Your toddler holds up a bath toy and announces, "Look — my fis!" You crouch down, smile, and do the thing every loving parent does on instinct: "Fish, sweetheart. Can you say fish?" They blink at you. "Fis." You try again, slower this time. So do they. Somewhere around the fourth round, the game collapses, the toy loses its magic, and you walk away wondering whether you helped at all.

Here is the uncomfortable truth: you probably didn't. And there is a sixty-year-old observation in child language research that explains why — one that begins, almost unbelievably, with that exact word.

The child who corrected the adults

In 1960, the psycholinguists Jean Berko and Roger Brown described a small boy who called his inflatable toy a fis. When an adult played along and asked, "Is this your fis?" the child refused the word flatly: no, it was his fis. The adults could go around this loop as long as they liked. Only when someone finally asked, "Is that your fish?" did the boy relax and agree: "Yes, my fis."

Read that again, because it is one of the strangest and most useful facts in all of early language: the child rejected his own pronunciation when it came out of an adult's mouth. He could hear, perfectly well, that "fis" was wrong. He simply could not stop his own mouth from producing it.

Researchers came to call this the fis phenomenon, and it has been observed over and over since. It tells you something that should change how you respond to every mangled word at your dinner table: your toddler already knows what the word is supposed to sound like. The error is not in their ears. It is in their mouth.

You're correcting the wrong system

Saying a word involves two separate systems that mature on wildly different schedules. The first is perception — the stored template of what "fish" sounds like. That system is precocious; babies are tuning it before their first birthday. The second is production — the real-time coordination of tongue, lips, jaw, breath, and vocal folds, roughly a hundred muscles executing a motor plan in a fraction of a second. That system is a construction site until age four or later.

When you say "Fish. Can you say fish?" you are handing your toddler information they already possess and demanding a motor skill they do not yet have. It's a bit like correcting a beginning skater's crossover by shouting the word "crossover" more clearly. The knowledge was never the problem.

The same is true for grammar. In one famous exchange recorded by the psycholinguist David McNeill, a child said "Nobody don't like me," and his mother patiently corrected him — say "nobody likes me" — eight times in a row. On the final round, the child brightened and produced: "Oh! Nobody don't likes me." And when Roger Brown and Camille Hanlon combed through recordings of real parent–child conversation, they found parents rarely correct grammar at all — they respond to whether a child's statement is true — and children acquire grammar just fine anyway. Explicit correction turns out to be neither common nor necessary.

There's a quieter cost, too. Toddlers talk for two reasons: to get things and to share things. "Say it right" converts a moment of sharing into a pop quiz. Do that often enough and a child learns that talking to you sometimes triggers an exam — and the surest way to avoid failing is to attempt less. The correction meant to speed things up quietly shrinks the number of attempts, and attempts are the entire engine of speech development.

The recast: a correction that doesn't feel like one

So what should you do — just let "fis" and "pasghetti" stand? No. You use the tool speech-language pathologists reach for first: the recast. A recast means you repeat what your child said, in its correct form, as a natural part of your reply — with no demand attached.

Child: "My fis!" You: "Yes! Your fish. He's a fast fish, isn't he?"

That's it. No "can you say," no expectant pause, no test. It looks almost too gentle to work, but it attacks the problem at exactly the right joint. The language researcher Michael Saxton called the mechanism direct contrast: because your correct version arrives within a second or two of the child's own attempt — while their version is still active in their memory — the mismatch between the two is maximally noticeable. The child's brain gets a clean side-by-side comparison at the only moment a comparison is possible.

Studies of parent–child conversation have repeatedly linked recasts to faster uptake of the corrected forms, and recasting is a core technique in professional speech therapy for exactly this reason. It also has a social elegance that explicit correction lacks: the child got what every speaker wants — to be understood and responded to — and the correct model rode in for free. The conversation keeps moving, which matters enormously, because back-and-forth conversational turns are among the strongest predictors of language growth we know of.

Your toddler will not repeat the word correctly on the spot. That is fine. The fis phenomenon already told you the template is being stored; production catches up on its own timetable, fed by thousands of these tiny contrasts.

When to keep an ear on it

Almost all early mispronunciations are systematic simplifications that fade on schedule. A rough clinical rule of thumb: strangers should understand about half of what a child says at two, and nearly all of it by four. If your child is far off that pace, or — notably — if they don't react when you playfully imitate their errors (no protest, no sign they hear a difference), mention it to your pediatrician and ask about a hearing check and a speech-language evaluation. One short appointment beats a year of quiet worrying, and early support works best early.

Your next moves

  • Retire "Can you say ___?" for one week. Every time the urge hits, recast instead: repeat the word correctly inside a warm, natural reply. Notice whether your child talks more by Friday.
  • Pick one target word your toddler reliably flubs and give it the double-model treatment: use the correct form twice in your response ("Your fish! What a big fish"), then move on. Twice sounds natural; five times sounds like a lesson.
  • Answer the message before the grammar. If they say "doggy runned away," respond to the drama first ("He did! He ran away so fast!"). The recast is smuggled inside the empathy.
  • Run the fis test once, gently. Playfully use their version of a word ("Is that your pasghetti?") and watch. If they protest or look at you funny, celebrate privately — their perception is ahead of their production, exactly as it should be.
  • Jot down intelligibility, not errors. Once a month, note roughly how much of your child's speech a stranger would follow. That single number tells you more about whether to seek help than any individual mispronounced word.

This is also the philosophy behind Acorn, our daily three-minute first-words session for one-to-three-year-olds. Acorn never asks a toddler to perform a word on command — it works the way the science says words are actually built: clear, warm models of one word at a time, repeated across days, with zero testing, zero ads, and zero upsells. If you'd like those three minutes of modeling handed to you each morning so you can spend the rest of the day just talking with your kid, Acorn is waiting at acorn.lumenlabs.works.