Social Science & Medicine 51 (2000) 749±758
www.elsevier.com/locate/socscimed
Dierences between Internet samples and conventional samples of men who have sex with men: implications for p research and HIV interventions Michael W. Ross a,*, Ronny Tikkanen b, Sven-Axel MaÊnsson b a
WHO Center for Health Promotion Research and Development, School of Public Health, University of Texas, P.O. Box 20186, TX 77225, Houston, USA b School of Social Work, GoÈteborg University, PO Box 720, SE 405 30 GoÈteborg, Sweden
Abstract The Internet is becoming a new erotic oasis for obtaining sex online or in person. We reviewed the literature on cybersex and compared dierences in data from samples of homosexually active men obtained on identical questionnaires from a conventional written questionnaire, distributed through the mailing and contact lists of a large national gay organization in Sweden, and through the same organization's website and chat room. A total of 716 written questionnaires and 678 Internet questionnaires were obtained. The Internet sample was younger, more likely to live in small towns or cities, live with parents or a girlfriend, and have lower formal education. They are less likely to have previous sexual experience solely with other men (one in three of the Internet sample vs. 1 in 14 of the written sample de®ned themselves as bisexual) and more likely to visit erotic oases such as bathhouses, video clubs and erotic movie houses. They also visited Internet chat rooms more frequently (86% of the Internet sample vs. 50% of the written sample). One third of the Internet sample wanted the opportunity to talk with an expert about HIV compared with a quarter of the written sample. Sexual practices between the two samples were generally similar, although the Internet sample reported signi®cantly less body contact, kissing, hugging, mutual masturbation, and more condom use for anal intercourse with steady partners. Over four times as many of the Internet samples reported sex with women in the past year as the written sample. These data indicate that Internet data collection is feasible and that this mode of data collection, despite the nonrandom and self-selected nature of both types of samples, is likely to be more signi®cantly oriented toward the young, geographically more isolated, and more behaviorally and self-identi®ed bisexual respondent than conventionally distributed written questionnaires. 7 2000 Elsevier Science Ltd. All rights reserved. Keywords: Male homosexuality; Internet; Survey research; HIV; Sweden
This study was funded by the Swedish National Institute of Public Health (FHI) and the Swedish Council for Social Research (SFR). * Corresponding author. Tel.: +1-713-500-9652; fax.: +1713-500-9602. E-mail address:
[email protected] (M.W. Ross). p
Introduction The utility and appropriateness of collecting data using the Internet has been approached by several researchers. However, it raises a number of questions when the data concern sexual attitudes and behaviors. This paper reviews some of the relevant questions and
0277-9536/00/$ - see front matter 7 2000 Elsevier Science Ltd. All rights reserved. PII: S 0 2 7 7 - 9 5 3 6 ( 9 9 ) 0 0 4 9 3 - 1
750
M.W. Ross et al. / Social Science & Medicine 51 (2000) 749±758
seeks to determine whether the mode of data collection (paper and pencil, computer, Internet) may produce any systematic bias into sampling and data variability, and if so, what direction these biases may be in. Such data have implications not only for virtual or actual sexual contact respectively on and via the Internet, but also potentially for services using this medium. The Internet refers to the system of interconnected computers creating a matrix of information ¯ow that provides the backbone for E-mail (electronic mail), WWW (World Wide Web), FTP (®le transfer programs), Usenet (special interest groups), IRC (Internet relay chat rooms), and similar services (DeGuzman & Ross, 1999). The Internet has served as a mode of data collection in a number of contexts previously. Stanton (1998) used identical questionnaire items to gather data from two employee samples from commercial ®rms, nonpro®t organizations, university and government oces. The WWW sample comprised 50 individuals, the paper version comprised 181 individuals. Stanton found that in his research (on supervisory fairness) it was not possible to ascertain the response rate from the WWW survey since it was based on individuals who were contacted by E-mail and requested to ®ll out the on-line survey, and to pass on the request to colleagues. The paper and pencil survey had a response rate of 50%; the WWW survey was posted on the web server of a large US public university. Comparison of the two modes indicated that the Internet data had signi®cantly fewer missing data points, and that on yes/no questions, the Internet data had greater variability (using higher variability as a gross measure of higher data quality). On the six-point Likert scales the trend was in the same direction but not statistically signi®cant. Examining the factor structure of the scales, Stanton found psychometric equivalence (factor invariance) between the two samples. These data indicated that the quality and usefulness of the data from the Internet were good, in fact, the level of missing data was lower in the Internet sample although the possibility of slightly lower reliability was noted. While Stanton noted that demographically, the two samples were highly comparable on male:female ratio, time in present job, age, and full-time employment, the Internet respondents worked signi®cantly more hours per week. Stanton (1998) notes that where sampling need to be controlled, for example where a controlled environment is necessary in their administration, Internet surveys are at a disadvantage. Where the purpose of a survey is to pinpoint areas of concern, identifying preferences for a set of alternatives in a company, developing or piloting instruments, then the Internet may be appropriate. The question of measurement equivalence between Internet and pencil and paper question-
