In sub-Saharan Africa more women are infected with HIV/AIDS than men and new prevention methods are urgently needed. One major attribute of female-initiated HIV prevention methods is that they can be used covertly, without a male partner's knowledge. Using mixed methods, we explored the predictors and dimensions of covert use of the diaphragm in a randomized controlled trial that tested its effectiveness for HIV prevention. The Methods for Improving Reproductive Health in Africa (MIRA) trial was conducted in Zimbabwe and South Africa, and data collection took place between September 2003 and January 2007. This study is a secondary analysis of quantitative and qualitative data from participants randomized to the intervention group, and their male partners. It includes survey data from 2316 women (mean age=28.3), 14 focus group discussions (FGD) conducted with 104 women, and 7 FGD and 10 in-depth interviews with 37 male partners. The median follow-up for trial participation was 21 months (range: 12-24). At their final visit, approximately 9% of women had never disclosed to their primary partners that they were using the diaphragm (covert use). In multivariate analysis, predictors of covert use included being older, not co-habiting with the partner, having a partner who did not use condoms, and being from South Africa. Qualitative analysis revealed that covert use was not dichotomous, but ranged along a continuum, which we categorized into five levels (i.e. full disclosure; mostly open use; occasional covert use; mostly covert use; and completely covert use). We discuss the critical role of the option of covert use for many women in the context of an HIV prevention trial, as well as gender power dynamics which may influence women's decisions about disclosure.