An empirical study of ordinal condom use measures.

Each year, approximately 12 million Americans acquire a sexually transmitted infection (STI; Institute of Medicine, 1997). STI transmission rates could be substantially reduced through increased use of male condoms (Cates, 2001; Holmes, Levine, & Weaver, 2004; Warner et al., 2004). Therefore, it is not surprising that many STI-prevention interventions focus on increasing participants' condom use (Chesson & Gift, 2000; Sheeran, Abraham, & Orbell, 1999), or that efforts to evaluate the effectiveness of these interventions frequently include measures to assess changes in condom use (Catania, Binson, Van Der Straten, & Stone, 1995; Mullen, Ramirez, Strouse, Hedges, & Sogolow, 2002; Stephenson, Imrie, & Sutton, 2000). Because condom use, like other sexual behaviors, is private and cannot ethically or practically be directly observed, these measures rely on participants' ability to self-report their condom use accurately (Crosby, 1998; Holtgrave & Pinkerton, 2000).

One popular method for assessing self-reported condom use asks respondents to classify their recent condom use along a 5- or 7-point continuum that ranges from, for example, "never" to "always," with intermediate levels of condom use labeled (again, for example), "rarely," "some of the time," "most of the time," and so forth (Crosby, 1998; Jaccard, McDonald, Wan, Dittus, & Quinlan, 2002; Sheeran & Abraham, 1994; Sheeran et al., 1999). The accuracy of ordinal condom use measures depends on participants sharing common understandings of the terms used to label categories, both amongst themselves and with study investigators. However, several studies have demonstrated that these terms are subject to individual interpretation and that respondents do not always interpret the category labels as intended (Biehl & Halpern-Felsher, 2001; Cecil & Zimet, 1998; Laswad & Mak, 1997; Ohnishi et al., 2002; Schaeffer, 1991; Weir, Roddy, Zekeng, Ryan, & Wong, 1998). For example, Cecil and Zimet found that 23% of college students believed that using condoms 2 times out of 20 is "never" using condoms, and a similar proportion believed that using condoms 18 times out of 20 is "always" using condoms. Importantly, the propensity of respondents to label less than 100% condom use as "always" using condoms could mislead researchers into believing that "always" users are practicing safer sex that will protect them from STI acquisition.

This study examined college students' use of the ordinal condom use labels "never," "rarely," some of the time," "most of the time," and "always" to determine whether they applied these labels consistently when the number of sex acts and the numerical frequency of condom use were varied. Part of this investigation was descriptive in nature. We were interested in determining how the labels were applied to different frequencies of condom use (i.e., 0%, 5%, 25%, 50%, 75%, and 100%). Based on the previous research (e.g., Cecil & Zimet, 1998), we anticipated that there would be between-subject variation in the assigned labels, indicating subjectivity in how participants interpreted these labels.

We also were interested in investigating whether the number of intercourse acts influenced how condom use labels were assigned. Participants were presented with scenarios describing various levels of condom use for either 20 or 100 acts of intercourse. If condom use labels are assigned based on the proportion of sex acts for which condoms are used, which is a fundamental assumption underlying the use of ordinal labels in assessing condom use, then the same label should be assigned to condom use for 10 of 20 acts as is assigned to condom use for 50 of 100 acts. Conversely, if different labels are assigned to these situations (i.e., if the total number of acts influences responses), one may question the validity of using ordinal labels to assess condom use in samples where the number of acts is non-uniform.

To test further whether the assignment of condom use labels was based on proportions, we compared two different formats for presenting condom use frequency information: one that stated the percentage of times condoms were used by a hypothetical couple and one that stated the number of acts for which condoms were used, out of a total of 100 acts of intercourse. Mental conversion of the number of protected acts out of 100 to a percentage requires only the simplest arithmetical operations. Therefore, if proportions were being used, one would expect close agreement in the assigned condom use labels for these two condom use frequency information presentation formats.

Finally, we sought to replicate the findings of Cecil and Zimet's (1998) study, in which participants problematically applied the labels "never" and "always" to scenarios in which condom use was neither 0% nor 100%, respectively. The results of this study are discussed with respect to their implications for the use of ordinal condom use measures.