naires has also been addressed by King and Miles (1995), who found that there was no evidence that administration mode for cognitive tests had any eect on measurement equivalence. There are also signi®cant advantages for computer-based questionnaires. Turner, Ku, Rogers, Lindberg, Pleck and Sonerstein (1998) found that for sensitive data such as on male±male sexual behavior and injecting drug use, prevalence was higher by a factor of three for audio-computer-assisted forms where data were typed directly into the machine compared with written (conventional pencil and paper) forms. There is also the advantage that data transcription errors are avoided and out of range values may be avoided by not being accepted and ¯agged. Binik, Binik, Mah and Kiesler (1999) suggest that Internet surveys are less likely to be distorted in the direction of social desirability. Sexual contact on the Internet The Internet has become a sexual marketplace which in some ways is the equivalent of the gay bar or the cruising area. As an entry to an `erotic oasis', as well as a virtual reality `erotic oasis' itself, it contains many features of traditional areas for homosocial and homosexual contact while adding some new ones. Wysocki (1998) studied heterosexual chat rooms in the United States and found that the most common reasons for having sex online included the need for anonymity, time constraints on participants' personal lives, having the ability to share sexual fantasies with other people, to participate in online sexual activity, and to meet other people with similar sexual interests. Her sample had an average age of 35, and spent a mean of 1.6 h per day, communicating with a mean of 13.3 h per week. Over 57% reported that they had already met someone face to face with whom they became involved as a result of meeting through the chat room. We are unaware of any equivalent data on men who have sex with men (MSMs), but anecdotal evidence suggests that gay chatrooms may be even more oriented toward locating partners for physical sex. Weinrich (1997) comments that gay personal-ad contacts can take place on the Internet freed from many of the problems of disproportionate romantic supplyand-demand that plague many corresponding Internet sites for heterosexuals. It is important to describe the novel features that the Internet oers in addition to the facility to extend old functions. While the Internet may provide opportunities for a much greater range (and time frame) for meeting other homosexual men, it can in addition provide the possibility of `virtual sex' as well as an opportunity for disguising or changing sexual characteristics. For the person who may be geographically or temporally isolated from others, the Internet provides a potentially huge set of possible
M.W. Ross et al. / Social Science & Medicine 51 (2000) 749±758
responders. Further, given the possibility of `cybersex' (erotic interactions through cyber discourse: Blair, 1998) which is both safe from the transmission of sexually transmissible pathogens and may be terminated at any point by either party, it oers the opportunity of anonymous, commitment-free interaction in much the same way as public sex environments may oer the latter. However, it also oers the possibility of disguising or selecting sexual characteristics in the way only a totally dark `backroom' sex environment might approach. Even so, a `backroom' sex environment still oers tactile cues to physical characteristics. The degree of discrepancy on the Internet, assuming that there is no subsequent physical contact, is almost limitless: there have been reports of gay men masquerading as women on heterosexual sexual chat rooms (Ford, 1998). Blair (1998, p. 209) notes that issues of power equality on the Internet are enhanced because the system is based on discourse. Whereas the traditional gay erotic oasis will usually have a power structure based on appearance and thus desirability, the anonymity of cyberspace reduces this (although it could be argued that cyberspace is not a leveler, insofar as the person who is the most pro®cient verbally will have an advantage). Blair also notes that absent from this discourse is the physical self, only words on a screen and a moving cursor that mark the progress of the exchange. The person in this medium is `free to initiate or refuse the exchange, manipulate the ¯ow of discourse, or passively enjoy the ¯ow'. However, Blair (1998, pp. 210±211) also makes the point that `net sex' does have its sensory limitations. The sexual discourse of the net may be combined with autostimulation or expand to an oine physical relationship. As she observes, oine physical coupling has in many cases been enhanced by online exchanges, and many