Characteristics of men and women who complete or exit from an on-line Internet sexuality questionnaire: a study of instrument dropout biases.

Despite the increase of research into the Internet and sexual behavior, interpretation of Internet and sexuality data are limited by a lack of clarity regarding sample biases, among other methodological issues. Since it is very difficult to obtain a random sample of Internet users on sexuality-related sites, and since respondents to Internet questionnaires are usually subject to some sort of self-selection bias, there are difficulties in interpretation of data. Cooper, Scherer, Boies, and Gordon (1999) have already observed that respondents to website surveys may be more frequent Internet users or have a greater interest in the subject matter, and Ochs, Mah, and Binik (2002) have noted some of the sampling biases that relate to questionnaire data obtained from the Internet. A second area of potential bias is that of questionnaire completion: Anecdotal evidence has suggested that a large number of people who commence responding to sexuality-related Internet-based questionnaires drop out before completion of the questionnaires. Thus, this source of bias relates to whether those who complete the questionnaire are likely to be typical of those accessing and responding, or whether they are unrepresentative in systematic ways. Finally, the dropout rate itself is not well known, nor is the point at which dropout is most likely to occur or rates for various subpopulations including gender.

Research into the correlates of sexual contact on the Internet suggests that the Internet has a greater and more instantaneous reach than any other medium to facilitate encounters that result in face-to-face sexual activity (Bull & McFarlane, 2000). The potential for more efficient disease transmission is highlighted by a syphilis outbreak associated with men who have sex with men (MSM) seeking partners on the Internet (Klausner, Wolf, Fuscher-Ponce, Zolt, & Katz, 2000). McFarlane, Bull, and Rietmeijer (2002), using an on-line survey, found that young adults using the Internet to meet sexual partners had quite different patterns of sexual behavior compared to those who did not use the Internet to meet sexual partners. In a study comparing MSM who responded to a questionnaire on the Internet to those completing the same questionnaire distributed conventionally to members of a gay organization in Sweden, Ross, Tikkanen, and Mansson (2000) found that those responding to the Internet questionnaire were significantly younger and more likely to come from a rural area or small town, to be bisexual and have significant heterosexual experience, not to be as "out" as homosexual, and to be less well educated. While there were few differences in actual sexual behavior between the two samples, these data made it clear that Internet samples can differ on many important demographic variables from conventional questionnaire samples.

The question of potential biases in Internet research is a central one. Wallace (1999) indicates that there has been little solid research as to who participates on the Internet and at what level. Mustanski (2001) also notes that Internet studies frequently rely on volunteer altruism, although a study by Frick, Bachtiger, and Reips (2001) found that randomizing participants to an incentive or incentive-free condition with the same Internet-based questionnaire produced fairly small dropout rates, 9.5% compared with 18.5% respectively. Mustanski (2001) identifies participation bias as one of the crucial problems facing on-line research and suggests that this is one reason why Internet-based questionnaires are typically less than 15 minutes long. Important questions, however, are (a) at what point dropout is likely to occur and (b) what characterizes those who are likely to drop out compared with those who continue. Here, the use of Internet technology makes it possible to analyze the dropouts compared with a standard paper and pencil questionnaire where such a comparison would be difficult unless all incomplete questionnaires were returned. An understanding of this issue will help us better decipher the nature of the bias involved.

To assess the possible biases and their directions in those who complete versus those who do not complete an Internet questionnaire investigating Internet sexual practices, we carried out a study of those responding to an Internet questionnaire in Sweden. Our research questions were (a) what is the rate of dropout in respondents to an Internet questionnaire; and (b) what demographic characteristics, including gender, are associated with dropout versus retention?

