. This study is designed to reduce some of the uncertainty about SNT by examining the efficacy and theoretical validity of a brief, small-group, peer-led intervention with a large sample of college students (n = 874). Changes were assessed with pre- and post-intervention assessments of drinking and perceptions of peer drinking. In order to assess efficacy in a rigorous manner, we randomly assigned groups of first-year college students to receive either (a) standard alcohol programming (SAP) or (b) SAP plus an intervention designed to change perceptions of alcohol norms (PAN). Three sets of analyses were conducted to assess theoretically and practically meaningful issues. In the first set of analyses, repeated measures analysis of variance (ANOVA) was used to see if the PAN intervention was associated with changes in both perceptions of peer drinking and self-reported drinking. The second set of analyses used ANOVA to look at intervention effects in subgroups of students in order to assess the generalizability of the PAN intervention across the larger sample. The third set of analyses used path modeling to assess the hypothesis that the PAN intervention effects would be completely mediated by the changes in perceptions.

METHOD

Design and Procedure

Sections of a class course for first-year students designed to ease the transition from high school to college (i.e., University 101) were randomly assigned using a random number table to receive either the standard alcohol presentation (SAP) or SAP plus the Perception of Alcohol Norms (PAN) intervention. Both interventions took one class period (i.e., about 45 minutes for SAP and 60 minutes for PAN). Approximately one month before and about one month after the intervention, a research assistant visited the class to administer surveys. Thus, this is a randomized study with pre- and post-intervention assessment.

Participants

Pre-intervention data were collected at the beginning of the semester from 1,411 students in University 101 classes. After the alcohol intervention, follow-up data was gathered from 1,152 students, all of whom had either the PAN or SAP interventions. We collected baseline data from 80 sections of University 101 and follow-up data from 70 sections. Data from 10 sections was lost due to scheduling problems with the post-intervention data collection.

Of the students who completed the follow-up survey, 874 of 1,152 students were matched up to the original 1,411 students who participated in the baseline survey. Almost 300 students were not matched up because (a) they did not participate in the baseline survey, or (b) they did not include identifying information in either or both of the pre- or post-test conditions. The 874 students who participated in both the baseline and follow-up surveys will be referred to as the repeated measures group, of which 402 received the PAN intervention and 472 received the SAP intervention.

A summary of selected participant characteristics follows: 61.4% female, 75.2% Caucasian, 18.2% African American, 3.3% Asian/ Pacific Islander, and 1.0% Hispanic. The demographic characteristics of the entire incoming class were as follows: 55.7% female, 74.7% Caucasian, 17.6% African American, 3.3% Asian/Pacific Islander, and .9% Hispanic. Thus, with the exception of gender, the demographic characteristics of the sample of students from the University 101 classes were equivalent to the entire class of first-year students.

Measures

The self-report surveys took about 15 to 20 minutes to complete. The major loci of the surveys were the quantity and frequency of drinking, self-reported alcohol-related problems, and perceptions of alcohol use by peers. Additional content included in the survey that was not pertinent to the current investigation included questions about life satisfaction, alcohol expectancies, and self-reported history of psychiatric diagnosis.

Quantity of Drinking. The survey asked about the quantity of drinking in a variety of ways including self-report of (a) the number of drinks on each individual day of the previous week, (b) the number of drinks they have then when they "party", and (c) the number of drinks they have on a "typical day of drinking." Because the "number of drinks when partying" had the highest incremental validity of the quantity of drinking measures for predicting alcohol-related problems, we chose to use it as our primary measure of the quantity of drinking in our analyses. Secondary analyses with other variables used to measure quantity of drinking yielded the same pattern of results as with "number of drinks when partying." The test-retest reliability for this measure (Pearson r) was .84.

Frequency of Drinking. The survey provided two ways to assess the frequency of drinking, including serf-report of (a) the number of days they drank in the past week, and (b) the number of days they drank in the past month. These variables were highly correlated, with a Pearson r of .85. The frequency of drinking variable with the greatest incremental validity for predicting alcohol-related problems was the "number of drinking days in the past month". Therefore, we used the monthly frequency as the primary measure in our analyses. The test-retest reliability for this measure was .82 (Pearson r).

