Train the mouse to find the cheese by clicking out a path through the maze and selecting train mode a few time. Then test using the test mode. The real trick is to get the mouse to associate the door it picks from the start room to know which door to pick at the maze exit room to find the cheese. To do this, first train all the possible start/end combinations while using the same maze path. Then train a different maze path by clicking out just the rooms from entry to exit. Then go back to testing the complete start to goal paths. You may have to give it a couple of times to get used to the new maze path. If it succeeds, it means that the knowledge about what to do at the maze exit is independent of the intermediate task of getting through the maze itself.