I like this a lot. It would be interesting to combine this with objects in a discourse context drawing on some rich psycholinguistic research.

Also, to look at interactions with task demands (point to the brush, pick up the brush - not a great example, but it's early).