The Interventions

One meeting of each section of University 101 was conducted by peer leaders trained and supervised by the University's Alcohol and Drug Education Program housed in the Department of Student Life. The peer leaders provided an intervention called Standard Alcohol Programming (SAP). This intervention focused primarily on interactive presentations focusing on values clarification and the risks of drinking and its consequences. Props such as the Fatal Vision goggles were used to enhance the information about the relationship between blood alcohol content and impairment (www.fatalvision.com).

Half of the sections of University 101 were randomly assigned to receive a new intervention designed to change Perceptions of Alcohol Norms (PAN). However, the PAN intervention did not replace the SAP presentation. The two interventions were integrated, with the PAN intervention adding 10 to 15 minutes to the previous standard alcohol presentation (SAP) that lasted about 45 minutes.

The PAN intervention had three major in-class components, with a fourth component that was delivered outside the classroom. First, the students listed reasons for drinking and the peer leaders called attention to the many social influences on drinking. Second, the students graphed (a) their own quantity and frequency of drinking, and (b) their perceptions of their peers' quantity and frequency of drinking. Then, the peer leaders revealed the mean quantity and frequency of drinking reported in recent campus-wide surveys, asked the students to add this information to the graphs, and then led a discussion that compared the students' perceptions of perceived alcohol norms with the actual survey data (i.e., correct normative data).

A major difference between the social marketing campaigns and the small group PAN intervention is that the students' misperceptions are examined interactively in a supportive environment. Based on the reactions in class and the responses in focus groups, the normative alcohol information often surprised the students because the discrepancy between their perceptions and the actual survey data tended to be large. Social psychologists have known for a long time that persons typically reject information that is at odds with their beliefs (Festinger, 1954). Thus, in this project, the students who had the greatest misperceptions were expected to be the least likely to be influenced by mass social marketing. In the small group setting, however, the students' peers can facilitate acceptance of the norming message via explanation, anecdotal corroboration, and personal examples that verify the social norming data.

Another method of reducing resistance to the norming message was to provide a logical explanation for how students may develop misperceptions of drinking norms. Thus, the third part of the PAN intervention was an interactive discussion of a hypothetical party in which a small minority of drunk and poorly-behaved students might be misperceived as representing the behavior of almost everyone at the party. Based on informal observations in classes and in the focus groups, this story had a remarkable effect on calming students' reactions when there was a discrepancy between perceived and actual drinking norms. The key components of this discussion are (a) showing how risky drinking is often overestimated due to the salience of the behavior of a few heavy drinkers, and (b) how persons who were not even at the party can become "carriers" of the misperception by telling stories about the extreme behavior of only a few of the partygoers. This discussion helped to explain how over-estimations about peer alcohol use are then likely to be formed, and how these misperceptions may overshadow the true social norm. The peer leaders then linked the concept of social influence on drinking with the idea of overestimation of norms and how that might lead to higher levels of personal drinking.

A fourth part of the PAN intervention was a campus-wide social norming media campaign. Thus, all students on campus were targeted to receive information about drinking norms via media materials, such as full-page advertisements in the student newspaper. An example of a media message is a full-page newspaper ad with an attractive photo of students stating that 71% of the students on campus have 0 to 4 drinks maximum when they party. As discussed later, the effect of this campus-wide intervention was not studied in this project. Thus, the direction and magnitude of the influence of the campus-wide media campaign is unknown.

Peer Leader Training

The student service arm of the USC Alcohol and Drug Education Program delivers the University 101 alcohol presentations. Students facilitated several small group discussions each semester, usually with a co-leader. Prior to leading the group discussions, peer leaders received a manual and several hours of supervised practice during a weekend retreat. The manual explicitly outlined how to implement the interventions and provided suggestions on how to answer difficult questions that were often asked by students in the University 101 classes. During the weekend retreat, the experienced peer leaders honed their skills and shared their knowledge with the rookie peer leaders. Also, the peer leaders assisted in making revisions to the presentations, thus possibly increasing their investment in the quality of the intervention.

