I must admit I don't think it is a great test. Counting coloured T-shirts and spotting waving bunnies has nothing to do with driving safely. If your going to set up test like this, it needs to be more realistic to have an impact, i.e. hit the spacebar when you spot a hazard, whilst listening to someone talking. You could prove that they have paid attention by asking questions about the conversation afterwards.
Do this with and without the converstation and compare results. This would have a greater impact IMO, though it might not hold interest though...