Multisensory research using fMRI presents specific challenges for appropriate experimental designs and analysis strategies. We report about a series of experiments investigating letter-speech sound integration using different experimental designs, paradigms and analysis methods, including block and event-related designs, adaptation paradigm, passive vs active performance, conjunction analysis as well as standard vs advanced brain alignment techniques for group analyses. While some results were rather consistent, some design and analysis choices lead to unexpected findings. We will discuss the implications of the obtained insights in the context of general multisensory fMRI research.