So the user forms an association between the shape of the button and its function, that way they are easier to differentiate in the dark.

Edit: I totally beat the stackexchange answer by 4 minutes :)