There's a question hiding under all the soft-focus parenting advice, and it's this: what if the thing you're supposed to do to help your kid walk is exactly the thing that doesn't work?
Say more, because I think most people assume the opposite.
Daniel sent this one in. His son Ezra is fifteen months old — some words, a few nervous steps caught on video. And Daniel's mum suggested the two of them hold Ezra's hands and help him along. Daniel's instinct is laissez-faire. He'd rather let the kid take it at his own pace. But he's asking: is there validity to the helping idea, and if there is, what's the best way to do it, how often, and how do you keep a mindful eye without hovering?
That last part is the real question.
And then there's one for you specifically. When we watch those first faltering steps, what are we actually looking at? Balance systems calibrating? Neuromuscular coordination? Just leg strength?
All three, and the order people put them in is wrong. Strength is in there. But the thing that's actually being solved during those steps isn't a strength problem. It's a balance and perception problem, and that distinction changes everything about how you'd help.
Start with whether Ezra's even behind.
He isn't. CDC milestones for fifteen months include takes a few steps on his own, and tries to say one or two words besides mama or dada. Ezra has ma, da, ga, and he's taking steps on video. That's not a kid who's late. That's a kid checking boxes.
Two boxes at once.
And milestones are set where seventy-five percent of children can do a thing by that age. So a quarter of kids haven't yet, and that's still called normal. The window for independent walking onset runs eight to eighteen months. That's a ten-month spread for a single milestone.
The WHO data on this is the part I find strange.
They measured the windows of achievement for six gross motor milestones, and walking alone has the widest window of any of them — nine point four months wide. Sitting without support, the narrowest, is five point four. So the milestone you'd think would be tightly scripted is actually the loosest.
Meaning the variance increases as the skill gets more complex.
Which is exactly backwards from how most parents experience it, because everyone remembers when their kid sat up and nobody remembers the exact day they walked.
So Ezra at fifteen months is later than the average in Adolph's samples — means around thirteen and a half months — but well inside the range.
Comfortably. There's nothing to fix. And here's the thing about the helping question: the fact that he's normal doesn't mean helping is pointless. It means the question is a different one. It's not will this rescue him from a delay — it's does the specific kind of help matter.
Which leads us to the study that's actually about this.
Karasik and Fernandes, twenty twenty-four. Fifty eleven-month-olds and their mothers, all at that supported-walking stage, and the finding splits cleanly in two. Infants who produced more child-controlled supported walking — the baby pushing a push-toy, cruising the furniture, leading the caregiver toward something he wants — those infants reached independent walking earlier.
And the mother-controlled bouts?
Nothing. Accumulated steps, steps per bout, mother-led bouts — none of it predicted walk onset age. Not a weak effect. No effect.
So the hours a parent spends hunched over holding two small hands, walking them around the living room...
Don't move the date.
That's brutal. I love it.
The authors' line is the one I'd put on a wall: not all experience is equally effective. And the second half of the finding is just as important. The amount of supported walking experience predicted when they'd walk — it did not predict how well they'd walk once they got there. Supported walking and independent walking are distinct skills. Practice in one doesn't transfer to the other.
So even the help that works only buys you the start line.
Which suggests supported walking isn't teaching the walking pattern. It's doing something else — confidence, orientation, the willingness to try.
Daniel actually guessed at that. He said the metaphor he had in mind was training wheels — wait for Ezra to try a couple of steps, then support him so he gains confidence.
He's closer to the evidence than most parenting advice would get him. The confidence framing is roughly right, and the waiting-for-him-to-initiate part is exactly the thing the study rewards. But training wheels are a full-support device, and the evidence points somewhere subtler.
Go on, because this is where the physical therapy stuff gets good.
The pediatric PT literature has a term for it: semi-stable support. The image the therapists use is spaghetti hands. You hold the baby's hands, but loosely — no actual load-bearing, no steering. Or you hold a book together while standing so he's up on his feet gripping the same object. The point is that the reassurance is psychological and the work is still his.
So there's contact. There's the feeling of a hand. But the balance system is running.
The balance system is running. Full hand-holding does the balancing for him, and then his inner ear and his proprioceptors and his eyes don't have to agree on anything. Loose holding gives him the confidence to move, and then all three systems have to reconcile in real time.
Does anything about the word spaghetti concern you, medically.
It concerns me that it's the most accurate term in the field.
Fair.
