Televised debates are regularly accompanied by an extensive coverage in social networks like Twitter. When receiving messages via second screens, people are embedded in their idiosyncratic, personalized social network, possibly leading to biases or in other words: filter bubbles. Given that self-selective processes are at work, though, it is usually hard to determine what the effects of second screens and filter bubbles are. In order to be able to test for such effects, we have conducted a lab as well as a survey experiment: Respondents were randomly allocated to groups, which were then treated with specific, but biased and real Twitter coverage while watching the most recent televised debate from Germany. The lab experiment was run as a live experiment on debate night. Results of both experiments show: Filter bubbles do matter.