It is expected that in the near-future people will have daily natural language interactions with robots. However, we know very little about how users feel they should talk to robots, especially users who have never before interacted with a robot. The present study evaluated first-time users' expectations about a robot's cognitive and communicative capabilities by comparing robot-directed speech to the way in which participants talked to a human partner. The results indicate that participants spoke more loudly, raised their pitch, and hyperarticulated their messages when they spoke to the robot, suggesting that they viewed the robot as having low linguistic competence. However, utterances show that speakers often assumed that the robot had humanlike cognitive capabilities. The results suggest that while first-time users were concerned with the fragility of the robot's speech recognition system, they believed that the robot had extremely strong information processing capabilities.