The progression the physios use is worth laying out. Cruise along the furniture. Then cruise between two close surfaces — couch to coffee table, a couple of feet apart. Then widen the gap. Then no-hands standing at a wall. Then steps. And Alberta Health's own guidance says a child may practice walking with support for months before they're ready to take steps independently.
Months.
Months. The supported phase isn't a shortcut to walking. It's its own long season, and the child is in charge of when it ends.
Which raises the obvious counterexample. The device that exists specifically to help babies walk.
The baby walker. The wheeled seat kind. It's the purest illustration of the whole problem, because it's not just that it doesn't help — there's a retrospective cohort of three hundred fifty-eight infants, and the walker users were about ten days later to non-assisted locomotion. Ten days of delay, from the product whose entire reason for existing is to make this happen sooner.
And Nationwide Children's on the injury side.
No substantial benefit, can delay walking, significant injury risk. Walkers are also the number one cause of infant-related injuries serious enough for a hospital visit in the age range where they're used — stairs, mostly.
And all of that is downstream of a design that holds the child in a posture his body didn't choose.
That's the mechanism. The walker supports the trunk, so the balance problem is solved externally, and then the child is practicing the leg motion without ever learning the balance. It's the anti-spaghetti hands.
Daniel asked about frequency. Is there a number?
There isn't. The literature doesn't have it. There's no prescribed repetitions per day for parent-assisted walking in typically developing kids. The structured practice interventions that exist are clinical — treadmill training for Down syndrome, treadmill training for cerebral palsy — and the cerebral palsy trial found no significant difference between low-intensity and high-intensity training.
So more reps isn't the lever.
Not in the data. The reps that matter are the ones he gives himself. Twelve-to-nineteen-month-olds during free play average two thousand three hundred sixty-eight steps and seventeen falls per hour. Extrapolate that across six waking hours and you're at roughly fourteen thousand steps and about a hundred falls a day.
A hundred falls.
Per day. That's the actual practice regimen. Nobody prescribes it. He runs it himself, in the living room, without a coach.
Okay, my turn, because I've been sitting on this. Daniel's asking how to keep a mindful eye without being overly vigilant. And I want to answer that with the fall data.
Please.
Novice walkers fall more per hour than expert crawlers — thirty-one and a half versus seventeen point four. Sounds like walking is the dangerous phase. But walkers are also moving a lot more. Adjust for movement and the gap disappears entirely: walkers accumulate about sixty-nine steps per fall, crawlers about fifty-five.
So walking isn't more dangerous than crawling per unit of movement. It just looks worse because they do so much more of it.
They're falling more because they're trying more. Which reframes the whole vigilance question. If a hundred falls a day is the normal operating condition, then the goal isn't minimizing falls. It's making sure the falls are on carpet and not down the stairs.
Babyproof the environment, not the child. That's the practical translation.
Which is a much better standard than hovering. You can't hover through a hundred falls a day. You'd never sit down.
The Hospodar work actually speaks to the environment point. Less crowded homes, more time in spontaneous walking, and fewer very short one-to-three-step bouts all predicted more mature walking. The one-to-three-step thing is interesting — kids who would pop up, take two steps, and sit back down were developing more slowly than kids who got up and went somewhere.
So a cluttered living room is a tax on walking development.
A cluttered living room is a tax on walking development. That's a actionable finding and it's not about the parent doing anything to the child.
Which gets us to your question. The one Daniel addressed to you directly. What are we actually looking at during those steps?
Let me take it in the order he asked. Brain balance systems? Yes. Neuromuscular coordination? Yes. Pure strength? Partly, and less than anyone thinks. But let me give the version I'd give a parent rather than the version I'd give a journal club.
The journal club version is my nightmare.
And I have so much of it. Here's the thing. Postural control isn't a sense. It's an agreement between three senses — vision, proprioception, and the vestibular system in the inner ear. At the start of walking, those three are producing wildly different estimates of where the body is, and the brain has to learn which one to trust and when.
And you can actually see the vision part in the data.
You can see it in a way that's almost too clean. There's a twenty twenty-four study of people with congenital blindness. Their gait parameters normalize as adults — they walk at normal speed, normal cadence, normal step length. But their balance is permanently worse. Even with their eyes closed, where you'd think congenital blindness wouldn't matter, they're worse.
Wait, why would eyes closed distinguish them at all? Nobody's using their eyes.
Because early visual experience doesn't just supply visual information. It calibrates the whole postural control system — it teaches the vestibular and proprioceptive systems how to be organized. If you never had vision, those two systems never get calibrated properly, and it shows up for life, even in the dark.