METHOD

Participants

This study was part of a larger investigation of college students' perceptions of various sexual health behaviors. Participants were 282 undergraduate students attending a Midwestern university who were recruited from the psychology subject pool. Participants were told that they would be asked to complete a survey to assess their perceptions about different sexual behaviors and were informed that some of the items would be about their own sexual behavior. Anonymity of responses was emphasized. The mean age for participants in this study was 19.49 years (SD = 2.71). Almost half were male (48.6%), and the majority self-identified as White (86.7%).

Measures

The study questionnaire assessed demographics, sexual behaviors, sexual attitudes, and several topics related to the measurement of sexual behavior (Bogart, Cecil, Wagstaff, Pinkerton, & Abramson, 2000; Cecil, Bogart, Wagstaff, Pinkerton, & Abramson, 2002). Primary data were derived from a series of 15 questions that asked participants to judge how a hypothetical couple would label their condom use (e.g., as "some of the time") as a function of how frequently they used condoms. Five condom use frequency levels were assessed: 5%, 25%, 50%, 75%, and 95%. For each condom use level, frequency information was presented in one of 3 formats: (a) by stating the number of times that condoms were used out of a total of 20 acts (e.g., "A couple had vaginal intercourse 20 times. On 10 of those occasions, they used condoms"); (b) by stating the number of protected acts out of a total of 100 acts (e.g., "A couple had vaginal intercourse 100 times. On 50 of those occasions, they used condoms"); or (c) by stating the percentage of time that condoms were used out of a total of 100 acts of vaginal intercourse (e.g., "A couple had vaginal intercourse 100 times. They used condoms 50% of the time"). Participants then were asked, "How often do you think [the couple] would say that they used condoms?" There were 5 response options: "never," "rarely," "some of the time," "most of the time," and "always." Two additional items were used to assess participants' beliefs regarding how the hypothetical couple would label their condom use if they used condoms either for 0 or 20 of 20 total acts of intercourse.

Procedure

Participants completed a self-administered questionnaire in a group test setting with up to 12 persons in each group. The anonymous questionnaire took approximately 30 minutes to complete. This study was approved by the Institutional Review Board at the university where data were collected.

Data Analysis

Standard descriptive statistics were used to summarize the study population and their condom use labeling patterns. The Mann-Whitney U test was used to test for gender, age, and race/ethnicity group differences in the assignment of ordinal frequency labels (e.g., "never," "rarely") to each of the 17 condom use items. The Mann-Whitney U test is appropriate when the independent variable has two levels (e.g., male vs. female) and the dependent variable is measured on an ordinal scale (Kiess, 2002). This test is the nonparametric equivalent of the t-test.

The Friedman Test for Related Samples was used to determine whether, for each of the 5 condom use levels (5%, 25%, 50%, 75%, and 95%), the ordinal frequency labels assigned by participants differed as a function of the condom use information presentation format (number of protected acts out of 20 or 100 total acts, or percentage of acts for which condoms were used). The Friedman Test is a nonparametric test used for within-subjects designs in which each respondent provides matched ordinal-scale measurements (e.g., on the 3 condom use information presentation formats; Langley, 1968). This statistical test is analogous to a two-way, repeated measures analysis of variance (Marascuilo & Serlin, 1988).

The Wilcoxon Signed Rank Sum test was used to contrast pairs of condom use information presentation formats (e.g., number of protected acts out of 100 total acts vs. percentage of acts for which condoms were used) when the Friedman Test indicated that significant "omnibus" differences existed among the 3 condom use information presentation formats. The Wilcoxon Signed Rank Sum test is used for within-subjects designs in which the independent variable has only two levels and is measured on an ordinal scale. The Wilcoxon Test is the nonparametric equivalent of the paired t-test (Kiess, 2002).