METHODS

Procedure

The questionnaire, which was administered in the Swedish language, was launched through a Swedish portal site called Passagen (http://www.passagen.se) that is part of the Eniro Corporation. Passagen is ranked one of the top four Internet domains in Sweden considering the number of unique visitors. We placed a banner on the website for 2 weeks from June 10 to June 23, 2002, announcing our study; it appeared randomly on the portal as well as on portal subsites. There was no way to control where the banner would appear, nor was it possible to predict for whom the banner would show; thus, for all practical purposes, its appearance was random. The banner appeared 2,004,709 times in the 2-week period. During these 2 weeks, Passagen had 818,422 and 893,599 unique visitors per week respectively, and the total number of visits was approximately 2 million with approximately 14 million pages viewed. By "clicking" on the banner, 10,644 visitors were actually linked to the study site (Figure 1).

[FIGURE 1 OMITTED]

By clicking on the banner, the viewer was linked to an introduction site located on a server within the University of Gothenburg website. The introduction site had the University logo and described the project, the nature and number of the questions, the funding source, and material related to ethics and confidentiality, including the fact that the questionnaire was anonymous. The introduction site also informed participants that this survey was limited to people 18 or more years old. By clicking on an 'accept" button, the viewer was linked to the questionnaire, which was also placed on the University server. Below the questionnaire and visible at all times was a set of boxes numbered 1 to 75 and corresponding to each question (many questions contained more than one item; we measured 165 items total). Different colors indicated whether a question had been answered or not, and it was possible up to completion for respondents to return to a particular question to revise an answer. The system was running on an Intel-based 2 x 450 Mhz server placed within the University of Gothenburg website with a 10 GB connection both ways.

Each respondent opened a session with the server, and this session was active until the questionnaire was finished or the respondent quit. If the respondent quit, the session was automatically shut down. All responses and changes of responses were logged and saved continually. This format makes it possible to analyze missing values, when and where respondents drop out, and other variables which might be related to their discontinuing participation, such as gender and age.

Instrument

We based the questionnaire on two earlier instruments. The first was used in an earlier study done in conjunction with MSNBC, one of the largest American portals (Cooper, Scherer, & Mathy, 2001). The second was used in the Sex in Sweden Survey (Lewin, 1998). The instrument in this study consisted of 75 questions broken down into seven sections. (1) Section 1 had 24 demographic questions including items on Internet access and experience of relationships and sexuality. Section 2 consisted of 13 questions focusing on perceptions of on-line love and sexual behavior. Section 3 had 7 questions on on-line sexual activity in the workplace and relevant policies and regulations for such behavior. Section 4 consisted of 17 questions dealing with both on-line and off-line

sexual experiences. In Section 5, respondents were to answer 4 questions including 12 statements about the Internet and sexuality to help make clearer their attitudes about this phenomenon. Questions here asked, for example, if cybersex is cheating, if sex on the Internet is better suited for men, if the Internet fosters equality between genders, and similar questions. Items were rated on a 3-point Likert scale (1 =fully agree, 3 = do not agree). Section 6 had 8 questions around issues of online sexual problems and STDs. The last question contained a 10-item Kalichman scale (Kalichman, Johnson, Adair, & Rompa, 1994) on sexual compulsivity. The final item of the instrument provided respondents with an opportunity to comment on the questionnaire.

Due to the format and technique used in this survey, different ns were obtained for items throughout the questionnaire. The first question (age) was answered by 3,614 persons, and the last question (10-item Kalichman scale) by 1,851 persons. This provides a possibility for special analyses of those who did not complete the entire questionnaire. For example, half of the males had dropped out by item 23, whereas half of the females had dropped out by item 49.

Sample

Portal site. Of those who visit the Passagen portal site, 54% are men and 46% are women. Twelve percent of the visitors are between 18 and 24 years old. The two largest groups are ages 25 to 34 (22%) and 35 to 49 (27%). Eighteen percent are 50 to 64 years old and 4% are 65 or more years old. A total of 17% are under age 17. More than half of the portal population consists of the following groups: full-time students (17%), service occupation workers (13%), educational workers (10%), administrative workers (8%), and technical workers (7%). Ten percent do not currently have an occupation.