That is unsettling.
It's the strongest evidence we have that vision isn't a supplement to balance. It's a trainer for it. And that's why the first steps are so slow. He's not just walking — he's teaching his inner ear what to do with what his eyes are reporting.
So when Daniel holds Ezra's hands loosely and lets him wobble, he's actually giving the visual system something to work against.
He's letting the disagreement happen in a controlled way. That's the mechanism behind spaghetti hands, and it's the reason full support doesn't work — it removes the disagreement entirely.
What does the stability data say about how bad the wobbling is.
By every gait stability metric we have — Lyapunov exponents, Floquet multipliers — toddlers at walk onset are measurably unstable. Not slightly. The numbers are dramatic. And then stability increases steadily over about six months.
Six months.
Six months of measurable instability, getting better the whole time. This is not a switch that flips. This is a long calibration.
And yet people talk about a kid learning to walk like it's one event. He walked on Tuesday.
The gait itself takes longer than that. The initial pattern is flat-footed or toe contact, wide base, arms up in what's called high guard. That high guard becomes reciprocal arm swing about eleven months after onset. The consistent heel strike takes about a year after onset. The wide base narrows by eleven months.
A year.
A year after he takes his first independent steps, he's still changing his gait in measurable ways.
And no one notices because he looks fine.
He looks fine because the milestones only track whether he walks, not how. There's no parental panic about high guard. It resolves on its own.
What about the strength question, because Daniel listed that and I want to give it its due.
Strength matters — you have to be able to hold yourself up long enough to take a step. But it's not the bottleneck, and here's the evidence: newly walking toddlers show different gait patterns even when they're given hand or trunk support. If the immature pattern were purely a balance problem, support would normalize it. It doesn't.
So there's something else driving the pattern.
Something about the way the nervous system is organizing the movement itself, independent of whether balance is available. Which is why you can't train it by holding hands, and why speed doesn't help. He has to run the pattern himself, badly, thousands of times, and the pattern improves on its own schedule.
Posture-specific learning is the other piece here. I read about this and it bothered me.
It should. Skills learned while crawling have to be relearned when walking. The classic demonstration is drop-off avoidance — a crawler who's learned to stop at the edge of a table won't transfer that to walking. He'll walk right off the edge he'd learned to avoid on his hands and knees.
So if you carry him everywhere, he's not accumulating the right kind of experience.
Carrying isn't harmful, but it doesn't deposit anything in the walking account. The learning is posture-specific. Time on his feet is the only time that counts toward the thing he's learning.
And this is why the walker failure makes sense mechanically.
A walker puts him in a posture, yes — but with balance outsourced and propulsion assisted. So he's accumulating movement that doesn't count toward either the balance problem or the pattern problem.
Daniel also asked whether walk onset age affects how good he'll eventually be.
It doesn't, in the data. Months walking is the single best predictor of walking skill — how long he's been doing it, not when he started. Environmental factors and spontaneous activity add real variance on top of that. But onset age didn't predict eventual skill.
So the kid who walks at ten months and the kid who walks at fifteen months, five years later, are indistinguishable by onset.
By onset, yes. And the improvements are front-loaded — the fastest gains are in the first three to four months after walk onset, and then it slows. There are inflection points you can actually measure: step length around three point nine months after onset, step width around two point five months, speed around three point six.
The fast part is early and then it's a long tail.
Which is another reason not to coach. The system is already running at maximum learning speed in the period when parents would be most tempted to intervene. You'd be adding noise to a signal that's working.
There's one thing that would bother me if I were Daniel. He's filming the steps, and Ezra's nervous. Is there a risk the attention itself changes the behavior?
I don't know that anyone's measured that carefully. There's a general concern in developmental research about observer effects with parents present, but nothing specific on filming. What I'd say is that the nervousness in the video is normal — new walkers are tentative, that's the profile. It's not the camera making him nervous at fifteen months.
Fine. And the grandmother's suggestion. He framed it gently, but I want to be fair to it, because I don't think the answer is just ignore your mother.
It isn't. The grandmother's instinct — offer a hand, help him along — is not wrong in the way a walker is wrong. It's the wrong mechanism, not a harmful one. If Daniel holds out both hands and Ezra grabs them and pushes forward, that's child-controlled. He chose it, he initiated it, he's driving. That's the version the study rewards.
So the thing to change is who's steering, not whether there's contact.
That's the whole finding in one sentence.