RESULTS

Changes in Perceptions

This intervention effect was tested by looking for a significant group (PAN or SAP) by time interaction (pre- and post-intervention survey) intervention, which would suggest that one group changed more over time than the other. As expected, there were significant group by time interactions for perceptions of both the frequency of peer-perceived drinking, F (1,843) = 80.22, p < .01, and the quantity of peer drinking at F (1,848) = 29.74, p < .01. Examination of the group means (see Figures 1 and 2) shows that, comparing the pre to post assessments, greater reductions in perceptions of the quantity and frequency of peer drinking was observed in the PAN intervention relative to the SAP intervention.

Changes in Self-Reported Drinking

As with perceptions, this hypothesis was tested using repeated measures ANOVA to find a group by time interaction. Unlike with perceptions, however, the hypotheses regarding self-reported drinking were not fully supported. There was a significant group by time interaction, F (1,863) = 5.14,p < .01 for the self-reported frequency of drinking. Inspection of the means in Figure 1 shows that the PAN group showed a greater reduction than the SAP group. It is noteworthy, however, that in the pre-intervention survey, students in the PAN intervention reported that they drank more frequently than those in the SAP condition (see non-overlapping confidence intervals in Figure 1). Thus, the interaction is manifested such that the SAP group did not change and the PAN group reduced their drinking frequency to a level equivalent to the SAP group. This pattern of results makes it unclear if there was a therapeutic benefit of the SAP intervention on the frequency of drinking.

[FIGURE 1 OMITTED]

As indicated by a non-significant group by time interaction for quantity of drinking, there appears to have been no effect of the Caution for Figures 1 & 2: The error bars correspond to 95% confidence intervals. SAP refers to the Standard Alcohol Program. PAN refers to the Perception of Alcohol Norm intervention, which was the brief social norming intervention added to the SAP intervention. Self refers to self-reported drinking. Peers refers to perceptions of drinking by peers. PAN intervention on the self-reported quantity of drinking, F (1,862) = .96, N.S. Instead, students in the PAN and SAP interventions experienced similar reductions in quantity of drinking from the pre- to post-intervention surveys (see Figure 2).

[FIGURE 2 OMITTED]

Subgroup analyses

To assess the stability of effects across subsets in the sample, analyses were also conducted on the following subgroups: males (n = 323), females (N = 523), abstainers (N = 265), all drinkers (N = 597; excludes abstainers), and binge drinkers (N = 438). Because we were interested in how the PAN intervention interacted with pre-intervention drinking status, we used the pre-test criteria to identify the three drinking subgroups. To be considered an "abstainer", students had to report consuming no alcohol on all of the pre-intervention survey questions and endorse the response "I do not drink at all." The "all drinkers" subgroup consists of all students who reported drinking any amount on the pre-intervention survey. The "binge drinkers" subgroup includes only those students who reported having 4 or more drinks for females or 5 or more drinks for males in one sitting in a typical week on the pre-intervention survey.

There were no subgroup by intervention by time interactions. However, we pursued the possibility of subgroup differences further by replicating with each subgroup the repeated measures ANOVA analysis for peer perceptions and self-reported drinking. As with the total sample, there was no effect of the PAN intervention on the quantity of drinking in any subgroup. There were greater reductions in perceptions of the quantity and frequency of peer drinking in the PAN versus SAP treatment conditions for females, drinkers, and binge drinkers. Unlike the total sample, there was no effect of the PAN intervention on perceptions among males and abstainers, and the subgroup analysis found no effect of the PAN intervention on the frequency of drinking in any subgroup. Thus, the subgroup analyses suggest that the PAN intervention is not differentially effective for any particular group because all groups showed no change in measures of drinking.

Path Modeling

Two sets of path models were conducted using the SAS Proc Calis program. Each of these analyses examined three sets of models. Model 1 is the mediating model in which all effects of the PAN intervention on serf-reported drinking are constrained to be mediated by perceptions of peer drinking. Model 2 tests for unanticipated direct effects in which all effects of the PAN intervention are modeled as having no effects mediated by perceptions. Model 3 is the mixed model in which direct and indirect effects are modeled simultaneously. To control for baseline differences between subjects, paths were allowed between pre- and post-test measures of perceptions of peer drinking and self-reported drinking. Also, perceptions of peer drinking and self-reported drinking were allowed to covary at pre-test. All of the path models had fit indices in the acceptable range (e.g., NH and NNFI greater than .9), except for a significant model Chi-square (p < .05) which is usually discounted with large samples such as in this study (Hu & Bender, 1995).