Study respondents. The sample participating in this survey came from the population of those who visited the Passagen portal site during the 2 weeks from June 10 to June 23, 2002. Because the survey was related to sexuality, which brought up ethical and legal complications for involving minors, it was decided to restrict participation to adults (over the age of 18). If a respondent filled out the questionnaire and claimed to be under 18, we removed that case from the database. An upper age limit was set at 75 years, both because of the small population claiming to be older and so that we could compare the results with previous research (Cooper, Morahan-Martin, Mathy, & Maheu, 2002; Lewin, 1998). Data on the sample characteristics are presented in Table 1. For those who reported engaging in online sexual activity (n = 2035), the mean age for females was 29.7 years (SD = 10.6) and for males 32.0 years (SD = 10.5).

The gender distribution of those who engaged in on-line sexual activity in this study was 55% males and 45% females, the same percentages found in overall use of the Internet in Sweden (Nielsen Netratings, 2002, January), and statistically almost identical to the percentages of those who visited the portal site where the questionnaire was launched (54% males and 46% females). Fewer than 3% claimed to be from a country outside Sweden, and almost all of these were from adjacent Scandinavian countries (Denmark, Norway, Finland).

Analysis

We analyzed data using SPSS Version 11.0. Respondents were considered to have dropped out if they did not complete the second and third-last items on the questionnaire (the last items on the Kalichman Sexual Compulsiveness Scale). Kaplan-Meier analysis indicated a significant difference between male and female dropout rates. Thus, we analyzed further data for males and females separately, comparing those who dropped out before completion of the questionnaire to those who completed it. Because of the relatively high dropout rate of men by the end of the basic demographic data (half of the dropouts had done so by the end of this section), we confined ourselves to analysis of these demographics. Data were subject to comparison using Chi-squared tests for categorical data (with Yates correction for discontinuity where appropriate) and t tests (unequal variance assumed where Levene's F for equality of variances was significant at p < .05) for linear data, with significance set at the .05 level. Finally, we used multinomial logistic regression with dummy variable coding to compare those who exited or were retained on the demographic and other variables that were statistically significant.

RESULTS

Table 1 displays demographic characteristics of the participants. The sample comprised roughly equal numbers of males and females, of whom about 40% were single, and the great majority were Swedish with most of the remainder being Swedish speakers in adjacent Scandinavian countries. About one third lived in one of the greater metropolitan areas in Sweden (Stockholm, Gothenburg, or Malm6). However, just one fifth grew up in one of those metropolitan areas. Just over one third lived alone, and just under one third lived with children. The sample was predominantly well-educated, with nearly 40% having some postsecondary education. About half were working, although nearly one quarter of the women were students. Nearly two thirds worked with computers with Internet access, and over 80% had Internet access at home and about half at work. Additional details on the sample are provided by Cooper et al. (2003) and Mansson et al. (2003).

Figure 2 shows that exit survival scores are curves of negative acceleration, with most dropout occurring by Item 25 for half the males who would drop out, compared to Item 49 for the women, reflecting almost double the retention number of items for women. Kaplan-Meier survival analysis indicated that this difference was significant (F = 30.4, df= 1, p < .000). Further, 50.6% of men had dropped out by the end of the questionnaire compared to 42.9% of women, again significant ([chi square] = 20.4, df = 1, p < .000).

[FIGURE 2 OMITTED]

The differences between the exiters and completers (summarized in Table 2) indicated that heterosexual men were significantly more likely to exit than bisexual or homosexual men. While the trend was similar for women, this was not significant. For men, those exiting were significantly more likely to be in a sexual relationship (married, cohabiting, partnership), whereas women exiters were significantly less likely to be in a relationship. For men, exiters were significantly more likely to have lower educational levels, with a gradient toward the better educated being less likely to exit, while for the women there was no significant relationship between education and exiting. A similar pattern occurred for geographical location, with men in the large metropolitan areas (Stockholm, Gothenburg, Malmo) being least likely to exit and those in a small town or rural area being more likely to exit. In contrast, there was again no significant association between exiting and location of residence for women.

