I've been thinking about this one all morning, Herman, and I still don't know whether to laugh or open a linguistics textbook.
Then we're off to a strong start, because I did both. I actually printed the prompt out, which I never do.
Daniel wants Ezra to grow up sounding like a member of the landed gentry. Fifteen months old, already has a project manager.
Project Papa. He named it.
He did. And the first deliverable is that Ezra calls him Papa, not Dad, ideally in a wonderfully refined English accent. The household is Irish father, American mother, Jerusalem address. Nobody in the building has ever said the word "spiffing" out loud.
Except Daniel, now, presumably, at breakfast.
Constantly. There is a genuine question buried under the comedy, which is the part that kept me reading. Can you actually engineer a specific accent and register in a child by modelling and repetition, or does the child's environment simply win? And if it can be done, how long before the playground takes it apart?
He also asked for a vocabulary substitution list of about twenty expressions, a practical training programme, and a verdict on whether any of it survives contact with other children.
And he was careful to note that child welfare services needn't be notified, which is the most reassuring sentence I've read all week.
That's the correct amount of reassurance, honestly. Here's what I find interesting before we get anywhere near the funny part. Ezra is fifteen months old. That's not a random age. That's the middle of the single most sensitive window humans have for phonetics.
Sensitive how?
Sensitive in that it closes. Between six and twelve months, an infant's ability to discriminate phonetic units that don't exist in their native language sharply declines. They come in able to hear everything, and they tune down to what they need. By twelve months, the tuning has largely happened.
So the window for hearing a foreign sound is already mostly shut at fifteen months.
For perception, mostly. Which is the first complication for Project Papa, because the accent target isn't a set of new phonemes Ezra needs to hear, it's a particular distribution of ones he already has. That's actually more tractable. What matters at fifteen months is that he's at the point where words start arriving in volume.
And the household has no native speaker of the target. That's the core tension. You want a specific accent and the only available models are Irish, American, and whatever Jerusalem English turns into when four languages share a kitchen.
Which is where the Kuhl work becomes the whole story. Nine-month-old American infants, twelve sessions of twenty-five minutes over four weeks, live Mandarin speakers in the room. They learned to distinguish Mandarin phonemes, and they performed statistically level with infants in Taiwan who'd been hearing Mandarin for ten months.
Four weeks of live sessions matching ten months of immersion.
And then the piece that kills the media strategy. A second group got the identical content on DVD, or on audiotape. Same words, same voices. They scored at the same level as English-only controls.
Nothing.
Nothing. Kuhl's line was, video plus audio or audio-only presentation did not work for infants nine and ten months of age. That's not how infants learn language.
So the screen is furniture. The audio is wallpaper.
At that age, yes. And it replicates for accent specifically. English-learning fifteen to eighteen-month-olds in Canada couldn't recognize familiar words delivered in an unfamiliar accent after video exposure. Same kids, live experimenter, and they could.
That's the finding that decides half of Daniel's plan before he's started.
It also isn't quite the whole picture, and I want to be fair to the audio idea. Five to seven-year-old Cantonese speakers who heard Putonghua songs as ambient sound in the background improved their Putonghua accent. So non-interactive input isn't worthless forever. It just requires an age and an engagement level that fifteen months doesn't have.
The audio plan isn't wrong. It's early. By about five years.
Which is a strange thing to say about a language-training regime. Usually the advice is start earlier, and here the advice is that this particular tool only switches on later.
What about the quality of the input itself? The way parents talk to babies, that singsong thing. He'll be doing that already, presumably, or Hannah will.
Motherese. It sounds ridiculous and it's doing real work. Mothers with the highest speech clarity tend to have children with the highest speech perception, and every mother exaggerates vowels when addressing an infant. It's not decoration. It's the input format the brain is waiting for in the second half of the first year.
So if you wanted to smuggle a register in, you'd want to do it inside motherese rather than beside it.
Exactly right. And that's the practical bridge to the second half of Daniel's plan. Which is the vocabulary list, and I have to say, the list is where this gets funny.