In the models pertinent to the frequency of drinking, their worst fit was with the direct effects model (Model 2) and there was an equally good fit between the mediating and mixed models (Models 1 and 3, respectively). This is based on the Chi-Square for the models which was as follows: Model 2 (143.74), Model 1 (49.79), and Model 3 (49.78). Using the parsimony criterion (Hu & Bender, 1995), such that the model with equivalent fit but fewer paths is preferred, Model 1, the mediating model, should be preferred to the mixed effect model.

In this mediating model for drinking frequency, the effect of the PAN intervention on perceptions of the quantity of peer drinking was small (.30, t = 10.00,p < .001) and the effect of perceptions of other peer drinking on self-reported drinking was very small (.14, t= 7.12,p < .001). These paths are multiplied to get the effect of the PAN intervention on self-reported drinking that was mediated by perceptions. The resulting tiny effect size of .04 is consistent with the non-significant difference between the PAN and SAP groups in the mean quantity of self-reported drinking at post-test.

In the models pertinent to the quantity of drinking, the direct model (Model 2) had the highest Chi-Square (105.09), followed by the mediating only model (Model 1 Chi-Square = 51.27), then the mixed model (Model 3 Chi-Square = 44.29). The difference in Chi-square between Models 1 and 3 was statistically significant (6.98, df = 1, p < .05), so the mixed model is statistically superior for the quantity of drinking. In this model, the effect size of the PAN intervention on the quantity of drinking was calculated by adding (a) the mediating path (i.e., the product of the path for the effect on perceptions times the path between perceptions and self-reported quantity of drinking) with (b) the direct effect of PAN on self-reported drinking. The effect on the PAN intervention on the perceived quantity of peer drinking was .24 (t = 7.94,p < .001) and the path between peer perceptions and self-reported quantity of drinking was .09 (t = 4.95, P < .001). The product of these paths (.022) is added to the direct effect of the PAN intervention on drinking (-.05, t = 2.65, p < .01) to give a total effect size of -.03. This tiny effect is consistent with the non-significant intervention by time interaction for quantity of drinking.

DISCUSSION

These results replicate and extend our previous finding that the PAN intervention resulted in significant changes in perceptions of peer drinking but no meaningful effect on self-reported drinking. Subgroup analyses suggest that the results on drinking were consistently disappointing across the whole sample. In fact, the PAN intervention may have been less effective on the perceptions of peer drinking by males and abstainers than on the perceptions of females, drinkers, and binge drinkers. Path modeling showed that there were small positive effects of the PAN intervention on perceptions of the frequency and quantity of drinking (.30 and .24, respectively) but only a very weak correlation between self-reported drinking and perceptions of drinking by the typical student (i.e., only .14 for frequency and .09 for quantity). This translated into tiny effects of the PAN intervention through the theoretically specified mediating variable of perceptions, which is consistent with the disappointing results of the ANOVA.

The path modeling added some important information that was not revealed by the ANOVA. The model focusing on the frequency of drinking indicated that the effects, albeit small, of the PAN intervention on frequency were completely mediated by perceptions. Thus, for frequency of drinking, the PAN intervention was true to SNT. However, the effects of the PAN intervention on the quantity of drinking was a very small positive effect mediated via perceptions plus a very small negative direct effect. These effects combined to produce what appears to be a very small, statistically nonsignificant, negative effect of the PAN intervention on the quantity of drinking. Thus, the path model suggests that the PAN intervention should be examined for content that might encourage higher quantity of drinking. The negative direct effect is an important evaluation consideration that we would not have found if we had limited our analysis to ANOVA.