For current living status, men who exited were significantly more likely to be living with others (spouse, parents, friends) rather than alone, and a similar significant pattern was observed for women. Those males living in a household without children were significantly less likely to exit, and while there was a similar pattern for women, it was not significant.

Those men who exited were significantly older by a mean of 2 years than those retained, whereas there was no significant age difference between the women who exited or were retained. Nor were there any significant differences between those who exited or were retained, male or female, on the mean number of hours per week they were active on the Internet or the mean number of hours per week they were on-line for sexual interests. However, the type of Internet connection was a significant variable in male, but not female, dropout: Men were significantly more likely to drop out if they had a slower connection (modem compared to DSL), whereas the connection speed had no relationship to dropout in women.

We also compared by gender those who either exited or were retained on two Internet sexuality questions, where the remaining n = 101 for males and 100 for females. There was no significant difference between the exiting and retained groups on whether the respondent had had cybersex (males, [chi square] = 0.03, df= 1, ns; females, [chi square] = 0, df = 1, ns) or met a partner with whom they had had sex in real life (males, [chi square] = 0, df= I, ns; females, [chi square] = 0.02, df = 1, ns).

On regression analysis, the equation was significant ([chi square] = 40.48, df = 11, p = .000), with three variables individually significant: type of Internet connection (broadband or modem; B = -.28, Exp(B) = .76, p = .0020), sexual orientation (homosexual/bisexual or heterosexual; B = .570, Exp(B) = 1.77, p = .001), and gender (B = .17, Exp(B) = 1.18, p = .05). Approaching significance was living arrangement (alone or with others; B = -.41, Exp(B) = .67, p =. 15).

DISCUSSION

These data are subject to several limitations, including the fact that the sample represents volunteers and that we have little information on their representativeness, although the sample characteristics appear to be close to those of the characteristics of general Passagen website users. Further, because of the high dropout rates in the early section of the questionnaire, especially among men, dropout characteristics on later items should be interpreted with the caution that they are simply comparing middle and late dropouts. Finally, we had no way of determining if any respondents had completed the questionnaire more than once.

The data show that dropout follows a curve of negative acceleration, with differences in survival rates between males and females (males dropping out significantly earlier and at a higher rate). The half-life for male dropout is at Item 25, whereas for females it is Item 49, reflecting almost a doubling of the retention time for females. Further, half of men and 40% of women dropped out. The shape of the curves in Figure 2 suggests that there is a relatively rapid rate of dropout that flattens over time, and that those remaining after about 25 items for men and about 50 for women tend to continue. This has implications for Internet questionnaire design, suggesting that sexuality questionnaires for males need to be shorter than for females to avoid significant dropout bias.

For males, self-identified heterosexuals were significantly more likely to exit, as were men currently in a relationship. For women, however, the relationship pattern was the reverse, with women in a relationship less likely to exit. The reversed relationship between men and women recurs with living with children, although since this is likely to be highly associated with being in a relationship, this is not surprising. For men, there was a significant gradation in the relationship between education and premature questionnaire exit, with the less educated more likely to drop out. While there was a similar trend in women, it did not reach significance. Geographically, again there was a similar pattern for men, who were significantly more likely to exit if they were in a smaller city or small town or rural area, whereas for women there was no apparent relationship between geography and dropout. These data suggest that without correcting for dropout, homosexual and bisexual men are likely to be overrepresented in Internet questionnaire samples, as are better educated and more urban men and those living alone. The different trends for relationship situation for men and women are intriguing and suggest that longer questionnaires (over about 30 items) will underrepresent men in relationships and overrepresent women in relationships. There was a significant difference in the age of men retained in the questionnaire, with dropouts being significantly older, but this relationship was not significant for women. Interestingly, there were no significant differences in Internet use in dropouts and those retained between either men or women, suggesting that the patterns of Internet use are not significant in determining the risk of dropout in terms of time per week active on the Internet or hours per week on-line for sexual purposes.