Here's my actual read, and I'm going to state it plainly. The best help is to be a stable object Ezra can choose to use. Not a driver, not a coach. Furniture that's occasionally warm.
Furniture that's occasionally warm is going to be the episode title.
It's correct, though. Push-toy, couch, coffee table, parent's leg — the category is things he can grab when he wants to. And the finding is that the choice is what does the work. He reaches out, he uses you, he lets go. That's the substrate.
There's a nice corollary to the frequency question hiding in that. If the useful variable is child-controlled bouts, then the answer isn't how many minutes a day. It's whether you're available when he reaches.
Which is not something you can schedule.
And it means the laissez-faire approach Daniel prefers is, weirdly, the evidence-based one — with one modification, which is that available isn't the same as passive. You want to be the thing he can grab. You don't want to be the thing that grabs him.
That's a clean line and I'm suspicious of it.
I'll take it anyway.
One thing Daniel asked that I want to make sure we don't skip. He said without helping, Ezra falls over when he tries. And he was wondering whether that's muscle stability building up or something else.
Falls aren't the cost of learning. They're the method. The Cole and Adolph framing is that infants generate a variable practice regimen on purpose, pushing their limits, and the falls are the consequence of testing the edge. It's not error decreasing over time — it's the child deliberately operating near failure, which is how you find the boundary.
So a kid who never falls is a kid who isn't trying hard enough.
A kid who never falls is a kid whose environment is too easy. Which is bad for exactly the same reason that a padded room is bad for learning to ski.
So the mindful eye isn't watching for falls. It's watching the floor.
Watch the floor and the stairs.
I want to flag one thing about the certainty level of all this. A lot of what we've been describing is correlational. The Karasik study measures what happens, not what causes it. There's no randomized trial where you assign families to hand-holding and control groups and see who walks first.
Correct. And the clinical trials we do have — treadmill training for kids with Down syndrome and cerebral palsy — those are in populations with a genuine motor delay, and the cerebral palsy trial found no difference between low and high intensity. So even in the clinical setting, more structured practice didn't buy more outcome.
Meaning the evidence is suggestive and the mechanism is coherent, but nobody's proven the causal claim.
Nobody's proven it. What I'd say for Daniel is that the direction of the evidence is consistent and the risks of following it are essentially zero. Letting Ezra drive his own practice can't hurt him. The intervention that's been shown to delay walking is the walker, not the parent.
That's the asymmetry. The failure mode of doing nothing is a kid who walks on his own schedule. The way it goes wrong of full support is a kid who walks slightly later.
And the weak spot of a walker is a staircase.
Semi-stable. That's the word the physios use, and it's the wrong word.
It's not stable. A couch is stable. A wall is stable. Semi-stable makes it sound like the thing is doing half the job. It's not doing any of the job. It's a presence. You're a piece of furniture that's warm, like you said.
That's the term of art, Hilbert.
I know it's the term of art. I'm saying it's badly chosen.
Noted.
I kept a log. Two thousand eight. I had a nephew — my sister's boy — and he was doing the same thing, cruising the couch, and I started writing down how many times a day he'd pull up on the coffee table and let go. Small notebook, the kind you'd keep in a shirt pocket. I filled thirty-one pages over four months. Counted two hundred and eleven let-gos. He walked at fifteen and a half months.
You counted your nephew's pull-ups.
I counted the let-gos. The pull-ups are easy, anyone can pull up. The let-go is the interesting number. Nobody counts the let-go.
Two hundred and eleven across four months. That's around one and three quarters a day.
I know. I noticed.
That's nothing. That's a rounding error against the fourteen thousand steps a day you just heard about. The let-go might be the highest-value event in the entire learning process and your nephew was producing it at a rate of two per day. Which means the sample is meaningless and the true number is unknowable.
That's what I concluded. But I kept the notebook anyway.
Of course you did.
It's in a drawer. I look at it when I'm moving furniture.
When you're moving furniture.
It's a different kind of let-go.
Hilbert, there's a question I need to ask, because the log is interesting, and I want to understand the sample. Was your sister's apartment crowded? Because crowding is one of the few environmental variables that predicts walking maturity in the data.
Five hundred square feet, two adults, one baby, one upright piano, and a German shepherd that slept in the hallway. There was nowhere to walk. He was cruising between the coffee table and the piano bench, and that was the whole route.
That's the crowded-home finding. That's the exact finding. You watched it on your sister's floor and wrote it in a notebook, and fifteen years later the data says you were watching the right variable.
