Are Our Recruitment Practices for Sex Studies Working Across Gender? The Effect of Topic and Gender of Recruiter on Participation Rates of University Men and Women.

Research on sexual behavior, sexual coercion, and related topics in university settings has been plagued by uneven proportions of men and women in the final samples. This is an observation that became clear to us in our own research and is confirmed by examining the sex breakdowns in the published work of other researchers (e.g., Abbey & Hamish, 1995 - 70% female; Grant, Folger, & Homak, 1995 - 73% female; Patton & Mannison, 1995 - 73% female; Perrott & Webber, 1996 - 84% female). While some studies have reported more balanced numbers of men and women, it is often the case, as it was in one of our studies (Campbell, Schellenberg, & Senn, 1997), that the comparable number of women and men is illusory--created by continued efforts to recruit male participants long after sufficient numbers of females are recruited or by excluding some female data to balance the n's (e.g., Struckman-Johnson & Struckman Johnson, 1992). An obvious etiology of the problem is the unequal number of males and females in psychology courses. However, research done prior to 1975 on general trends in recruitment and volunteering (see, for e.g., Rosnow & Rosenthal, 1976) suggested that the gender discrepancy in participation may also be due, at least in part, to increased volunteering of women in general, and to female students volunteering for studies where recruitment is done in undergraduate classes.

Whether this difference in volunteering exists or not, we hypothesized that our subject matter, the study of sexual behaviors and attitudes, was accentuating the problem. Though research has shown that volunteer effects are considerably less for questionnaire research like ours than for more intrusive sex research (Clement, 1990; Strassberg & Lowe, 1995; Wolchik, Braver, & Jensen, 1985), the subject matter may still have some influence on participation. Countering this hypothesis is the research suggesting that intrusive topics could decrease female volunteering, indirectly increasing the proportion of males in the final samples. For example, Rosnow and Rosenthal (1976) reported a reversal of the pattern of greater female volunteering when the topics were "physically or emotionally stressful" (p. 99). It is possible therefore, that the trend of lower participation by men we observed is present in all types of studies and is not only a problem with sex research.

Morokoff (1986) summarized the rate of volunteering specifically for sexuality studies using research which allowed breakdown by participant gender. Eight investigations were included with methods ranging from questionnaire administration to video exposure to physiological measurement. The earliest study reported was conducted in 1956, the latest in 1985. In three of these studies, men volunteered at higher rates than women; in the remainder there were no sex differences. Across studies the volunteering rates for both men and women ranged from 15 to 73 percent.

There are no similar data available for sexual coercion studies, as researchers have not appeared to spend the same amount of time writing about recruitment and sampling biases in this literature. Some of the best-known research in the sexual coercion field has avoided these issues by utilizing captive student audiences, that is, students who are given class time to complete questionnaire packages (e.g., DeKeseredy & Kelly, 1995; DeKeseredy, Schwartz, & Tait, 1993; Koss & Dinero, 1988; Koss, Leonard, Beezley, & Oros, 1985; Malamuth, Socklowskie, Koss, & Tanaka, 1991; Schwartz & DeKeseredy, 1997). This mode of participation is discouraged or disallowed at many institutions; therefore, recruitment in the field tends to be carried out using the more usual methods (Lewin, 1985; Shotland & Goodstein, 1992; Smith, Pine, & Hawley, 1988), including asking students to volunteer by signing up on bulletin boards (e.g., Gidycz, Nelson Coble, Latham, & Layman, 1993) or sign-up sheets in class for later research appointments. We have found no studies of sexual coercion that use these more usual methods of recruitment and provide data that would allow the calculation of the response rate. In fact, sexuality studies appear to be one of the only areas for which researchers report response rates with university student volunteers. Since most sexual coercion studies use a cover story (e.g., "Legal Decision Making," Smith et al., 1988) or describe the study more broadly as a study of sexual experiences and attitudes, sexuality studies probably give us the best indicator of possible response rates.

For questionnaire studies, the best recent estimate of volunteering rate comes from Strassberg and Lowe (1995), who had university undergraduates complete a few questions related to sexuality as part of a first year "mass testing" and then indicate whether or not they would be willing to participate in a subsequent sexuality questionnaire study. Forty-eight percent of the students agreed to participate, with 77% of those students showing up for the study which resulted in a final response rate of 37%. There was no sex difference found in rate of volunteering. However, this study does not provide a comparison topic which would allow us to judge whether this response rate is the same or different from recruitment on other topics.