Each intermediate category on a 5-point ordinal condom use frequency scale with anchors at "never" and "always" theoretically should encompass approximately one-third of the total range: "rarely" [approximately equal to] 1 to 33%; "some of the time" [approximately equal to] 33 to 66%; and "most of the time" [approximately equal to] 66 to 99%. Accordingly, both 5% and 25% condom use normatively should be labeled as "rarely" using condoms, 50% condom use as "some of the time," and 75% and 95% condom use as "most of the time." Participants' responses were compared to these normative expectations.

RESULTS

Table 1 displays the ordinal frequency labels assigned to 5%, 25%, 50%, 75%, and 95% condom use in each of the 3 condom use frequency information presentation formats: number of protected acts out of 20 or 100 total acts, or percentage of acts for which condoms were used. As indicated in the table, there was substantial interpersonal variability in the labels participants chose to use. For example, 41.1% of participants labeled condom use for 1 out of 20 acts of intercourse as "never" using condoms, 44.7% as "rarely" using condoms, and 14.2% as "some of the time," "most of the time," or "always." The Mann-Whitney Test was used to determine whether the category labels assigned by participants differed by gender (male vs. female), age ([less than or equal to] 19 years vs. >19 years), or race/ethnicity (White vs. non-White). A significant between-groups difference was found for only 1 of the 15 of condom use level/condom use frequency information presentation format combinations: male participants selected higher-valued frequency labels than did female participants for 95% condom use when condom use information was presented by specifying the percentage of time the hypothetical couple used condoms (z = -2.34, p = .02). There were no differences by age or race/ethnicity.

The Friedman Test for Related Samples was used to assess whether, for each of the 5 condom use frequency levels (5%, 25%, 50%, 75%, and 95%), the ordinal frequency labels assigned by participants differed as a function of the condom use frequency information presentation format. This analysis revealed significant differences as a function of the condom use information presentation format for each of the 5 condom use levels (all p's < .001; see Table 1). These results suggest that, if participants converted the number of protected acts out of 20 or 100 total acts of intercourse to the corresponding proportions, they did not do so accurately.

Follow-up analyses were conducted to gain further insight into the differences revealed by the Friedman Test. The Friedman Test is an "omnibus test." Significant findings on this test indicate that differences exist among the 3 condom use information presentation formats, but not whether the information presentation formats differ from each other on a pairwise basis. Pairwise contrasts were conducted using the Wilcoxon Signed Rank Sum Test. As indicated in Table 2, 11 of the 15 pairwise contrasts were statistically significant (p < .006), including 4 of the 5 comparisons between the "of 20 total acts" and the "of 100 total acts" condom use information presentation formats, and 7 of the 10 comparisons involving the "percentage of time" information presentation format.

With regard to the use of the "never" and "always" labels, only 78.0% of participants believed that a couple that used condoms for 0 to 20 acts would say they "never" used condoms. However, the "never" label also was applied to 1 out of 20 acts by 41.1% of participants. In contrast, nearly all (98.9%) participants believed that using condoms for 20 out of 20 acts would be characterized as "always" using condoms. But most participants (69.9%) thought this label also would be applied to condom use for 19 of 20 acts, and 13.8% believed it would be applied when condoms were used for 15 of 20 total acts. Participants' responses did not differ as a function of gender, age, or race/ethnicity group.

As discussed above, normative expectations suggest that both 5% and 25% condom use should be labeled as "rarely" using condoms, 50% condom use should be labeled "some of the time," and 75% and 95% condom use should be labeled "most of the time." As indicated in Table 1, these were the modal responses in this study with two exceptions. However, fewer than two thirds of respondents endorsed the modal response in most scenarios. Moreover, as noted previously, a substantial number of participants applied the "never" and "always" labels to intermediate condom use levels, particularly 5% and 95% condom use (see Table 1).

DISCUSSION

This study replicated and extended previous research (Cecil & Zimet, 1998; Jaccard et al., 2002) on the use of ordinal condom use measures. As hypothesized, there was substantial interpersonal variation in the labels assigned to particular condom use rates. For example, about one third of participants characterized using condoms for 5 of 20 acts as "some of the time," and about two thirds labeled it as "rarely" using condoms. These findings indicate that ordinal condom use labels are subject to individual interpretation (Cecil & Zimet). The lack of fixed between-subjects reference points calls into question the validity of ordinal condom use measures.