couples have met and assumed real-time relationships as a result of net-sex encounters. There are three Internet possibilities: embellishment of real-world circumstances, creation of a pure fantasy scenario, and `computer sex', where one party describes online what he would like the other party to do and may achieve orgasm. Recent advances in computing also make it possible for parties to download still pictures of themselves (or, if a physical meeting is not a possibility, pictures of their assumed computer persona) which may cover the range of revelation from the conventional to the highly explicit. Still more recently, the coupling of the computer with video technology makes possible real-time video interaction. In the foreseeable future, Dery (1994) argues, virtual reality technology will expand the possibilities substantially into real-time hyperreality. From the perspective of erotic oases, Henriksson (1995) has noted that public conveniences, which
751
were the major venue for men having sex with other men, declined in number from the middle of the century, and appeared to be replaced with erotic movie houses, video clubs and saunas (bathhouses). As technology has changed, so have venues and avenues which facilitate sexual contact between men. The Internet may simply be the latest opportunity to appear. However, as public venues may be subject to legal restrictions and patrons may run the risk of arrest or exposure, virtual environments for the present may oer the added protection of greater anonymity, lack of illegality, and protection from stigma. They may also oer a psychological defense Ð that as a sexual act is `virtual', then the act is not `really' homosexual or unfaithful (Wysocki, 1998). Internet sexuality (and by extension, Internet interventions) are an example of the diusion of a technological innovation (Rogers, 1995). Its characteristics, including its compatibility, ¯exibility, reversibility, relative advantage, low complexity, costeciency and low risk, all make it an ideal innovation for the sexual arena. These characteristics, along with the eectiveness of the Internet for creating a computer-supported social network which facilitates the types of social exchange that allow people to create community (Haythornthwaite, Wellman & Garton 1998) make it an ideal medium for a sexual minority. Where a minority and its culture is characterized primarily by sexual interaction and by stigma (Ross, FernaÂndez-Esquer & Seibt, 1995), the possibility of enhancing sexual contact and yet maintaining a degree of secrecy makes the Internet a particularly eective medium for enhancing such a subculture. This is illustrated by the massive growth of `permissive' computer-mediated communication (the characteristics of the system which oer the greatest ¯exibility for new and unexpected uses: Galegher & Kraut, 1990) in the area of sexually-related IRCs. Shaw (1997, p. 136) observed that the ten most popular chat rooms on America Online, the largest subscriber service, were sexual, of which three were gay. He suggests that it is an ```undernet' of self-discovery and shrouded experimentation which both resonates with and parallels the lived experiences of many gay men''. While the concept of Internet addiction has been raised, Grith (1998) questions whether much of the excessive use of the Internet is actually an addiction as classically de®ned. Nevertheless, Shaw (1997) notes that it is not unusual for his gay respondents to spend an average of 10±15 h per week online. Some of the more compulsive aspects of using gay sex channels may provide a further possible resonation between cybercruising and the gay subculture.
752
M.W. Ross et al. / Social Science & Medicine 51 (2000) 749±758
Public health and disease prevention From the perspective of public health, cybersex alone is totally safe, excluding the possibility of transmission of computer viruses which may aect the health of the participants' hardware and software but not their persons. However, the possibility exists of a transition from cybersex to physical sex. A recent lay description of experiences in a gay `chatroom' has characterized participants as falling into the categories of ``desperately needy, terminally horny, criminally insane, hiding from wife/girlfriend, and all of the above'' (Mannella, 1999, p. 32). This characterization of users as seeking sex may be exaggerated, but the suggestion that almost all local users are seeking sexual contact may not be. Recently, Nieves (1999) reported a cluster of seven syphilis cases (of the 17 reported for 1999 in San Francisco) in homosexually active men who met their partners through an Internet chat room. These seven men in turn identi®ed 99 partners, the great majority met through IRCs. It is clear that from the public health perspective, the Internet can be a signi®cant source of sexual partners and consequently disease spread. It is possible, we argue, that sexual interactions which by their nature on the Internet are unsafe may in person be unsafe. Where individuals have set up and consummated on-line a scenario of acts while interacting on the Internet which would be unsafe in person, a sexual script has been written. We use the concept of sexual script here as de®ned by Gagnon and Simon (1973) as a repertoire of acts and statuses that are recognized by social groups. This follows from Goman's (1959) view of self presentation as