We designed the current study to test whether the topic of research is responsible for the disproportionate representation of males to females in university samples in a context that is common to recruitment practices in many university psychology departments (Landrum & Chastain, 1999; Lindsay & Holden, 1987). Recruitment occurred in the first 2 weeks of classes with bonus grades being given for participation. Because recruitment took place early in the semester, no student had already fulfilled their bonus grade quota. Common recruitment wisdom (e.g., Rosnow & Rosenthal, 1976) suggested that recruitment should ideally be done by a woman. We wondered whether this advice still held in the late 1990s. There was also the possibility that same-gender recruiters would make students more comfortable or confident about participating in a study, particularly when the topic involved sexuality (Wiederman, 1999). We therefore used both a female and a male graduate student as recruiters and included the gender of the recruiter as a variable in our study. Thus, the study tested whether the probability of volunteering for, and later participating in, a study differed as a function of the gender of the participant, the topic of the research, the gender of the recruiter, or by some combination of these three factors.

METHOD

Participants

Recruitment of undergraduates in the psychology department at this Canadian university had to be done in person in the first 5 minutes of introductory psychology classes.(1) This procedure required that random assignment of students to the experimental conditions be done in blocks rather than individually. Fifteen sections of introductory psychology were offered in the winter 1997 term. One section was very large (n = 405), with the others varying in size from 25 to 105. The largest and smallest sections were excluded from the recruitment process to ensure roughly equal final n's following assignment of the blocks. One large section was not included because we were unable to contact the instructor. Thus, the potential participant group included 12 sections of the course with 789 students registered. Four hundred and ninety-seven (63%) of these students were female. Two hundred and ninety-one students (37%) signed up for the study and, of these, 153 showed up for the study. Thus the overall recruitment success was conservatively estimated at 19% of the total number of students available to participate. Of those who volunteered by signing up, 53% actually showed up for their appointments. Students received a 1% bonus grade for participating.

Procedure

We were unable to gain permission to stand in classes long enough to count manually the number of males and females present on the day of recruitment. Therefore, we used the next best available method. The registrar of the university provided a breakdown of student sex by section for the introductory psychology course. These figures were used to estimate the total number of potential participants available for recruitment. We acknowledge that these figures are overestimates since attendance is rarely perfect.

Two graduate research assistants, one male and one female, learned the recruitment speeches. Both assistants were attractive, were in their mid 20s, and had good communication skills. They worked together to practice these speeches until they were able to deliver them enthusiastically and uniformly. The recruitment speeches were identical except for the topic. One version of the recruitment asked students to participate in a study of memory and real life, the other in a study of sexual and social experiences in campus life (our cover title for sexual coercion and sexuality research). These generic descriptions had similar content to those used in other studies in the department with one exception. We did not want to risk influencing the participation rates by having large numbers of debriefed students talking about the study and its purpose before their peers had participated. We reduced the likelihood of this occurring by running the entire study in a 1-week period and telling students that if they missed their appointment they would not have a second chance.

The 12 sections of introductory psychology were randomly assigned to the four conditions (topic X gender of recruiter). Each assistant went into the assigned classes and delivered the appropriate recruitment speech. The assistant left the room once all questions had been answered. A recruitment folder was passed around the class and students signed up for a convenient time. A variety of times of day and days of the week were offered, including night-time slots for night students. The folder always made it all the way around the room and was picked up at the end of the class by the assistant. Students tore off a reminder slip that contained the place, time, and date of their scheduled appointment. No reminder calls were made.

When participants showed up for their appointment, they were given a one-page explanation of the purpose of the study stressing the importance of participant recruitment strategies for psychological research. They signed a form to ensure registration of their one bonus grade for participation while the research assistants surreptitiously recorded the sex of the participant. Any questions or concerns were addressed at this time.

Data Analytic Strategy

Data for the study consisted of frequency counts of the number of potential participants in each of the conditions, the frequency and gender of students who signed up for the study, and the frequency of women and men who later showed up for the study (see Table 1). These data were analyzed by multinomial logit model which tests the significance of all possible models that combine the various categorical independent variables used in the study, along with their interactions, on a binary response measure (Demaris, 1992). In this case, the response was whether or not the participants chose to sign up or show up for the study. A logit model is the most applicable technique for these data given that the distribution of responses were clearly uneven, with probabilities of signing up or showing up for our study often below 25% which violates parametric assumptions (Tabachnick & Fidell, 1996).