Interpretations of ordinal condom use labels not only differed from one individual to the next, but they also differed from the meanings typically assigned to them by researchers (Biehl & Halpern-Felsher, 2001; Cecil & Zimet, 1998; Fortenberry, Cecil, Zimet, & Orr, 1997). For many participants, "never" using condoms did not mean never using condoms (i.e., 0% condom use). Instead, the "never" label also was applied to infrequent (but non-zero) condom use. Conversely, some participants labeled 0% condom use as using condoms "rarely" or "some of the time." Similar comments applied to the "always" label. From a public health perspective, "always" using condoms means consistent (100%) use, without exception. Although using condoms for 19 of 20 sex acts reduces the risk of STI transmission, it does not qualify as "always" using condoms in this strict sense. For an STI with a per-act transmission probability of 0.2 (e.g., gonorrhea; Hooper et al., 1978), the cumulative probability of transmission from an infected person to an uninfected sex partner equals 0.33 if condoms are used for 20 of 20 acts of intercourse, but is 50% larger (0.46) if 1 of the 20 acts of intercourse is unprotected. Mischaracterization of less than 100% condom use as "always" using condoms, which was common in this study, can distort evaluations of the effectiveness of STI-prevention interventions with regard to the impact of consistent condom use.

Study participants applied different condom use labels depending on whether the total number of acts of intercourse was 20 or 100. This limits the reliability of ordinal condom use measures. Specifically, the label assigned to a particular condom use rate (e.g., 25%) may depend on the respondent's overall level of recent sexual activity, which could differ from one assessment point to the next. Moreover, for the 5%, 25%, and 50% condom use levels, participants' responses differed when condom use frequency information was presented by stating the number of unprotected acts out of 100 total acts rather than as the percentage of time that condoms were used for 100 total acts. This finding suggests that, in the "out of 100 total acts" scenario, participants were not basing their responses on the equivalent proportion of protected acts. Notably, the conversion from 50 out of 100 total acts to 50% condom use does not require advanced mathematical skills.

A main limitation of the present study is that participants were asked how a hypothetical couple would characterize their condom use rather than asked to characterize their own condom use behavior. This strategy was adopted to reduce the potential for self-presentation bias in reporting sexual behaviors (Bogart et al., 2000; Finkelstein & Brannick, 2000; Hannon, Hall, Nash, Formati, & Hopson, 2000). However, it leaves open the question of whether similar results would have been obtained had participants been queried regarding their own recent sexual behavior. We believe they would have. Two of the main findings--that ordinal label assignment varies from person to person and that the "never" and "always" labels are applied problematically--should not depend on whether a hypothetical couple's or a participant's own condom use behavior is assessed.

Another possible limitation is potential bias due to the order in which the condom use items were presented in the survey. Although the individual items were randomly ordered (e.g., 95%, 25%, 50% condom use) and the "of 20 total acts" and "of 100 total acts" items were randomly interspersed, item ordering and the ordering of the instrument sections themselves were identical on all surveys. Conceivably, this could have influenced some participants' responses. Future research should counterbalance the presentation of individual items and survey sections to determine whether or not presentation order affects participants' responses.

Finally, the results of this study may have limited generalizability to populations other than college students. Ordinal condom use measures have been used with adolescents, low-income men and women, gay men, and other groups whose interpretations of frequency labels such as "never" or "some of the time" might differ from college students'. Further research is needed to address this issue. Nevertheless, intermediate frequency labels (e.g., "some of the time") are inherently subjective, so interpersonal variability would be expected in most every population.

Ordinal condom use assessment measures have limited discriminability. Part of the difficulty is the limited capacity of the English language to express fine nuances between proportions. "Never" and "always" provide firm (though problematic) anchors, but intermediate condom use rates are not as easily mapped onto ordinal labels. Ideally, the "never" and "always" labels would be applied only to 0% and 100% condom use, respectively, leaving the 3 intermediate categories of a 5-point Likert scale ("rarely," "some of the time," and "most of the time") to describe condom use rates from 1% to 99%. Thus, each intermediate category encompasses approximately a 33% range of potential condom use rates. These wide ranges limit the sensitivity of ordinal condom use measures to detect epidemiological significant differences (or changes) in condom use, such as, for example, the difference between 67% condom use and 99% condom use. Increasing the number of intermediate categories increases sensitivity but exacerbates the language issue.