These data indicate that there is a high and early dropout rate for sexuality-related questionnaires, higher than that noted by Frick et al. (2001). This may be due to the nonsexual content of Frick et al.'s questionnaire and the lack of an incentive in the present study. While the present study was a volunteer questionnaire, the significantly higher dropout rates for men suggest that they may be unwilling to devote much time away from sexual pursuits on the Internet, compared to women. As in this study the participants could see the number of questions in the total questionnaire and the number they had completed, it may be that they made a decision based on time early in the questionnaire to drop out. The ability to assess progress in the questionnaire is unusual and may have contributed to dropout. This suggests that critical data questions should be placed as early as possible in a questionnaire. The technology employed in this study, which allows dropout point to be determined, may also be useful to assess dropout bias. Of particular interest is the fact that men but not women with slower Internet connections were significantly more likely to drop out. This, combined with the higher dropout rate for men, suggests that time and impatience are factors that are more important to male than to female respondents.

These data also suggest that estimates of the population of those who will complete an Internet sexuality questionnaire will vary depending on the length of the questionnaire. Given the progressive dropout rates in this study, simple univariate analyses that base demographic characteristics on initial demographics and subsequent analyses on those remaining are likely to progressively misrepresent the demographics and items appearing later in the study. The acknowledged bias toward urban and better educated respondents in Internet studies may also, to a limited extent, be an artifact of those who will continue with a sexuality questionnaire on the Internet rather than those who will initially commence it. A remaining question that these data do not directly answer is related to the motivations for dropout, although time and attention span are strongly implicated by the data on the connection speed being related to dropout in men.

It is likely, however, that the bias for Internet sexuality-related questions is not as large as might be expected, given the nonsignificant differences between exiters and those retained on the two sexuality questions relating to having cybersex and meeting partners on the Internet with whom the respondent had had sex in real life. A strong caveat, however, is that the questions on cybersex (Item 47) and meeting partners in real life (Item 51) occurred at the point where over half of the respondents who would drop out had already done so; thus, these questions are only a measure of potential differences between late dropouts and those who were retained.

Finally, it is likely that dropout rates may also be influenced by variables including culture, language, content, and, as already noted, incentives. It would seem unwise to generalize these data beyond incentive-free sexuality questionnaires in Western countries or to replicate these data without regard to the specific sexuality content areas of the questionnaire. However, it is clear from these data that dropout does constitute a significant bias in Internet sexuality research and that it varies significantly by gender, sexual orientation, Internet connection type, relationship status, and other demographic characteristics. Dropout biases therefore need to be considered in interpreting the results of Internet sexuality questionnaires, particularly given the high rates of dropout before questionnaire completion found in this study.

Table 1. Sample Characteristics (%)

                                     Males               Females
Variable                           (n = 1846)           (n = 1637)

Age (Mean [+ or -] SD)         32.5 [+ or -] 11.3   29.2 [+ or -] 10.5
Marital status
  Single                              39.2                 36.0
  Married                             20.0                 11.8
  Cohabiting                          22.7                 28.0
  Divorced/widowed                     5.8                  8.5
  Live apart                           9.7                 13.4
Nationality
  Swedish                             91.7                 91.4
  Other Scandinavian                   2.4                  2.8
  Other European                       1.3                  1.9
  Non-European                         1.1                  0.7
Residence
  Large city                          34.9                 29.3
  City                                40.7                 43.9
  Town or rural                       17.7                 21.1
  Other Scandinavian place             1.1                  1.4
Grew up in
  Large city                          22.4                 20.8
  City                                35.1                 34.9
  Town or rural                       30.4                 34.8
  Capital or large city
    outside                            2.5                  2.2
  Town or rural area outside           1.8                  2.4
Living arrangement
  Alone                               38.1                 35.6
  Parents                             12.8                 18.0
  Spouse/partner                      40.1                 37.1
  Friend(s)                            2.8                  5.2
Living with children
  Yes                                 33.7                 31.6
  No                                  59.9                 64.5
Highest education
  Elementary school                    6.5                  7.7
  High school                         41.0                 43.0
  University                          38.4                 39.5
Occupation
  Working                             58.8                 45.9
  Student                             14.1                 23.8
  Unemployed                           5.7                  8.0
  Retired                              1.7                  1.1
  Homemaker, parental leave            0.5                  3.3
  Sick leave                           1.9                  3.8
  Other                                4.8                  6.1
Work with computers
  With Internet access                64.9                 62.9
  Without access                       5.6                  7.2
Internet access
  At home                             80.0                 84.1
  At work                             54.2                 42.6
  At school                           15.4                 25.4
  At Internet cafes                   5.1                  4.6
  At other places                     3.5                  6.4