Table 1. Frequency of Potential Participants - Students Who Signed up
and Students Who Showed up

                           Potential       Potential
Topic by                  pool - male    pool - female
student sex                recruiter       recruiter

Sex study                     217             196
  Males                        83              63
  Females                     134             133
Memory study                  164             212
  Males                        63              83
  Females                     101             129
Entire study
  Males (% of total)
  Females (% of total)

                           Potential       Signed up      Showed up
Topic by                 pool - across    (% of those    (% of those
student sex               recruiters      recruited)     recruited)

Sex study                     413          194 (47)       101 (25)
  Males                       146           52 (36)        29 (20)
  Females                     267          142 (53)        72 (27)
Memory study                  376           97 (26)        52 (14)
  Males                       146           32 (22)        14 (10)
  Females                     230           65 (29)        38 (17)
Entire study                  789          291 (37)       153 (19)
  Males (% of total)          292 (37)      84 (29)        43 (15)
  Females (% of total)        497 (63)     207 (42)       110 (22)

Logit models require testing the significance of each independent variable individually or in combination, along with their interactions, to determine which of the models provides the best fit of the data under the most parsimonious conditions (Knoke & Burke, 1980) (see Table 2). For each model tested, expected cell frequencies for each condition are estimated and compared with the observed frequencies using a likelihood-ratio statistic ([L.sup.2]) which is similar to a chi-square. If the [L.sup.2] statistics is significant, the model is rejected. In circumstances where two models provide good fit of the data (i.e, are not significant), they can be compared by computing a chi-square test of the difference between the [L.sup.2] statistics of the two models.

Table 2. Fit of Models Predicting Signing up for the Study

Model no.   Variables entered               [L.sup.2]   df     p

 1          P                                 47.86     6    <.0001
 2          T                                 22.27     6    <.002
 3          R                                 41.11     6    <.0001
 4          P, T                              17.99     5    <.01
 5          P, R                              36.67     5    <.0001
 6          R, T                              14.11     5    <.02
 7          P, T, R                            9.39     4     .052
 8          P, T, R, P(*)T                     8.47     3     .037
 9          P, T, R, P(*)R                     9.09     3     .028
10          P, T, R, T(*)R                     4.56     3     .207
11          P, T, R, P(*)T, P(*)R              8.09     2     .018
12          P, T, R, P(*)T, T(*)R              3.45     2     .176
13          P, T, R, P(*)R, T(*)R              4.23     2     .120
14          P, T, R, P(*)T, P(*)R, T(*)R       2.97     1     .085

Note. N for all models = 789. P = Participant gender; T = Topic
of study; R = Recruiter gender.

Once the best fitting and most parsimonious model is determined, the results of a logit model can be interpreted in a manner similar to an Anova. The model provides a test of significance (z test) for each main effect or interaction entered (see Table 3). An added feature of this technique is that the odds of a particular response (e.g., to sign up for the study) can be calculated for each significant independent variable or interaction. We use this approach in the result section to estimate the odds of signing up or showing up for the study as predicted by each significant independent variable and interaction included in the model that provides the best fit of the data.

Table 3. Estimated Parameter Effects and Calculated Odds of Signing up
for the Study for Best Logit Model

                            ([Tau])                 Odds of signing up
Variables                  Parameter    z value     (times more likely)

Participant gender           -0.36     -2.24(*)
   Women                                                    1.41
   Men                                                      0.71
Topic of study               1.13       5.11(***)
   Memory study                                             0.43
   Sex study                                                2.29
Recruiter gender             0.73       3.62(**)
   Woman                                                    0.62
   Man                                                      1.62
Interaction term (Topic      0.68       2.19(*)
 X Recruiter gender)
  Memory study
   Female recruiter                                         0.94
   Male recruiter                                           1.06
  Sex study
   Female recruiter                                         0.49
   Male recruiter                                           2.03

(*) p < .05. (**) p < .01. (***)p < .001.