As with all studies, this study has several important limitations. Before reviewing these limitations, however, we should stress that this study has many strengths including a large sample size, randomization to treatment condition, psychometrically sound measures, and a comprehensive data analysis. The students who led the interventions were carefully trained and supervised. After a site visit during which he reviewed the training manual and observed the students delivering the PAN intervention, an expert consultant on social norming theory determined that the intervention was theoretically sound. We also submit that similar small group interventions have been promoted in The Report on Social Norms (2001), which includes a mention of the PAN intervention as a promising intervention. Thus, we rigorously evaluated a carefully planned, face-valid intervention based on SNT that is being disseminated widely throughout colleges and universities in the United States. Therefore, our results should have important implications for assessing both the theoretical validity of SNT and the efficacy of a subgroup of brief interventions based on SNT.

With respect to limitations, one of the limitations of this study is the lack of a no-treatment comparison group. To provide the least ambiguous estimates of intervention effect sizes it would be helpful to not only compare the two treatment groups with each other (i.e., SAP and PAN), but also compare the treatment groups with a no intervention control group. However, there are ethical arguments against withholding an alcohol intervention to students. This is in spite of the fact that programs such as SAP have the reputation of being very weak and have an estimated effect size of around .17 (see Latimer and Cronce, 2002). Thus, we did not feel it was ethically defensible to have a no treatment control group. Consequently, our design puts PAN to a particularly stringent test.

Another consideration that may have lessened the estimated effect size of PAN was that this was conducted in the midst of a media campaign on campus that included a weekly ad printed in the student newspaper about drinking norms on campus. This may have confounded the comparison between the PAN and the SAP intervention. However, focus groups with students in University 101 classes suggest that this media campaign may have minimally influenced students who did not get the PAN intervention. In the focus groups the majority of students who were not in the PAN intervention either were confused by the media message or did not believe the media message. It was only students who had been in the PAN intervention that really noticed the ad, understood it, and believed it to a certain degree. Moreover, students in the PAN intervention decreased their perceptions much more than students in the SAP intervention. This indicates that, at the very least, the media campaign did not completely obscure the effects of the PAN intervention. To the contrary, it appears that the PAN intervention may have made students more receptive to social norming messages, thus creating a PAN by media campaign interaction. Accordingly, in the methods section we described the media campaign as part of the PAN intervention.

This study is also limited in the fact that the sole data source is a self-report. It is possible that students may respond to the questionnaire in a socially desirable manner and under-report their drinking behaviors. There is concern that self-reports do not corroborate with students' actual behavior. To date, however, self-reports are one of the best and most economical ways in which to obtain data about drinking-related behaviors (Baer, Stacy, & latimer, 1991).

Another potential limitation was that there was no effect of the intervention on males, either for perceptions or actual drinking. This could be due to the fact that most of the peer leaders implementing the PAN intervention were female. It is possible that males did not find the peer leader credible or reliable. Because of this potential effect of gender on the effect of the PAN intervention, in future studies it may be important to (a) recruit a diverse mix of male and female peer leaders and (b) examine the effect of gender on intervention outcomes.

Finally, one might argue that our intervention was too weak and that a stronger SNT intervention might have produced meaningful change in self-reported drinking. Our data suggests that simply making the PAN intervention stronger will not produce the desired effects on drinking. This is based on the observed small correlations between the perceptions and self-reported drinking found in the path models (i.e., .14 for frequency and .09 for quantity) suggest that an intervention that had a huge effect on perceptions would have only a small effect on drinking. For example, if there was a huge effect (e.g., .90) of the PAN intervention on perceptions there would be only a small effect of the intervention on drinking (e.g., .13 for frequency and .08 for quantity). Given the extensive resources that would be needed to achieve a huge effect on perceptions, using a stronger version of the PAN intervention that focuses on the drinking of the typical student does not seem to be a fruitful or economical intervention method.

