Ming from Agents of Speech shows how to teach your child language by modeling target phrases repeatedly, like "I want water," then demonstrating the action and fading the model as they pick it up. Skip the corrections and questions.
If you want a child to say what you want to hear, then you must speak as many times as possible about the thing you want to hear. Whatever you want your child to hear, you're going to have to model it instead of asking questions, instead of correcting the child whenever they say it wrong. Because it becomes a vicious cycle of, oh, I need to say something incorrect and then mom says something right, and I get rewarded. Rather than that, we want to teach the kid incorporated immediately and actively, and we would like model what they should be saying instead. If a child wants water, then before the child even talks, you're going to say like, I want water, right? That's easy. Or mommy, pour water. And then you demonstrate and you do it one more time so you pour little by little so that the child keeps on trying and getting your model and trying to imitate you, right? And then when the modeling becomes so easy and you can fade it out and then the child can actually say it by him or herself, all right?