RESULTS

Manipulation Check of Potential "Cluster" Effects

A potential issue with the design of this study is that response rate may differ as a function of the cluster (in this case, the class) in which one is recruited. Several characteristics of the class (e.g., high rates of expressed participation, very negative or positive class atmosphere) may have influenced the degree of participation of students in that class independent of the condition to which the class was assigned. To determine whether the effects of our assigned variables were independent of cluster effects, we first compared the response rate of each set of three classes assigned to each condition. For all four conditions (i.e., sex study with male and female recruiter, and memory study with male and female recruiter) the data revealed no significant cluster effect, [[Chi].sup.2] (2, Ns from 164 to 216) = from 2.34 to 4.89, all ps [is greater than] .087.

Results of Logit Model Analyses

Goodness of fit data for the various models predicting signing up for the study are reported in Table 2. Model 7 includes all main effects and shows a poor fit of the data. Models 1 to 6 are even more parsimonious as they include only one or two independent variables, but they clearly do not fit the data. Of all the models tested, only models 10, 12, 13, and 14 fit the data relatively well (i.e., the [L.sup.2] statistics are not significant). A comparison of these models indicate no significant difference in fit ([[Chi].sup.2] values between 0.33 and 1.59, n.s.). For parsimony we selected model 10, which includes all main effects and the topic of study X recruiter gender interaction, as the best prediction model ([L.sup.2] = 4.56, df = 2, N = 798, p = .20).

In Table 3, we report the estimated tau parameters ([Tau]) for each variable entered in model 10 as well as their respective z values. The [Tau] parameters are similar to the coefficients in ordinary regression analysis. We also report the calculated odds of signing up for the study that result from each significant variable.

Table 3 shows significant effects for all three independent variables and for the interaction of topic of study X recruiter gender. The variable with the largest tau coefficient was the topic of study ([Tau] = 1.13, z = 5.11, p < .001) where the odds of having participants sign up for the sex study were 2.29 times higher than the odds of signing up for the memory study (47% versus 27% respectively). Recruiter gender also had a strong significant effect on recruitment ([Tau] = 0.73, z = 3.62, p < .01). The odds were 1.62 times higher that students who were recruited by a man would sign up for the study in comparison to students recruited by a woman (43% versus 32% respectively). However, these two main effects are explained by a significant topic of study X recruiter gender interaction ([Tau] = 0.68, z = 2.19, p [is less than] .05). The likelihood of participation hardly differs in the memory study condition where 27% of students recruited by a woman and 28% of students recruited by a man signed up for the study, [[Chi].sup.2] (1, N = 376) = .06, n.s.. Thus, the odds of participation in the memory study condition are about equal for female and male recruiters (0.94 and 1.06 for female and male recruiters, respectively). However, we find a highly significant difference in the sex study condition, [[Chi].sup.2] (1, N = 412) = 12.49, p [is less than] .001. Here, 55% of participants signed up for the study when recruited by a male and only 37% signed up when the recruiter was a female. In this condition, the odds of participation if recruited by a man were 2.03 times higher than when recruited by a woman.

The gender of participant variable also shows a significant effect on recruitment ([Tau] = -0.36, z = -2.24, p [is less than] .05). Overall, women were more likely than men to sign up for a study regardless of the condition to which they were assigned (40% versus 32% for women and men respectively). The calculated odds of having women sign up for a study were 1.41 times higher than they were for men.

We also used logit model technique to determine whether our independent variables could predict the odds of showing up for the study. We tested all 14 models used to predict whether participants would sign up for the study, but none of the models fit the data adequately. We also reran these analyses with only those students who had signed up for participation as opposed to all potential participants. Again, none of the models fit the data. These findings suggest that the results for showing up are likely confounded within the results of signing up for the study and would not be meaningful if interpreted by themselves.

DISCUSSION

We designed this study to examine the effects of research topic and recruiter gender on the volunteering rates of male and female students. One possible explanation for the low numbers of men in published studies of sexual coercion is simply their lesser availability in the undergraduate Psychology classroom. Our findings confirmed that while there is an imbalance in enrollment by gender, male participants are further reduced in number because men are not volunteering for research at a rate equal to their enrollment. While men made up 37% of the students in Introductory Psychology classes, on average only 29% of the men chose to volunteer. By contrast, 42% of females signed up for the research. This differential volunteering is then magnified by the much larger number of women in the classes so that women become 71% of the volunteer pool while men end up being only 29% of the pool. Whether this gender difference in volunteering is due to the somewhat greater cooperation or helping behavior of female students for social tasks (Eagly & Crowley, 1986), differential attendance in classes (a possibility we could not find any support for in the literature), or some other factor is unclear. What is obvious, however, is that if the research knowledge gained on the basis of recruiting undergraduate volunteer samples is to reflect the experiences and views of male and female participants equally, greater efforts must be made to encourage men to volunteer in our studies. While this phenomenon appears to be a general problem not specific to sex research, it may be of even greater significance in studies of sexuality, sexual coercion, and other types of research where men and women can be expected to have greater differences in attitudes and experiences than in other areas of their lives.