An alternate approach to increase sensitivity to fine gradations in condom use is to ask respondents the percentage of time they used condoms, using either a fill-in-the-blank format or percentage ranges (e.g., 0 to 10%, 11 to 20%, etc; Jaccard et al., 2002; Sheeran & Abraham, 1994). Little information is available regarding the reliability or validity of this approach. Asking respondents to assign a percentage to condom use behaviors seems to presuppose either (a) accurate recall of instances of protected and unprotected intercourse, together with an understanding of percentages and a modicum of arithmetical ability, or (b) an innate non-arithmetical facility for estimating percentages from recalled behavior. The results of this study do not support the notion that participants first converted condom use information (e.g., 5 of 20 or 25 of 100 total acts) into an equivalent percentage (i.e., 25%) and then assigned an ordinal label to this percentage. However, we did not assess the percentages participants would have assigned to 5 of 20 or 25 of 100 total acts (for example), and therefore this study cannot directly address participants' accuracy in assigning percentage values to various condom use scenarios.

A third approach, which permits assessment of infinite gradations in condom use, is to ask respondents to specify both the total number of acts of intercourse and the number of protected acts (Pinkerton et al., 1998b). The researcher can then convert this information into a condom use percentage and, as well, can infer the number of unprotected sex acts (Jaccard et al., 2002). From an epidemiological standpoint, the ability to infer the number of unprotected acts from the total number of acts and the number of protected acts is critical because it mainly is the number of unprotected acts that determines STI/HIV risk (Crosby, 1998; Jaccard et al.; Pinkerton et al., 1998a). Unlike ordinal or percentage-based condom use measures, the count-based approach does not require that respondents perform mental arithmetic or classify their condom use into a small number of subjective categories, but it does require that they accurately recall both the total number of acts and the number of condom-protected acts. Notably, to obtain estimates of the number of unprotected acts from either an ordinal or percentage measure, the total number of acts also must be assessed, introducing the same potential for recall bias as a count-based approach.

Empirical support for a count-based approach to sexual behavior assessment is found in a number of studies that demonstrate both the reliability and validity of this assessment strategy (e.g., Jaccard et al., 2002; McLaws, Oldenburg, Ross, & Cooper, 1990; Ramjee, Weber, & Morar, 1999). In one exemplary study, Jaccard and colleagues (2002) surveyed young adults on a weekly basis over a 1-year period, as well as at 1-, 3-, 6-, and 12-month intervals, and then compared the cumulative totals obtained by summing the weekly reports to the corresponding items on the extended recall period surveys. This study found good agreement between weekly and extended-period reports of condom use, particularly for the 3- and 6-month assessments.

The main limitation of count-based measures, as with all retrospective measures, is that reports of past behaviors are subject to several forms of recall bias (Catania, Gibson, Chitwood, & Coates, 1990), such as recency effects (more recent events are better remembered than are more distal events) and uniqueness/salience effects which cause frequent and less salient events to be less accurately recalled than rarer and/or more salient events (Baddeley, 1990). Consequently, validity is greatest when the recall period is relatively brief and sexual activity levels are not too high (Jaccard et al., 2002; Kauth, St. Lawrence, & Kelly, 1991).

Accurate assessment of condom use rates is necessary to evaluate the effectiveness of STI-prevention programs. Ordinal condom use measures are compromised by issues of validity (i.e., between-subject variability in the assignment of category labels), reliability (due to the dependence of category labeling on the number of acts of intercourse during the assessment period), and limited sensitivity to detect potentially significant differences in condom use rates. The results of this study suggest that these measures can produce inaccurate characterizations of condom use, and therefore that ordinal measures should be used sparingly, if at all.

