This project analyzes teacher behavior using a corpus of 110 hours of recorded instruction in order to document the realities of second-language pronunciation instruction. Teachers of six intact classes were recorded during a seven-week oral communication course. Four classes implemented a focus on pronunciation that was integrated into the syllabus in one of two ways: form-focused (only) instruction ('explicit'), or form-focused and communicative at the same time ('communicative'). The other two did not have a specific focus on pronunciation ('NSP') and serve as the baseline. Our goal was to establish how much more time was spent on pronunciation when integrated compared to the NSP baseline, and to document how teachers implement various pedagogical choices while doing so. Our findings show that in the NSP groups, pronunciation instruction happened for a very short time and was mostly reactive. In comparison, the proportion of time with a pronunciation focus was 10 times greater in the integrated groups, where pronunciation-related episodes covered a quarter to a third of class time. Furthermore, a large portion of the integrated instruction was pre-planned, unlike in the NSP groups. The analysis of how this time was spent in the integrated groups further shows that aside from a core number of dedicated pronunciation lessons, a pronunciation focus was achieved by integrating activities with the course materials in variable ways and to varying degrees. This dataset shows that integrating pronunciation into existing syllabi is possible and does not happen at the expense of the other course goals. Therefore, integration appears to be a successful approach from a pronunciation instruction standpoint since the time accumulated also led to measurable improvements in comprehensibility after seven weeks for the observed students. We outline an integration model to help guide instructors towards implementing integrated pronunciation instruction even in courses without explicit pronunciation goals.