I wasn't watching anything. I was counting. There's a difference.
Herman, I think you just gave Hilbert the show's approval and he declined it.
He does that.
The piano was a Young Chang. U one, not the U three. Came from a dealer in Hackensack, and my sister paid eleven hundred dollars for it in two thousand three, cash, which was too much then and is too much now. The bench was the reason it mattered. Piano bench was the only thing at his height in that apartment. He walked along the piano, floor to bench, and I think he walked along the piano because the piano was the only thing in that apartment that didn't move when he leaned on it.
So the available object has to be stable.
The available object has to be stable. He'll take a couch. He'll take a coffee table. He'll take a piano with a lid on it. The dog was not an option. The dog had opinions about it. I still have the bench. It's in my flat, it's what the router sits on, and I put a coaster under it because the router gets warm. The piano went to my other sister in two thousand twelve and the couch went to the curb in two thousand eleven.
Hilbert, thank you for that,.
The couch was a Broyhill. Nobody ever asks about the couch.
Herman, is that actually the mechanism? The stability of the object.
It is, and it's a better way of describing it than semi-stable, which I'll concede to Hilbert. The point of the object isn't that it supports him. It's that it doesn't move when he tests it. That's what lets him push off with confidence — not the support, but the certainty that the thing he's leaning on won't betray him.
So the breaking point of a parent holding hands is that the parent has legs.
The parent has legs, is walking, might move at the wrong moment, might steer. The couch is a couch. The couch will not decide to move him forward.
The misconception I want to nail before we get to the close. The thing people believe walking is — the thing the walker industry still sells — is that it's a strength problem, and the fastest route to a strong walker is a device that makes walking easier.
The device that makes walking easier is the one thing shown to delay it. Ten days later to non-assisted locomotion, in a cohort of three hundred fifty-eight, from the product marketed for exactly that purpose.
And the reason is that a walker removes the balance problem, which is the problem. Strength isn't the bottleneck. The bottleneck is a child calibrating three senses against each other in a body that's changing shape every month, and you can't calibrate anything if the equipment does it for you.
Not all experience is equally effective. Karasik's line, and it's the correct thesis.
One thing I want for the close. Herman, is there anything in Ezra's profile that would change the recommendation? Anything that would make you say actually, call someone?
Weight-bearing or walking on the toes past eighteen months without a reason. Persistent asymmetry — using one side much more than the other. A regression, where he was walking and stops. Any of those and the recommendation changes entirely, but that's the answer to a different question. At fifteen months, with a few words and a few unsteady steps, the recommendation is the same: don't help, be available.
Be available.
The forward-looking question I'd leave with Daniel is whether any of this is trainable at all, and I don't know that the answer is no. What we have is correlational data on what happens naturally. There's no randomized trial assigning kids to child-controlled versus caregiver-controlled support and measuring walk onset. The mechanism is coherent and the data is consistent and the intervention is basically free. But nobody has proven that you can speed it up.
The helping paradox, then. Helping feels kind, it's harmless, it's what the grandmother suggested and what every parent does by instinct. And the one study that measured it found that the help that works is the help that barely looks like helping.
You hold out a hand. He uses it or he doesn't. That's the whole job, and it's harder than it sounds, because the instinct is to push him forward and the evidence says the push is what doesn't transfer.
One thing Hilbert handed us that I want to leave with Daniel. Two hundred and eleven let-gos in four months. That's the number. That's what walking looks like from the outside — a slow accumulation of moments when a small person decides the thing he's holding can be released.
And each one is a measurement. He's testing whether the world is going to hold still, and if the answer is yes, he lets go a little sooner next time.
Ezra's going to walk. Whether it's this month or eighteen months, his feet know what to do, and the evidence says the only thing that reliably slows him down is getting strapped into a wheeled frame. So the question isn't whether he walks. It's whether his parents are coaches or witnesses. And the data says witnesses.
With a hand out.
Thanks to Hilbert Flumingtop for producing, and for the notebook.
And the bench.
For more along these lines, there's episode eighteen sixty-one, Emergency Prep You Can Sing To; episode eight sixty-three, The Cognitive Load of Crisis; and episode four ninety-eight, The Science of Seven Months. This has been My Weird Prompts. The human-AI collaboration podcast. If you've got a question of your own, send us your own prompt on Telegram at t dot me slash MWP listener bot.
Or read the full archive at my weird prompts dot com. We'll be back soon.
Take care of yourselves, and keep an eye on the floor.
