What to Correct, and What to Let Go
The instinct when your child says “wabbit” is to say “no — rabbit.” It is a reasonable instinct and it is usually the wrong move, because a child who is corrected all day learns that talking to you is a test. Children who feel corrected talk less, and talking less is the opposite of what speech practice needs.
The technique clinicians use instead is called recasting: you say the word back correctly inside a normal sentence, without demanding a repeat. Your child says “I saw a wabbit,” you say “you saw a rabbit! Where was it?” They hear the correct version, they are not put on the spot, and the conversation continues. Do that ten times a day and you have delivered ten clean models with no arguments.
Outside a dedicated practice window, that is essentially all you should do. Inside the practice window you can ask for productions directly — that is what the window is for — and even then, correcting only the target sound and letting every other error pass is what keeps the session short and the child willing.
Model Rather Than Drill
The most reliable thing an untrained adult can do is provide a very large number of correct models. Clinicians call the deliberate version auditory bombardment: you load your own speech with the target sound while your child listens, with no requirement for them to say anything. Reading a book and stressing every target word, or narrating a snack while leaning on the sound, both count.
When your child does attempt the word, respond to the attempt rather than grading it. “That was close, I heard the ssss at the start” tells them what worked and what to aim at. “No, say it again” tells them they failed and gives them nothing to change. If three attempts in a row go wrong, model it once more and move on — the fourth attempt is rarely better and the mood usually is worse.
Resist coaching the mouth unless the SLP has shown you how. Telling a child to “put your tongue behind your teeth” for an R sound is a common piece of well-meaning advice that describes the wrong place, and a child who practises a wrong position gets very good at a wrong position.
Practise the Target the SLP Actually Set
Ask your child's therapist four specific things and write the answers down: which sound, which position in the word (start, middle, or end), which level (single words, phrases, sentences, or conversation), and what a correct production should sound like when your child manages it. Those four answers are the difference between useful home practice and enthusiastic noise.
Level matters more than parents expect. A child who can produce a perfect S in single words often cannot hold it in a sentence, and asking for sentences before they are ready produces a run of failures that makes them stop trying. If practice is going badly, dropping back a level is usually the fix rather than practising harder.
It is also worth asking what not to work on yet. Late-developing sounds like R, S, Z, L, and TH commonly are not mastered until six, seven, or eight, and a therapist who has left one alone has usually left it alone on purpose. Our sound guides list which errors are typical at which age, so you can see where your child sits before you worry.
How Much Is Enough, and When to Ask for Help
Five to ten minutes, a few times a week, done properly, beats twenty minutes of a child being corrected while they would rather be elsewhere. Frequency is what builds the habit; the number of correct productions is what moves the sound. If you get more than that, good — but do not treat a missed day as a failed week.
Signs practice is going well are usually behavioural before they are audible: your child starts self-correcting without being asked, or produces the target sound in a word nobody prompted. That is generalisation starting, and it is a better indicator than a scoring number.
Go back to the SLP if your child is becoming reluctant to talk, if they are frustrated to the point of avoiding practice entirely, if a sound they had mastered has slipped, or if you genuinely cannot tell whether their attempt was right. That last one is not a failing — hearing the difference between a good R and a nearly-good R is a trained skill, and asking is faster than guessing for a month.