The speech that these devices produce always use the same voice, without adapting it to the user, the context or the environment. To improve the speech to be more intelligible, we recorded how well people understood a robot in different circumstances. In the HomeLab we explored different acoustic environments, from a large open-plan living room to a small bedroom, to see how the acoustic quality of the room changes how well people understand the robot, and how the robot needs to change its speech patterns to overcome changes in the acoustic environment.