Beyond the issue of differential volunteering by sex, the current study confirmed previous research findings (e.g., Rosnow & Rosenthal, 1976) that people do not volunteer for research at random. The empirical research on mail surveys has suggested that sensitive topics or questionnaires with sensitive questions receive lower response rates than more neutral ones (Dillman, Sinclair, & Clark, 1993; Rosnow & Rosenthal, 1976: see also Johnson et al., 1989, for their discussion of their extremely low response rates for mail surveys on sexuality). In light of this, it was unexpected that the sex research condition had almost twice the volunteering rate of the more mainstream research condition. Volunteering rates for the sexual experiences study replicated Strassberg and Lowe's (1995) sign up rates (47% compared to their 48%) and show up rates (25% compared to their 37%) and suggest that sex and sexual coercion researchers can count on approximately half of the available students volunteering with standard classroom recruiting. When put along side Lindsay and Holden's (1987) claim that just over half of the Canadian students eligible to participate in research in extra credit systems do so, our data suggest that recruitment for sex studies is likely to obtain the maximum volunteering possible within this system.

We wondered whether this finding of much higher volunteering for sex research is an artifact of the age of our sample (estimated based on other studies with this group to have a mean age of 21 or 22) compared to the older age of people who generally receive mail surveys. Historical increases in the percentage of people who have experienced premarital heterosexual intercourse (Clement, 1990), increases in the number of sexual partners, and decreases in the age of first intercourse (Johnson et al., 1989) have been well documented. These changes would suggest that younger cohorts are more likely to be liberal in their sexual attitudes than older cohorts. Furthermore, volunteering for sex research is biased toward recruitment of participants with more liberal attitudes and beliefs about sex (Farkas, Sine, & Evans, 1978; Morokoff, 1986; Wolchik et al., 1985). It may therefore be the case that our sample, because of the disproportionate number of younger people, is reflecting a new trend toward heightened interest in sexual topics. This explanation is supported in part by the research on mail survey participation which has demonstrated that greater interest in a topic leads to dramatic increases in response rates (Martin, 1994).

Another potential explanation for the higher rate of volunteering for the sex study condition may be that university-based studies, with their greater perceived anonymity (large classes, studies run in groups, etc.), are less daunting than questionnaire studies where materials are mailed to one's home (and therefore it can be assumed by the participant that the researcher knows where they live even if they have been promised anonymity through special mailback procedures). It is possible that the difference in setting could affect the way personal topics impact on participation rates. To our knowledge, this is the first study done in a university setting that has demonstrated significant increases in the rates of volunteering for sex studies over other more mainstream topics for both men and women. Research by Blumberg, Fuller, and Hare (1974) pointed out that while many researchers assume that questionnaires with "a more interesting title and content ... may be expected to yield a better response, ... no one seems to have done systematic experiments varying content in this way" (p. 117-118). This study has tested the assumption and found support for it, but we await replication of the finding.

The use of female recruiters was recommended by Rosnow and Rosenthal (1976) in order to maximize volunteering. This advice has been adopted by many researchers such as Wiederman (1999) who used only a female researcher in his recruitment effort. He wondered about "[t]he extent to which the sex of the researcher may have [positively] influenced respondents' perceptions of the research and expressed willingness to participate ..." (p. 65). Our findings contradict these suggestions, at least for questionnaire methods. The combination of a male recruiter and sex research resulted in a volunteering rate of more than 50% for the study, almost double the rate obtained for the memory study. Furthermore, there was no positive effect of female recruitment even for female students. This result contradicted our previous practice and empirical wisdom to date, and so was a most interesting finding for which we have no definitive explanation. In the past researchers have assumed that comfort was increased by the presence of female recruiters and that comfort was critical to recruitment on sensitive topics. It may be that some other factor is a stronger influence on volunteering in University samples. For example, if perceived expert status affects recruitment, as Rosnow and Rosenthal (1976) suggest, then volunteering rates would be affected if participants perceive that men have more expertise on sexual topics than women. Future research should attempt to replicate and examine possible explanations for this finding. If replicated, this situation would be particularly troubling for female undergraduate and graduate researchers who often cannot afford to hire male assistants.