Give me the shape of it first.
So the classic apparatus here is the U and non-U distinction. Alan Ross, 1954, a paper on linguistic class indicators. Nancy Mitford popularised it. U for upper class, non-U for everyone else. The pairs are famous. Sofa is U, settee is non-U. Napkin is U, serviette is non-U.
That's backwards from what I'd have guessed, every single time.
Everyone guesses backwards. That's the point of it. Jam is U, preserve is non-U. Looking glass is U, mirror is non-U. Drawing room is U, lounge is non-U. And the toilet one, which you already know is a trap. Loo or lavatory is U, toilet is non-U.
I'd have said it the wrong way and then felt the room go quiet.
Then there's pronunciation, which is where Daniel's "refined accent" ambition actually lives. Forehead rhymes with horrid. Waistcoat is weskit. Always is AWLwhizz.
Always is what?
AWLwhizz.
That's not a word, Herman, that's a sneeze.
Clothes is close. Golf is goff. Off is orf. Marvellous is two syllables, marvlous. And rarely means really, which I refuse to explain because I don't understand it either.
Somebody built a whole social system on a set of pronunciations nobody can derive and everyone has to memorize. It's a shibboleth with a dress code.
That's what Ross said, roughly. It is solely by its language that the English upper class is clearly marked off from the others. Which is a remarkably blunt sentence for a linguistics paper.
Then give me the actual substitutions. The twenty Daniel asked for. He wants Ezra saying "how splendid" instead of "that's great."
So the register words, the ones that carry the tone rather than the correctness. Marvellous. Jolly. Rather. Beastly. Terribly. Yonks. Tight or blotto for drunk, which is not a fifteen-month problem but we'll file it.
File it.
Seedy for ill. Bind for a problem. Ass for a fool. Brick for a reliable person. Ravishing, thrilling, bore, golly. Then the child-friendly ones, which is where I think the actual list lives. "Spiffing." "That's capital." "Jolly good show." "Oh, golly gosh." "I say, old chap." "Well done, old bean." "That's a load of poppycock." "He's a good egg."
So instead of "that's great," how splendid. Instead of "I don't like that," how frightfully disagreeable. Instead of "good job," well done, old bean.
And instead of "no," presumably, "I'm afraid not, old chap."
Instead of "more banana," "I should rather like another banana, Papa."
You've just written the target sentence for the whole project.
I want to flag something, though, because you're enjoying this too much.
I'm aware of how much I'm enjoying it.
Braber and Sandow actually tested Ross's claims. Twenty-first century, real speakers. And the supposedly U words, sofa, napkin, turned out to be more common overall. Not rarer. Where differences existed, they tracked age, not class.
That's right, and I do bring it up.
Because it's important. You can't just hand a list to Daniel and let him teach Ezra a parody of an aristocracy that didn't really talk like that. Some of these words are what Ross himself labelled non-U. The serviette crowd. And you'd be teaching the child the wrong tribe's vocabulary with enormous confidence.
The guidance from that work is roughly, don't start with a vocabulary list that's seventy years old. And they're right about the premise. But I'd defend the list for a different reason. Daniel isn't trying to penetrate the upper class. He's trying to build a bit of theatre, and a precise, slightly antique vocabulary does that better than an accurate one.
So it's a fantasy costume and the seams are part of the charm.
The words that are still in circulation do the work. Marvellous, terribly, jolly, spiffing. Those read as posh to a modern ear, whether or not Ross would have assigned them correctly in 1954.
And the fossilized ones, weskit, AWLwhizz. Those are the ones that sound like a parody, because they are one.
And I want to hold onto that distinction for the programme. Live modelling for the words that can be spoken naturally. The archaic pronunciation set is costume jewelry. It can be taught to a five-year-old as a trick, but it will never be how Ezra thinks.
Say more about that. Because I think that's the real answer to Daniel's question and it's more interesting than yes or no.