Note. This research was supported, in part, by center grant P30-MH52776 and by postdoctoral training grant T32-MH19985 from the National Institute of Mental Health.

Manuscript accepted July 31, 2005

REFERENCES

Baddeley, A. M. (1990). Human memory: Theory and practice. Needham Heights, MA: Allyn & Bacon.

Biehl, M., & Halpern-Felsher, B. L. (2001). Adolescents' and adults' understanding of probability expressions. Journal of Adolescent Health, 28, 30-35.

Bogart, L. M., Cecil, H., Wagstaff, D. A., Pinkerton, S. D., & Abramson, P. R. (2000). Is it "sex"?: College students' interpretations of sexual behavior terminology. The Journal of Sex Research, 37, 108-116.

Catania, J. A., Binson, D., Van Der Straten, A., & Stone, V. (1995). Methodological research on sexual behavior in the AIDS era. Annual Review of Sex Research, 6, 77-125.

Catania, J. A., Gibson, D. R., Chitwood, D. D., & Coates, T. J. (1990). Methodological problems in AIDS behavioral research: Influences on measurement errors and participation bias in studies of sexual behavior. Psychological Bulletin, 108, 339-362.

Cates, W. Jr. (2001). The NIH condom report: The glass is 90% full. Family Planning Perspectives, 33, 231-233.

Cecil, H., Bogart, L. M., Wagstaff, D. A., Pinkerton, S. D., & Abramson, P. R. (2002). Classifying a person as a sex partner: The impact of contextual factors. Psychology and Health: An International Journal, 17, 221-234.

Cecil, H., & Zimet, G. D. (1998). Meanings assigned by undergraduates to frequency statements of condom use. Archives of Sexual Behavior, 27, 493-505.

Chesson, H. W., & Gift, T. L. (2000). The increasing marginal benefit of condom usage. Annals of Epidemiology, 10, 154-159.

Crosby, R. A. (1998). Condom use as a dependent variable: Measurement issues relevant to HIV preventing programs. AIDS Education & Prevention, 10, 548-557.

Finkelstein, M. A., & Brannick, M. T. (2000). Making decisions about condoms: Whose attitude is it anyway? Social Behavior and Personality, 28, 539-553.

Fortenberry, J. D., Cecil, H., Zimet, G. D., & Orr, D. P. (1997). Concordance between self-report questionnaires and coital diaries for sexual behaviors of adolescent women with sexually transmitted infections. In J. Bancroft (Ed.), Researching sexual behavior. Bloomington, IN: Indiana University Press, pp. 237-249.

Hannon, R., Hall, D. S., Nash, H., Formati, J., & Hopson, T. (2000). Judgments regarding sexual aggression as a function of sex of aggressor and victim. Sex Roles, 43, 311-322.

Holmes, K. K., Levine, R., & Weaver, M. (2004). Effectiveness of condoms in preventing sexually transmitted infections. Bulletin of the World Health Organization, 82, 454-461.

Holtgrave, D. R., & Pinkerton, S. D. (2000). Consequences of HIV prevention interventions and programs: Spectrum, selection, and quality of outcome measures. AIDS, 14(suppl. 2), S27-S33.

Hooper, R. R., Reynolds, G. H., Jones, O. G., Zaidi, A., Wiesner, P. J., Latimer, K. P., et al. (1978). Cohort study of venereal disease. I: The risk of gonorrhea transmission from infected women to men. American Journal of Epidemiology, 108, 136-144.

Institute of Medicine. (1997). The hidden epidemic: Confronting sexually transmitted diseases. Washington, DC: National Academy Press.

Jaccard, J., McDonald, R., Wan, C. K., Dittus, J., & Quinlan, S. (2002). The accuracy of self-reports of condom use and sexual behavior. Journal of Applied Social Psychology, 32, 1,863-1,905.

Kauth, M. R., St. Lawrence, J. S., & Kelly, J. A. (1991). Reliability of retrospective assessments of sexual HIV risk behavior: A comparison of biweekly, three-month, and twelve-month self-reports. AIDS Education and Prevention, 3, 207-214.