We are also concerned about the drop-out rate of almost 50% between the time participants volunteered for the study and when they are supposed to show up to take part in the research. This is a frustrating and time-consuming part of conducting research that can be reduced with reminder phone calls (Klawsky & Peterson, 1990) and by rescheduling missed appointments. The good news, however, is that the drop-out rate does not appear to further bias the demographic makeup of the group we are studying (as evidenced by the lack of significant findings at the "showing up" stage of recruitment). This finding would suggest that increasing initial volunteering should be the focus for obtaining a representative sample.

In contrast to authors of published research who use mail surveys, authors of published research conducted with university student samples very rarely present overall response rates based on the population of students recruited in classes. Our study would suggest that these rates are much lower than readers of the research might think, and often much lower than is deemed appropriate in survey research (i.e., minimum 50-70% - Dillman, 1991; Neuman, 1994). While the rates found here appear to be at the high end of what can be expected in extra-credit systems, it is clear that researchers should pay more attention and develop better methods to increase all students' involvement in the research process. Targeting male students who are represented in even lower numbers in final participant pools should be a high priority. We suggest that focus groups conducted with male introductory psychology students may need to be carded out to understand the reasons for non-volunteering among male students and to provide some solution to this methodological problem.

(1) This policy has since changed. The data collected here was used to support proposals for a participant pool that is now in place.

The study was supported by a grant from the Social Sciences and Humanities Research Council of Canada (410-94-0878). Our thanks is extended to Corinne Hardy-Morais and Brent Abello for their enthusiasm in gathering data for this project.

REFERENCES

Abbey, A., & Harnish, R. J. (1995). Perception of sexual intent: The role of gender, alcohol consumption, and rape supportive attitudes. Sex Roles, 32, 297-313.

Blumberg, H. H., Fuller, C., & Hare, A. P. (1974). Response rates in postal surveys. Public Opinion Quarterly, 38, 113-123.

Campbell, B., Schellenberg, E. G., & Senn, C. Y. (1997). Evaluating measures of contemporary sexism. Psychology of Women Quarterly, 21, 89-102.

Clement, U. (1990). Surveys of heterosexual behavior. Annual Review of Sex Research, 1, 45-74.

DeKeseredy, W. S., & Kelly, K. (1995). Sexual abuse in Canadian university and college dating relationships: The contribution of male peer support. Journal of Family Violence, 4, 41-53.

DeKeseredy, W. S., Schwartz, M. D., & Tait, K. (1993). Sexual assault and stranger aggression on a Canadian university. Sex Roles, 28, 263-277.

Demaris, A. (1992). Logit modeling. Newbury Park, CA: Sage.

Dillman, D. A. (1991). The design and administration of mail surveys. Annual Review of Sociology, 17, 225-249.

Dillman, D. A., Sinclair, M. D., & Clark, J. R. (1993). Effects of questionnaire length, respondent-friendly design, and a difficult question on response rates for occupant-addressed census mail surveys. Public Opinion Quarterly, 57, 289-304.

Eagly, A. H., & Crowley, M. (1986). Gender and helping behavior: A meta-analytic review of the social psychological literature. Psychological Bulletin, 110, 109-128.

Farkas, G. M., Sine, L. F., & Evans, I. M. (1978). Personality, sexuality, and demographic differences between volunteers and nonvolunteers for a laboratory study of male sexual behavior. Archives of Sexual Behavior, 7, 513-520.

Grant, J. M., Folger, W. A., & Hornak, N. J. (1995). College students' perception of victim responsibility in an acquaintance rape situation. College Student Journal, 29, 532-535.

Gidycz, C. A., Nelson Coble, C., Latham, L., & Layman, M. J. (1993). Sexual assault experience in adulthood and prior victimization experiences: A prospective analysis. Psychology of Women Quarterly, 17, 151-168.