Variable                                 t or [chi square]

Age (Mean [+ or -] SD)         t = 9.1, df = 3445, p < .000
Marital status
  Single
  Married
  Cohabiting
  Divorced/widowed
  Live apart                   [chi square] = 67.9, df = 4, p < .000
Nationality
  Swedish
  Other Scandinavian
  Other European
  Non-European                 [chi square] = 9.5, df = 3, p < .03
Residence
  Large city
  City
  Town or rural
  Other Scandinavian place     [chi square] = 17.6, df = 4, p < .002
Grew up in
  Large city
  City
  Town or rural
  Capital or large city
    outside
  Town or rural area outside   [chi square] = 7.0, df = 4, p < .13
Living arrangement
  Alone
  Parents
  Spouse/partner
  Friend(s)                    [chi square] = 30.6, df = 3, p < .000
Living with children
  Yes
  No                           [chi square] = 3.6, df = 1, p < .06
Highest education
  Elementary school
  High school
  University                   [chi square] = 1.39, df = 3, p < .71
Occupation
  Working
  Student
  Unemployed
  Retired
  Homemaker, parental leave
  Sick leave
  Other                        [chi square] = 135.7, df = 6, p < .000
Work with computers
  With Internet access         [chi square] = 1.6, df = 1, p < .02
  Without access               [chi square] = 3.9, df = 1, p < .05
Internet access
  At home                      [chi square] = 9.8, df = 1, p < .002
  At work                      [chi square] = 34.7, df = 1, p < .000
  At school                    [chi square] = 54.3, df = 1, p < .000
  At Internet cafes            [chi square] = 0.5, df = 1, p < .48
  At other places              [chi square] = 15.6, df = 1, p < .000

Note. Where responses total < 100%, the remainder are missing responses.

Table 2. Differences Between Exiters and Retained
Respondents by Gender

                                      Male              Female

                                Exit    Retained   Exit    Retained

Variable                                   Percentages

Sexual identity
  Heterosexual                  38.4    61.6       35.0    65.0
  Homosexual                    22.7    77.3 **    26.6    73.4
Relationship situation
  Not in one                    46.1    53.9       44.3    55.7
  In one                        53.6    46.4 **    38.3    61.7 *
Education
  High school                   53.8    46.2       40.8    59.2
  Some college                  47.8    52.2       42.7    57.3
  Graduate                      40.3    59.7 **    36.4    63.6
Geography
  Large city                    43.8    56.2       38.7    61.3
  Small city                    48.3    51.7       42.6    57.4
  Town or rural                 55.0    45.0       42.1    57.9
  Overseas                      56.3    43.8 **    34.8    65.2
Living companions
  Alone                         43.0    57.0       39.8    60.2
  Parents or friends            49.8    50.2       53.7    46.3
  Partner or spouse             51.1    48.9 **    36.1    63.9 **
Living with children
  Yes                           50.8    49.2       39.1    60.9
  No                            45.2    54.8 *     41.5    58.5
Type of Internet connection
  DSL                           36.8    62.3       35.0    65.0
  Modem                         45.5    54.5 **    38.3    61.7

                                              Means

Age                             34.4    32.3 **    28.9    29.9
Hours/week active on Internet   16.0    17.3       14.3    15.3
Hours/week online for sex        6.2     5.2        4.5     4.6

* p < 0.05. ** p < 0.01.