being analagous to playing a role in a play. Such a script, played out in person, may be unsafe; and, having been rehearsed on-line, may be more resistant to ad lib or change. Whether such `risk-priming' scenarios can be empirically demonstrated is unclear, but anecdotal reports suggest that they exist. For this reason alone, apart from reasons of sampling methods, homosexual contact on and subsequently o the Internet is an important public health issue. A second issue relates to the Internet as a source of information and particularly to populations who are unlikely to have access to the gay subculture and to be acculturated into it. Recently, there has been some debate concerning the role of gay subcultures in acculturation into safer sex norms (Ross et al., 1995; Seibt et al., 1995) which suggests that, as in the case of ethnicity in a new culture, acculturation can signi®cantly impact health behaviors. A case could be made that those in the virtual sexual marketplace may be less acculturated to the gay subculture. Alternatively, the case could also be made that a virtual equivalent of a real gay subculture exists, which allows participants to
access and observe gay interactions in a stress-free environment, at their own pace, and without pressure: desensitization rather than ¯ooding. In the case of a `virtual culture', there is a question of the relationship between the `real' and the `virtual'. What is the degree of overlap between gay subcultures and virtual cultures? This not only has implications for the diusion of health-related messages: it also raises questions about the existence of an equivalent of the MSMs described by Humphreys (1970) in his study of men who had sexual contact in public conveniences, who had sex with other men because it was less lonely than masturbation. Similar to such places, cybersex may provide opportunities for men to have virtual sex with men which they may have even less reason to consider homosexual, because there is no physical contact. Cybersex oers a middle place between fantasy and physical contact similar to that played by video erotica Ð as a means of arousal without the physical contact with another. It would be a mistake to characterize the Internet sexual subculture as monolithic Ð Rheingold (1993) comments that it is more an ecosystem of subcultures. Here, we intend to focus on the MSM sexual chatrooms, while recognizing that there is major variation even within this category. Computer mediated relating (Cooper & Sportolari, 1997) is assumed to be less involving, less rich, and less personal than face to face communication. It may also facilitate anonymous, compulsive contacts. Recent research has con®rmed the potential of the Internet for disruption or for the facilitation of dysfunction. Brenner (1997) reported on an Internet questionnaire study (not related to sexual contact) where responses came mostly from men in their mid-thirties who reported using the Internet for an average of 19 h per week. Some respondents reported this interfered with work and created social isolation. In a powerful longitudinal study, Kraut, Patterson, Lundmark and Kiesler (1998) reported that greater use of the Internet leads to shrinking social support and happiness, and increases depression and loneliness. Kraut and colleagues were reporting on individuals in diverse households in Pittsburgh over their ®rst two years online, and not on those who might be considered to be `addicted' to the Internet. One speculation for this ®nding is that people may be substituting weak online friendships for stronger real life friendships. As a search for sexual contacts (`cruising'), however, the Internet may be particularly useful: as Weinrich (1997) has observed, it allows a very large number of potential partners to be accessed and screened in a relatively short time. Cooper and Sportolari (1997) note that initial impressions are based on how one describes oneself, and that online, physical presence is not evident unless users choose to textually describe
M.W. Ross et al. / Social Science & Medicine 51 (2000) 749±758
themselves. They can make decisions about when and whether to disclose negative information and how to frame potentially positive information. Usually, in the MSM context, information is limited to height, weight, hair and eye color, and genital size and related characteristics as well as preference for particular sexual acts. Thus, like a conventional cruising area (Humphreys, 1970; Henriksson, 1995) information is conveyed by the textual equivalent of gestures Ð in recognition of the fact that in face to face situations, people make quick judgements based on physical attributes. The Internet is thus very appropriate for sexual contact where appeal is based on a few characteristics to the exclusion of other interactions. On the other hand, Cooper and Sportolari discuss relationships which are able to be extended oine through telephone contact and face to face meetings, with online contact acting to enhance self-integration, eroticization, autonomy and intimacy. But it is the opportunity for physical contact which places the Internet midway between video erotica and the public sex environments described by Humphreys (1970) and Henriksson (1995). However, unlike public sex environments, cybersex is more analagous to phone sex since there is not the immediate opportunity of physical consummation with another