Keiss, H. O. (2002). Statistical concepts for the behavioral sciences (3rd ed.). Boston, MA: Allyn & Bacon.

Langley, R. (1968). Practical statistics. New York: Dover.

Laswad, F., & Mak, Y. T. (1997). Interpretations of probability expressions by New Zealand standard setters. Accounting Horizons, 11, 16-23.

Marascuilo, L. A., & Serlin, R. C. (1988). Statistical methods for the social and behavioral sciences. New York: W. H. Freeman & Co.

McLaws, M. L., Oldenburg, B., Ross, M. W., & Cooper, D. A. (1990). Sexual behaviours in AIDS-related research: Reliability and validity of recall and diary measures. The Journal of Sex Research, 27, 265-281.

Mullen, P., Ramirez, G., Strouse, D., Hedges, V. L., & Sogolow, E. (2002). Meta-analysis of the effects of behavioral HIV prevention interventions on the sexual risk behaviors of sexually experienced adolescents in controlled studies in the United States. Journal of Acquired Immune Deficiency Syndromes, 30, 594-S105.

Ohnishi, M. Fukui, T., Matsui, K., Hira, K., Shinozuka, M., Ezaki, H., et al. (2002). Interpretation of and preference for probability expressions among Japanese patients and physicians. Family Practice, 19, 7-11.

Pinkerton, S. D., Holtgrave, D. R., Leviton, L. C., Wagstaff, D. A., & Abramson, P. R. (1998a). Model-based evaluation of HIV prevention interventions. Evaluation Review, 22, 155-174.

Pinkerton, S. D., Holtgrave, D. R., Leviton, L. C., Wagstaff, D. A., Cecil, H., & Abramson, P. R. (1998b). Toward a standard sexual behavior data set for HIV prevention evaluation. American Journal of Health Behavior, 22, 259-266.

Ramjee, G., Weber, A. E., & Morar, N. S. (1999). Recording sexual behavior: Comparison of recall questionnaires with a coital diary. Sexually Transmitted Diseases, 26, 374-380.

Schaeffer, N. C. (1991). Hardly ever or constantly? Group comparisons using vague quantifiers. Public Opinion Quarterly, 55, 395-423.

Sheeran, P., & Abraham, C. (1994). Measurement of condom use in 72 studies of HIV preventive behavior: A critical review. Patient Education and Counseling, 24, 199-216.

Sheeran, P., Abraham, C., & Orbell, S. (1999). Psychosocial correlates of heterosexual condom use: A meta-analysis. Psychological Bulletin, 125, 90-132.

Stephenson, J. M., Imrie, J., & Sutton, S. R. (2000). Rigorous trials of sexual behaviour interventions in STD/HIV prevention: What can we learn from them? AIDS, 14 (Suppl. 3), S115-S124.

Warner, L., Newman, D. R., Austin, H. D., Kamb, M. L., Douglas, J. M. Jr., Malotte, C. K., et al. (2004). Condom effectiveness for reducing transmission of gonorrhea and chlamydia: The importance of assessing partner infection status. American Journal of Epidemiology, 159, 242-251.

Weir, S. S., Roddy, R. E., Zekeng, L., Ryan, K. A., & Wong, E. L. (1998). Measuring condom use: Asking "Do you or don't you" isn't enough. AIDS Education and Prevention, 10, 293-302.

Heather Cecil

Penn State University--Capital College

Steven D. Pinkerton

Center for AIDS Intervention Research, Medical College of Wisconsin

Laura M. Bogart

Rand Corporation

Jelena Pavlovic

Kent State University

Allison M. Kimball

Penn State University--Capital College

Address correspondence to Heather Cecil, Ph.D., Penn State University--Capital College, 777 W. Harrisburg Pike, W157 Olmsted, Middletown, PA, 17057; e-mail: [e-mail].