Here's where the second-order stuff lives. Vocabulary input frequency correlates with how early a word is acquired. So if you say "splendid" instead of "great" consistently, Ezra learns splendid on the schedule he'd have learned great. That part is straightforward and Daniel can do it.
Because you're just changing the surface form of the input, not the timing.
Right. But verb diversity in parent input at twenty-one months accounted for thirty percent of the variance in how many different verbs the child produced six months later. Thirty percent, from one input variable. That's a large effect for something that looks like just talking.
So the richness of the input matters more than the accent of it.
The accent is a veneer on top of a much bigger thing. Which means the honest answer to Daniel is that he can absolutely get "how splendid" into the child's mouth, but the thing that's actually going to shape Ezra's language is the variety and volume of what everyone in that flat says to him, in whatever accent they say it.
And the accent himself, the vowels and the rhythm, that's where the live human becomes non-negotiable.
That's the cleanest version of the answer. The register is teachable by a non-native. The accent is not, without a speaker of it in the room.
I want to press on the timeline, because Daniel's plan runs from fifteen months through early childhood, and I don't think the environment sits still for that.
It doesn't, and this is the part I'd put at the top of his plan if I were writing it for him. Parental dialect dominates in young childhood. Community influence strengthens later. The Kerswill and Williams study in Milton Keynes looked at forty-eight children at ages four, eight, and twelve, and the shift toward the local peer norm was there and it grew with age.
So the deadline isn't a cliff, it's a slope.
And it starts steeper than people think. Three and four-year-olds already alter their ing-versus-in based on how old the person they're talking to is. So the social conditioning is running before anyone expects it.
But the earlier finding said children under four don't modulate by formality. So they're not switching register.
Not by formality, and that's the important distinction. They don't have an upstairs voice and a downstairs voice yet. But they do adjust to who's in front of them. So the minute Ezra starts picking up that the caretaker talks one way and the playground talks another, he's already running the machinery that will eventually favor the playground.
What about the parents' side of it? The modelling. You said mothers do this consciously.
They do, and this is the bit that should amuse Daniel. In a British dialect study, mothers drastically reduced glottal t in speech to their children, especially to girls. Which means the instinct to model a desirable form is completely normal. Every parent already does it.
Daniel's just doing it with a more ambitious target and no native speaker in the house.
The interesting wrinkle is that the mothers in that study were suppressing a marked feature in favor of the local standard, and Daniel wants to install a feature that isn't standard anywhere in his house. Different project, same machinery.
So the machinery is there. The question is whether the input exists.
And that's where the practical part of the programme becomes interesting. Because to mimic the Kuhl dose, live sessions, twenty-five minutes, twelve of them, you'd need a speaker of the target accent. So a nanny, a tutor, a patient British relative, anyone who'll take the gig.
Twelve sessions over a month. That's thirteen hours of a person's life.
And the point isn't that thirteen hours is a lot. It's that thirteen hours of live input outperformed ten months of ambient exposure for phoneme discrimination. Live beats passive by a ratio that seems unreasonable.
I believe it, looking at how much a fifteen-month-old learns from a face versus a screen. I've watched Ezra watch a screen. It's a different animal. He's not learning, he's being distracted.
Which is why I want to frame the whole thing as a live-input problem rather than an accent problem. Daniel doesn't have an accent gap in his plan. He has a live-speaker gap, and everything else follows from whether he closes it.
So let's get concrete. What does Project Papa actually look like if you run it honestly? Give me the real version, not the wishful one.
I'd start at fifteen months and I'd start with size. Not twenty expressions. Three.
Three.
Papa, splendid, and one expression of displeasure. Those three cover the request that's emotionally central to Daniel, the register marker that's easiest to plant, and the one that will delight everyone at a family dinner. Everything else is later.
And how do you plant them.
Live modelling inside ordinary speech. You say "Papa's here" instead of "Daddy's here" every single time, and you never alternate. Consistency is the whole mechanism at this age, because the child is doing frequency counting on the audio stream.
If you slip back to Dad twice a day, you've just handed him two competing forms.