Given that our path models suggest that simply having a stronger effect on perceptions will not lead to major changes in drinking, we are faced with the dilemma of either modifying our social norming intervention or seeking an alternative approach. A modification in SNT that might be worth exploring is the possibility that the target group in the PAN intervention needs to be changed from the "average or typical student" to more specific, meaningful subgroups, such as friends or a group with whom the student personally identifies. Germaine to this point, Thombs and Miller (2002) speculated that their social norming intervention may not have worked because it failed to produce changes in perceptions of drinking by friends even though their intervention resulted in significant changes in perceived drinking of other campus groups. Another possible modification is to address the potentially latrogenic feature that might inadvertently encourage higher quantities of drinking. However, the observed harmful effect of PAN was small (-.05) and removing this influence would probably not substantially change the impact of the PAN intervention on drinking.

With regard to alternatives to SNT, there are two promising possibilities presented in a recent review by Latimer and Cronce (2002). One approach is brief motivational feedback. However, this intervention was usually delivered as a multi-session individual intervention with very high-risk students. Therefore it is not clear if this intervention is feasible or appropriate for a general population of students. A group-based alternative for a general population of students is the Alcohol Skills Training Program (ASTP) developed by Marlatt and colleagues (see Latimer and Cronce, 2002). For campuses such as ours where there is strong infrastructure for implementing group interventions, the ASTP program should be a relatively economical and effective intervention option.

In conclusion, this rigorous evaluation of SNT found significant effects on perceptions of peer drinking and trivial changes in self-reported drinking. The path analysis suggests that simply strengthening the PAN intervention will not get the desired results. Thus, we are caught in a dilemma between entering uncharted intervention territory in an effort to revise PAN or adopting an intervention that has a better research base than PAN, such as the ASTP. Our experience suggests that SNT interventions should be regarded as highly experimental and should only be implemented in the context of rigorous evaluations. It would be interesting in future studies to compare a modified version of PAN (e.g., with a focus on perceptions of friends' drinking) with an established intervention like the ASTP.

References

Baer, J.S., Stacy, A., & Latimer, M. (1991). Biases in the perception of drinking norms among college students. Journal of Studies on Alcohol, 52, 580-586.

Berkowitz, A.D. (2001). Social norms interventions with small groups. The Report on Social Norms, 1, 1-4.

Berkowitz, A. D. (2003). The social norms approach: Theory and research. Internet article found at http://www.edc.org/hec/ socialnorms/theory.html

Cohen, J. (1992). A power primer. Psychological Bulletin, 112, 155-159.

Festinger, L. (1954). A theory of social comparison processes. Human Relations, 7, 117-140.

Hansen, W.B. (1992). School-based substance abuse prevention: A review of the state of the art in curriculum, 1980-1990. Health Education Research, 7, 403-30.

Hu, L. & Bender, P.M. (1995). Evaluating model fit. In R. Hoyle (Ed.), Structural equation modeling, issues, concepts, and applications (pp. 76-99). Newbury Park, CA: Sage Publications.

Larimer, M.E., & Cronce, J. M. (2002). Identification, Prevention, and Treatment: A review of individual-focused strategies to reduce problematic alcohol consumption by college students. Journal of Studies on Alcohol, Supplement 14, 148-163.

Schroeder, C.M., & Prentice, D. A. (1998). Exposing pluralistic ignorance to reduce alcohol use among college students. Journal of Applied Social Psychology, 28, 2150-2180.

Smith, B.H. (2003). Social Norms Theory to Reduce Problematic Drinking Among College Students: It is time for Experimentation, not Dissemination. International Center for Alcohol Policy Invited Opinion Series. www.icao.org/invited/invited.html

Smith, B.H., Stamper, G.A., & Gant, R.L. (in press). Changing misperceptions about college student drinking: Evaluation of a small-group, peer-led social norming intervention. Journal of Drug and Alcohol Education.

Thombs, D.L. & Hamilton, M.J. (2002). Effects of a social norm feedback campaign on the drinking norms and behavior of Division I student-athletes. Journal of Drug Education, 32, 227-244.

Wechsler, H., Lee, J. E., Kuo, M. C., & Lee, H. (2000). College binge drinking ill the 1990s: A continuing problem Results of the Harvard School of Public Health 1999 College Alcohol Study. Journal of American College Health, 48, 199-210.

Georgia Ann Stamper, Bradley H. Smith,

Rick Gant, Kristin E. Bogle 

University of South Carolina, Columbia