Table 1. Condom Use Frequency Label
Assignment as a Function of Frequency
Information Presentation Format (a)

                               % of responses

                         never     rarely      some
5% condom use
  1 of 20 acts           41.13      44.68     12.06
  5 of 100 acts          26.95      64.89      7.45
  5% of 100 acts          9.57      70.92     17.38
25% condom use
  5 of 20 acts            1.77      62.77     33.33
  25 of 100 acts          2.48      48.58     40.78
  25% of 100 acts         0.71      36.17     60.28
50% condom use
  10 of 20 acts           0.00       3.90     54.61
  50 of 100 acts          0.00       6.74     62.77
  0% of 100 acts          0.00       2.48     60.99
75% condom use
  15 of 20 acts           0.00       0.35      7.45
  75 of 100 acts          0.00       0.71     12.77
  75% of 100 acts         0.00       1.06     12.41
95% condom use
  19 of 20 acts           0.00       0.00      0.00
  95 of 100 acts          0.00       1.42      0.00
  95% of 100 acts         0.00       0.00      0.00

                                % of responses

                                              modal
                          most     always    response
5% condom use
  1 of 20 acts            1.42      0.71      rarely
  5 of 100 acts           0.71      0.00      rarely
  5% of 100 acts          2.13      0.00      rarely
25% condom use
  5 of 20 acts            2.13      0.00      rarely
  25 of 100 acts          7.80      0.35      rarely
  25% of 100 acts         2.84      0.00       some
50% condom use
  10 of 20 acts          39.72      1.77       some
  50 of 100 acts         30.50      0.00       some
  0% of 100 acts         36.52      0.00       some
75% condom use
  15 of 20 acts          78.37      13.83      most
  75 of 100 acts         84.40      2.13       most
  75% of 100 acts        84.04      2.48       most
95% condom use
  19 of 20 acts          30.14      69.86     always
  95 of 100 acts         52.13      46.45      most
  95% of 100 acts        56.03      43.97      most

                        Friedman
                          Test      [chi
                         mean      square]
                          rank    (df = 2)
5% condom use
  1 of 20 acts            1.78    103.6 ***
  5 of 100 acts           1.89
  5% of 100 acts          2.32
25% condom use
  5 of 20 acts            1.77    67.6 ***
  25 of 100 acts          2.04
  25% of 100 acts         2.20
50% condom use
  10 of 20 acts           2.10    26.5 ***
  50 of 100 acts          1.87
  0% of 100 acts          2.02
75% condom use
  15 of 20 acts           2.17    44.0 ***
  75 of 100 acts          1.91
  75% of 100 acts         1.92
95% condom use
  19 of 20 acts           2.25    96.6 ***
  95 of 100 acts          1.88
  95% of 100 acts         1.86

Table 2. Wilcoxon Signed Rank Sum Comparison of
Condom Use Frequency Information Presentation Formats (a)

Contrast                                        Z-score

5% condom use
  1 of 20 acts vs. 5 of 100 acts          -1.85    (p = ns)
  1 of 20 acts vs. 5% of 100 acts         -8.35    (p < .001)
  5 of 100 acts vs. 5% of 100 acts        -7.71    (p < .001)
25% condom use
  5 of 20 acts vs. 25 of 100 acts         -5.22    (p < .001)
  5 of 20 acts vs. 25% of 100 acts        -7.88    (p < .001)
  25 of 100 acts vs. 25% of 100 acts      -2.75    (P = .006)
50% condom use
  10 of 20 acts vs. 50 of 100 acts        -4.98    (p < .001)
  10 of 20 acts vs. 50% of 100 acts       -1.3     (p = ns)
  50 of 100 acts vs. 50% of 100 acts      -3.36    (p < .001)
75% condom use
  15 of 20 acts vs. 75 of 100 acts        -5.38    (p < .001)
  15 of 20 acts vs. 75% of 100 acts       -4.83    (p < .001)
  75 of 100 acts vs. 75% of 100 acts      -0.17    (p = ns)
95% condom use
  19 of 20 acts vs. 95 of 100 acts        -7.73    (p < .001)
  19 of 20 acts vs. 95% of 100 acts       -8.11    (p < .001)
  95 of 100 acts vs. 95% of 100 acts      -0.03    (p = ns)

(a) Presentation formats: number of protected acts of
20 total acts; number of protected acts of 100 total acts;
percentage of acts for which condoms were used.