REFERENCES

Bull, S. S., & McFarlane, M. M. (2000). Soliciting sex on the Internet: What are the risks for sexually transmitted diseases and HIV? Sexually Transmitted Diseases, 27, 545-550.

Cooper, A., Morahan-Martin, J., Mathy, R., & Maheu, M. (2002). Toward an increased understanding of user demographics in online sexual activities. Journal of Sex and Marital Therapy, 28, 105-129.

Cooper, A., Scherer, C., Boies, S., & Gordon, B. (1999). Sexuality and the Internet: From sexual exploration to pathological expression. Professional Psychology, 30, 154-164.

Cooper, A., Scherer, C., & Mathy, R. (2001). Overcoming methodological concerns in the investigation of online sexual activities. Cyberpsychology and Behavior, 4, 437-448.

Frick, A., Bachtiger, M. T., & Reips, U. D. (2001). Financial incentives, personal information and drop-out rate in on-line studies. In U. D. Reips & M. Bosnjak (Eds.), Dimensions of internet science (pp. 209-222). Lengerich, Germany: Pabst Publishers.

Kalichman, S. C., Johnson, J. R., Adair, V., & Rompa, D. (1994). Sexual sensation seeking: Scale development and predicting AIDS-risk behavior among homosexually active men. Journal of Personality Assessment, 62, 385-397.

Klausner, J. D., Wolf, W., Fuscher-Ponce, L., Zolt, I., & Katz, M. H. (2000). Tracing a syphilis outbreak through cyberspace. JAMA, 284, 447-449.

Lewin, B. (Ed.). (1998). Sex in Sweden: On sexual life in Sweden 1996. Stockholm, Sweden: Folkhalsoinstitutet.

Mansson, S.-A., Daneback, K., Tikkanen, R., & Lofgren-Martenson, L. (2003). Karlek och sex pa internet [Love and sex on the Internet]. Malmo & Goteborg: Natsexprojektet Rapport.

McFarlane, M., Bull, S. S., & Rietmeijer, C. A. (2002). Young adults on the Internet: Risk behaviors for sexually transmitted diseases and HIV. Journal of Adolescent Health, 31, 11-16.

Mustanski, B. (2001). Getting wired: Exploiting the Internet for the collection of valid sexuality data. The Journal of Sex Research, 38, 292-301.

Nielsen Netratings. (2002, January). Antalet svenska internet surfare okade med 23% under ett ar [The number of Swedish Internet surfers increased by 23% in a year]. Retrieved November 6. 2002, from http://www.nielsen-netratings.com/pr/pr_020212_sweden.pdf

Ochs, E. P., Mah, K., & Binik, Y. M. (2002). Obtaining data about human sexual functioning from the Internet. In A. Cooper (Ed.), Sex and the Internet: A guidebook for clinicians (pp. 245-262). New York: Brunner-Routledge.

Ross, M. W., Tikkanen, R., & Mansson, S. A. (2000). Differences between Internet samples and conventional samples of men who have sex with men: Implications for research and HIV interventions. Social Science and Medicine, 51,749-758.

Wallace, P. (1999). The psychology of the Internet. Cambridge, England: Cambridge University Press.

(1) The complete questionnaire can be obtained from Professor Sven-Axel Mansson at University of Gothenburg. P.O. Box 720, 405 30 Goteborg, Sweden.

This study was funded by a grant from the Swedish Social Science and Work Life Research Council.

Address correspondence to Dr. Michael Ross, WHO Center for Health Promotion and Prevention Research, School of Public Health, University of Texas. PO Box 20036, Houston, TX 77225; e-mail: [e-mail].

Michael W. Ross

University of Texas

Kristian Daneback, Sven-Axel Mansson, and Ronny Tikkanen

University of Gothenburg, Sweden

Al Cooper

Stanford University, California

Manuscript accepted August 13, 2003