person. Nor is there the `membrane' of the sexual versus the outside, asexual world to traverse, as Henriksson has commented. Thus, questions about MSM on the Internet raise issues about the nature of the participants in terms of identity and gay acculturation, in terms of the nature of risk and sexual scripting when cybersex extends into physical sex, and in terms of the information available to reduce risk. The Internet may also oer positive opportunities for diusion of information (Rogers, 1995) for those who are gay-identi®ed by part of the diaspora Ð distant in time (i.e., across time zones) and place from the gay satellite cultures where access to services and information are more available. DeGuzman and Ross (1999) have described, in detail, the possibilities for provision of HIV/AIDS information and counseling using the Internet and concluded that it is not only possible but that it would have advantages not available to conventional counseling and information services, including anonymity and availability at times and places not served by existing services. Thus, a second question in addition to the question as to what biases there may be in mode of data collection between Internet and written (paper and pencil) tests or computer-mediated tests, relates to whether Internet-sampled populations are less acculturated and more marginalized from gay satellite cultures or the heterosexual milieu. Binik et al. (1999) have commented that the bias with recruiting subjects from the Internet for sexually-
753
related studies will be dierent from that associated with traditional recruitment methods. They suggested that internet respondents are likely to be younger, richer, and better educated and more open than they might otherwise be. This study sought to characterize the demographic, sexual behavior and identity dierences between self-selected respondents to a conventional paper and pencil questionnaire and an Internetbased questionnaire. Respondents were accessed either through the RFSL (RiksfoÈrbundet foÈr sexuellt likaberaÈttigande Ð Swedish federation for lesbian and gay rights, the largest and oldest organization for gay people in Sweden) membership list (distributed written questionnaires) or the RFSL website. Methods Internet survey The Internet and especially RFSL's WWW-page is an important channel for men who have sex with men in Sweden. The RFSL's WWW-page (http:// www.rfsl.se) consists of news on sexual politics, information on HIV prevention and personal advertisements. Moreover, there is a chat room where men who have sex with men can communicate with each other on-line. The chat room has about 6000 visitors every day and over 1000 personal advertisements are posted on the WWW-page every week. It is not restricted to RFSL members. We constructed a homepage that is a part of the RFSL's WWW-page (http://www.rfsl.se/ annons/sexliv/index.html). The survey questionnaire was launched on RFSL's homepage in mid April 1998 and the responses were collected (downloaded) continuously until the end of September. The questionnaire could be answered directly on the homepage and the responses were anonymized using specially designed software. In total, we received 678 questionnaires, of which 40 were put aside as being non-relevant or `problematic'. Half of these 40 came from self-identi®ed heterosexual men who clearly reported that they did not have any interest or experience of sex with other men. The other half (20) were either incomplete or impaired by obvious inconsistencies. These were excluded from the ®rst analysis of the data presented here. However, only three questionnaires (0.4%) from the Internet sample have been excluded bearing signs of intentional deception or deviousness. This is a surprisingly positive result, which opens up interesting possibilities for future research and methodological development. Written survey The written survey was distributed using two dier-
754
M.W. Ross et al. / Social Science & Medicine 51 (2000) 749±758
ent channels. First, it was sent out to members of three dierent local branches of RFSL (Stockholm, GoÈteborg, UmeaÊ). Around 1,500 questionnaires were distributed this way. The other channel was through 43 key-persons in the gay community in dierent parts of Sweden who voluntarily gave or sent the questionnaire to friends, acquaintances or other people in their social network. Attached to the questionnaire was written information about the possibility of answering through Internet. In total, we received 716 completed questionnaires in the written format. This study was approved by the relevant committee for the protection of human subjects. Analysis Data were analyzed using SPSSX by comparing the Internet sample (n = 678) and the written questionnaire sample (n = 716) on chi-square tests (Pearson chi-square with Yates correction for discontinuity as appropriate for ordinal scale data). Interval scale data were compared using independent sample t-tests (assuming unequal variances where Levene's test for equality of variances was signi®cant at p R 0.05). In cases where the distributions were signi®cantly dierent from normal on the Kolmogorov±Smirnov one-sample test, data were log-normalized. Signi®cance was set at the 5% level for all tests. Results Results of the comparisons between the written questionnaire sample and the internet sample are presented in Tables 1 and 2. Several important demographic dierences