We are now at a point in time, where telephone surveys have to face many challenges and where Internet surveys are not yet fully fit to replace them. Mixed-mode surveys are offered as a promising solution. One of the most pressing challenges facing designers of mixed-mode surveys is combining telephone results with those from web surveys. There is considerable evidence that when different modes are used different results are obtained. Data from different modes may be different because the modes themselves lead to different response processes, or because different questions are employed in different modes (either by tradition or due to mode specific optimization of the questionnaire). For instance, because of the memory problems associated with verbal communication, telephone designers often use unfolding, or polar (end) labeling of response categories. The goal of this research is to identify means of being able to combine telephone and web survey results effectively, by choosing formats for asking questions that are less affected by mode, than are others. In a series of experiments we test multiple formats for asking questions (e.g. unfolding opinion questions in multiple steps vs. asking a complete question in one-step) within each of the two survey modes. These experiments go beyond many previous tests because they involved not only comparisons across modes of particular formats, but within modes of two different formats. Members of a high quality, probability-based Internet panel (LISS-panel, Centerdata) were randomly assigned to one of two modes: a computer assisted telephone interview or a web survey. Within each mode the same series of split ballot experiments on question format were conducted. This design enables us to disentangle question format effects from ‘pure’ mode effects. In addition, the rich background data of the LISS-panel members makes it possible to correct for potential nonresponse differences between the modes.