A system and method for making animals appear to talk is provided. The system includes a speaker mounted on an article (such as a collar) worn by the animal. A memory stores a plurality of pre-recorded messages. These pre-recorded messages may be generic phrases to be "spoken" by all types of animals or may be messages which are based on the characteristics of a particular animal. Using a selection circuit, a user is able to selectively output the pre-recorded messages from the memory to the speaker, thereby making the animal appear to talk.