two out of four of my test subject misinterpreted my instructions and actually drew what I had intended them to draw – a maze.
I made the mistake of not defining the condition that would make them stop drawing a line when that line was about to intersect a line that had been previously drawn.
Surprisingly only one exclaimed that I had got them to draw a maze while they were still in the process of drawing. The others only made the realisation after finishing and looking back at their work. The subject who made the realisation actually exclaimed “hah, now the mouse can’t get the cheese” – he purposely blocked off the cheese. The environment and interpretation of my intentions in giving them the task affected the image that he drew.
I think that this might be an interesting starting point for and idea to try on Mechanical Turk – getting people to generate mazes which other people then solve.