consistently stand out. First, Internet respondents are younger, more likely to live in smaller towns or cities, live with their parents or a girlfriend, and have lower formal education. Second, they are signi®cantly less likely to have previous sexual experience solely with other men (half of the Internet sample versus two thirds of the paper and pencil sample), and to have more sexual experience with women. Consistent with this, they are also much more likely to identify themselves as bisexual (one in three of the Internet sample compared with one in 14 of the paper and pencil questionnaire sample). Notable, however, is the lack of signi®cant dierence in their current sexual situation. With regard to gay acculturation, the Internet sample have been members of fewer gay organizations, but visit more sex environments such as saunas (bathhouses), video clubs, and erotic movie houses. Not surprisingly, they have also visited Internet chat rooms (Internet sex sites) much more frequently than the paper and pencil questionnaire sample: nevertheless, it is noteworthy that the pencil
and paper questionnaire sample also report extensive Internet experience. Half of the written questionnaire sample (49.7%) reported using chat rooms, compared with 86.3% of the Internet sample. One in three of the Internet sample, compared with a quarter of the written questionnaire sample, wanted the opportunity to talk with an expert about HIV, and one in four of both samples worried about being infected with HIV in the past year. Only a small proportion of each sample was aware that they were HIV infected. Reported sexual practices were generally similar between the two samples, although there were more of the Internet sample reporting receptive anal intercourse, and condom use for anal sex with steady (but not casual) partners. The Internet sample reported less analingus, and signi®cantly less mutual masturbation, body contact, and kissing and hugging. Where female sexual partners were reported (Internet sample, n = 175; Written sample, n = 40) there were signi®cant dierences in numbers, with the Internet group reporting more partners (and with greater numbers reporting female partners). Discussion These data oer some signi®cant insights into the biases associated with sampling modes for men who have sex with men. However, it must be emphasized that this is not `sampling' as usually methodologically de®ned, which refers to taking a subsample from a known population. In this case, the characteristics of the larger population are unknown and thus the biases of self-selection can only be speculated upon. Thus, this study more accurately reports on modal (Internet mode versus written mode) bias in self-selected samples than on sampling from populations which are knowable only with considerable diculty and expense even if studies were ethically acceptable. Practically, with the exception of random studies on large populations (Laumann et al., 1994), almost all studies on sexual behavior in MSMs are based on self-selected samples or clinical populations. Even random selection studies are biased by non-response and thus of limited generalizability (Ross, 1988; MaÊnsson, 1998). These data suggest, ®rst, that it is both possible and practical to obtain large samples of MSMs by using Internet questionnaires. The advantages for obtaining sexual and other sensitive data described by Turner et al. (1998) may be obtained while allowing for con®dentiality and a high degree of anonymity. These data do indicate a consistent set of biases for Internet data which may distinguish Internet samples from written (pencil and paper) questionnaires. However, such conventional questionnaires themselves are subject to bias and thus it is not possible to relate either of such self-
M.W. Ross et al. / Social Science & Medicine 51 (2000) 749±758
755
Table 1 Demographics Variable
Interneta
Writtena
p
1 steady partner Only casual partners Openness of sexuality (In/out of closet) Very open Quite open Quite closed Very closed Current steady female sex partner Current casual female sex partner Current steady male sex partner Perceived HIV status Positive Negative Ever had HIV test (yes) Mean number HIV tests Gay clubs Number of visits past 12 months Member gay organizations Sex clubs (video, sauna, sex-movie clubs) Number of visits past 12 months Internet chat rooms Number of visits past 12 months Ever felt need to talk with expert about HIV Worried might have been infected with HIV past 12 months Would use anonymous HIV home test
29.8 + 9.9 18.2 + 6.0 21.7 + 6.3
39.8 + 12.3 18.8 + 6.5 24.0 + 7.4
0.0001 0.06 0.0001
47.8 35.1 17.1
87.0 8.5 4.5
0.000001
49.0
38.2
0.001
49.1 26.5 7.4 15.4 1.5
65.9 28.5 2.9 2.1 0.6
0.000001
62.7 33.0 2.1 2.2
91.8 6.8 0.1 1.3
0.000001
51.6 14.6 19.3 9.0 5.4
59.5 2.8 31.0 1.3 5.5
0.000001
13.4 30.1 18.4 5.9 32.3
14.2 32.6 18.6 4.9 29.6
0.69
21.8 36.3 20.2 20.2 15.5 8.5 40.5
35.0 50.4 11.1 2.8 2.3 2.1 52.0
0.000001 0.000001 0.000001 0.00002
2.0 90.0 62.6 4.3 + 5.5 62.1 35.7 + 44.9 45.1 32.6 16.2 + 26.3 86.3 149.4 + 134.1 24.1 29.5 62.9
4.2 87.7 85.3 4.8 + 9.3 88.9 32.3+36.1 95.1 45.6 12.8 + 16.1 49.7 99.8 + 105.2 29.1 24.3 49.7
0.08 0.000001 0.31 0.000001 0.20 0.000001 0.000001 0.09 0.000001 0.0001 0.05 0.09 0.000001
a
Mean + SD or % as appropriate.
756
M.W. Ross et al. / Social Science & Medicine 51 (2000) 749±758
Table 2 Sexual practices Variable
Interneta
Writtena
p
