When writing questions with dichotomous response options, those administering surveys on the web or on paper can choose from a variety of formats, including a check-all-that-apply or a forced-choice format (e.g. yes-no) in self-administered questionnaires. These two formats have been compared and evaluated in many experimental studies. In this paper, we conduct a systematic review and a few meta-analyses of different aspects of the available research that compares these two formats. We find that endorsement levels increase by a factor of 1.42 when questions are posed in a forced-choice rather than check-all format. However, when comparing across a battery of questions, the rank order of endorsement rates remains the same for both formats. While most authors hypothesise that respondents endorse more alternatives presented in a forced-choice (versus check-all-that-apply) format because they process that format at a deeper cognitive level, we introduce the acquiescence bias hypothesis as an alternative and complementary explanation. Further research is required to identify which format elicits answers closer to the 'true level' of endorsement, since the few validation studies have proved inconclusive.