Teaching what questions
What has two common answer types: a thing ("What is she holding?") and an action ("What is he doing?"). Practise both, and point out which one the question wants.
Level 3 what questions are predictions and inferences ("What will the boy do next?", "What just happened?"). They are a good bridge to why, because the child has to use clues rather than just look.
How the three levels work
Every question word has three levels, following the usual clinical progression from literal to inferential questions. Level 1 is literal and obvious, with two choices: point to a big thing in the picture. Level 2 is literal but needs looking: a smaller detail, a spatial word, telling two people apart, with three choices. Level 3 is inferential, with three choices: use clues in the picture plus what you know about the world.
Every question was checked against its picture: the right answer is plainly true in the image and every wrong answer plainly false. A session is 10 questions of one type and level, at most two from the same picture.