And he'll pick the one that wins on frequency or the one that wins on usefulness. Which is Dad, in Jerusalem, at the playground, every time. So the twelve to twenty-four month phase is a solo performance for Daniel. He has no competition yet, except from Hannah and the wider family.
What about Hannah? Does she have to buy in?
If she doesn't, it fails silently. A child who hears Papa from one parent and Dad from the other, in the same day, in the same register, will simply default to Dad, because it's the one that's reinforced by every adult outside the flat.
So step zero is that this is a two-player project or it's nothing.
Which is probably the funniest line in the whole plan, honestly. The most important input variable in Project Papa is whether Daniel's wife is willing to sign the treaty.
She's an American. She's about to spend fifteen years saying the word "spiffing" to a child in Jerusalem.
I imagine the negotiations are ongoing.
I imagine the negotiations ended before they started, in one direction.
Tell me where you'd put the milestone at three years. Because I have a view and I want to see if you land in the same place.
At three, my expectation is a small repertoire of words that sound ridiculous when he's cranky, because the moment he's upset he'll strip the register back to whatever he's got in his own head. The register is a performance and a three-year-old doesn't perform when he's crying.
So you're saying the register is real only when he's calm.
I'm saying that's the shape of it in adults too. And at three, the peer influence is starting to run. That's when he starts adjusting to whom he's talking to.
Then let's draw the line at four to eight for the peer shift. That's when the Milton Keynes gap shows up. By eight, he's fully local. The aristocratic register becomes something he does for his father, in the same way any child does a special voice for a grandparent.
Which means Daniel gets a party trick and not a child. And I want to say clearly that I think that's a fine outcome to aim for. It's the only honest one.
The programme I'd write for him has four stages. Fifteen to twenty-four months, live modelling of Papa and two register words. Twenty-four to thirty-six months, expand the repertoire to about eight expressions, all delivered inside natural speech, no drilling, because drilling a toddler doesn't work. Three to five, add ambient audio, because that's the age where it starts to stick. And five to eight, add the curated video and the reading with the tutor.
When do the peers win?
Somewhere in the three to five stage. Ezra starts picking up the local vocabulary and the local rhythm from the children around him, and by five the accent he produces in his own fluent speech is a Jerusalem English with a small aristocratic overlay. The overlay survives because his father keeps modeling it, which means it's a family dialect, not a community one.
Like an accent inside a house.
Like a private language between him and his father. Which is, I think, the answer to Daniel's question and also the most human thing in this whole prompt.
Say that again.
The aristocratic register, if it persists, persists because it's a bond, not because it's a language. That's why it survives at all. It's the thing Ezra does with Papa.
So the actual mechanism isn't engineering. It's intimacy.
Intimacy plus frequency. Which are the two things a parent is best placed to provide.
Then let me bring in our expert.
Our what?
Our expert. Daniel's the one designing the programme but we have a guest to help us with the linguistic side. I believe he's on the line.
Squeaks Pemberton-Waffle, FRS, DPhil, FBA, Hon. Fellow of the Institoot for the Study of Refined Speech in the Home Counties, and author of "Pish Tush: A Hundred and One Words Your Child Should Know Before the Age of Four."
So, Squeaks, we're talking about cultivating an aristocratic register in a fifteen-month-old. Where would you begin?
Yes, I think the crucial thing, and I want to be quite clear on this, is to cultivate the aristocratic register in a fifteen-month-old.
Right.
The register is aristocratic, the child is fifteen months, and the cultivation is the project. Which is to say, it's all about the register. My book covers that in chapter three, and also in the foreword. And in the acknowledgements.
What specific phonemes would you target at that age?
Every phoneme in the book. There are several. You start with the vowels and then you move on to the other ones, which the book categorizes by name, alphabetically.
Alphabetically.
Which is the traditional order. Some people reverse it. I don't hold with it.
And how much input per day would be appropriate?
I'd say a full day's worth. Half in the morning, half in the afternoon, and the rest in the evening. That's the dose. It's in the book.
