This is pretty cool. Researchers from Carnegie Mellon University and Stockholm University have created a videogame that uses garbled speech to figure out how infants learn language:
To uncover how spoken sounds are decoded, the research team designed a video game narrated in deliberately distorted speech. The soundtrack, unintelligible in any language, was the only source of instruction for 77 adult players in the study. With just two hours of play, the participants could reliably extract word-length sound categories from continuous alien sounds and apply that learning to advance through the game.
Read more here.