Johnson, A. M., Wadsworth, J., Elliott, P., Prior, L., Wallace, P., Blower, S., Webb, N. L., Heald, G. I., Miller, D. L., Adler, M. W., & Anderson, R. M. (1989). A pilot study of sexual lifestyle in a random sample of the population of Great Britain. AIDS, 3, 135-141.

Klawsky, J. D., & Peterson, V. A. (1990). Four techniques for reducing pseudovolunteering. Psychological Reports, 66, 1087-1090.

Knoke, D., & Burke, P. J. (1980). Log-linear models. Newbury Park, CA: Sage.

Koss, M. P., & Dinero, T. E. (1988). Predictors of sexual aggression among a national sample of male college students. Annals of the New York Academy of Sciences, 528, 133-147.

Koss, M. P., Leonard, K. E., Beezley, D. A., & Oros, C. J. (1985). Nonstranger sexual aggression: A discriminant analysis of the psychological characteristics of undetected offenders. Sex Roles, 12, 981-992.

Landrum, R. E., & Chastain, G. D. (1999). Subject pool policies in undergraduate-only departments: Results from a nationwide survey. In G. D. Chastain & R. E. Landrum (Eds.), Protecting human subjects: Departmental subject pools and institutional review boards (pp. 25-42). Washington, DC: American Psychological Association.

Lewin, M. (1985). Unwanted intercourse: The difficulty of saying no. Psychology of Women Quarterly, 9, 184-192.

Lindsay, R. C. L., & Holden, R. R. (1987). The introductory psychology subject pool in Canadian universities. Canadian Psychology, 28, 45-52.

Malamuth, N. M., Socklowskie, R. J., Koss, M. P., & Tanaka, J. S. (1991). Characteristics of aggressors against women: Testing a model using a national sample of college students. Journal of Consulting and Clinical Psychology, 59, 670-681.

Martin, C. L. (1994). The impact of topic interest on mail survey response behavior. Journal of the Market Research Society, 36, 327-338.

Morokoff, P. J. (1986). Volunteer bias in psychophysiological study of female sexuality. The Journal of Sex Research, 22, 35-51.

Neuman, W. L. (1994). Social research methods: Qualitative and quantitative approaches (2nd ed.). Boston: Allyn and Bacon.

Patton, W., & Mannison, M. (1995). Sexual coercion in dating situations among university students: Preliminary Australian data. Australian Journal of Psychology, 47, 66-72.

Perrott, S. B., & Webber, N. (1996). Attitudes toward male and female victims of sexual assault: Implications for services to the victim. Journal of Psychology and Human Sexuality, 8(4), 19-38.

Rosnow, R. L., & Rosenthal, R. (1976). The volunteer subject revisited. Australian Journal of Psychology, 28, 97-108.

Schwartz, M. D., & DeKeseredy, W. S. (1997). Sexual assault on the college campus: The role of male peer support. Thousand Oaks, CA: Sage.

Shotland, R. L., & Goodstein, L. (1992). Sexual precedence reduces the perceived legitimacy of sexual refusal: An examination of attributions concerning date rape and consensual sex. Personality and Social Psychology Bulletin, 18, 756-764.

Smith, R. E., Pine, C. J., & Hawley, M. E. (1988). Social cognitions about adult male victims of female sexual assault. The Journal of Sex Research, 24, 101-112.

Strassberg, D. S., & Lowe, K. (1995). Volunteer bias in sexuality research. Archives of Sexual Behavior, 24, 369-382.

Struckman-Johnson, C., & Struckman-Johnson, D. (1992). Acceptance of male rape myths among college men and women. Sex Roles, 27, 85-100.

Tabachnick, B. G., & Fidell, L. S. (1996). Using multivariate statistics (3rd ed.). New York: Harper Collins.

Wiederman, M. W. (1999). Volunteer bias in sexuality research using college student participants. The Journal of Sex Research, 36, 59-66.

Wolchik, S. A., Braver, S. L., & Jensen, K. (1985). Volunteer bias in erotica research: Effects of intrusiveness of measure and sexual background. Archives of Sexual Behavior, 14, 93-107.

Manuscript accepted December 13, 2000

Address correspondence to Dr. Charlene Y. Senn, Department of Psychology, University of Windsor,401 Sunset Avenue, Windsor, Ontario, Canada N9B 3P4; e-mail: csenn@uwindsor, ca.