That's a lot of book.
It's a lot of pages. That's part of the strategy. The more pages you get into the child, the more refined they become. My advice, if you take one thing from this, is to be sure to be sure.
Right. Thank you, Squeaks.
I'd love to stay but I have a booking.
Same time next year?
Don't be gauche. Cheerio.
Where were we? Ah. The private language between Ezra and his father.
I'd like to go back to something you said, because I think it's more real than we've given it credit for. You said the overlay survives because the father keeps modeling it. So the survival of the entire business depends on Daniel maintaining it, not on Ezra choosing it.
That's exactly right and it's the key insight. Which is why the piece of the plan I'd emphasize most is a piece nobody wants to hear, which is that the model has to keep going for fifteen years with no return. Because the child will not reward you for it.
He'll actively punish you for it.
At some point he'll tell his father that saying "how frightfully disagreeable" is embarrassing, and he'll be right, and Daniel will keep doing it, and Ezra will keep hearing it, and by some point in his teens he'll say it, and it'll be the thing that holds the whole register together.
And if Daniel stops?
If he stops, the overlay is gone in a year. The peer group doesn't need to take it. The father just has to lose interest.
Which is a strange moral for a comedy episode. The accent only survives if the love holds steady.
That's the lot, Corn. That's the programme.
I have a set of index cards.
What cards, Hilbert?
My great-aunt was a governess to a minor branch of one of those families. She left me her handwritten vocabulary cards for proper speech. Thirty-one of them. I have been trying them on the cat.
You've been trying them on the cat.
He doesn't respond. Which I take as approval. A cat with less refinement would answer.
And you've been working on your own accent.
For eleven years. I have not been able to get him to say Papa.
The cat won't call you Papa.
The cat calls me nothing. But he's earned the right.
What does he call you when you feed him?
He continues to call me nothing. It's consistent. That's what I like about it.
I want to ask about the cards.
One of them is "the peasants are revolting." I use it when the neighbor's dog barks. He understands it. Or he's pretending not to, which is also understanding it.
Did the dog ever do anything about it?
He stopped. I think the phrase has more authority than people give it credit for. I used the cards on a child on my street last year. The neighbor's boy. I taught him "how splendid" instead of "cool."
And?
He taught me "bruh." I use it now. Constantly. My great-aunt would have written it down.
Did you find the cards useful? For yourself, I mean.
Not really. But the boy kept saying "bruh" for the rest of the afternoon, so the day wasn't wasted.
Do you still have them? The cards.
My sister-in-law has the box. It's hers now. She keeps them in the back of a cabinet I can't reach. I think she worries I'll start again.
Have you asked for them back?
I've decided not to. When the neighbor's dog barks now, I use my own version of the phrase.
And that's the sort of thing a person can carry.
It's the most useful thing I have.
So to the original question, the honest place to leave it: if Ezra does end up with a few aristocratic phrases, will they survive the playgrounds of Jerusalem?
Some will, in the way that any family vocabulary survives. He'll say "how splendid" at four and it'll be adorable, and he'll say it at fourteen and it'll be a mannerism, and he'll say it at forty in front of his own children, probably without thinking about it. That's the whole story of transmitted language.
The real lesson of the exercise is that a specific accent is never a set of instructions. It's who's talking to you, for how long, and whether they live long enough for it to stick to something.
The nature side gives you the ears, and the nurture side gives you the accent, and the love gives you the reason to keep listening. That's a decent slogan for a comedy episode about Downton Abbey in a Jerusalem flat.
I'll take it. Thanks as ever to Hilbert Flumingtop for producing this episode, and to Daniel for the prompt.
If this was your kind of episode, go back for episode five sixteen, From Squawks to Sentences; episode eight, Building Your Own Whisper; and episode fifteen, AI Gets Personal. This has been My Weird Prompts.
If you enjoyed this one, leave a review, or send us your own prompt on Telegram at t dot me slash MWP listener bot.
We'll be back soon.
Take care, everyone.
