The cliff at three: what happens when early intervention ends

Picture a family who has been living inside early intervention for a year and a half. There's a service coordinator who knows their kid. There's a therapist who comes to the house on Tuesdays and gets down on the living-room floor. There are goals everyone understands and a rhythm the whole family has built around. It works. It has become, quietly, part of how this family runs.

Then the child turns three, and almost none of it carries over.

The coordinator is gone. The Tuesday visits stop. There's a new set of people at a new office, the school district this time, asking for a new evaluation and talking about a different plan with a different name. Some families make this crossing smoothly. Plenty don't, and the ones who don't are almost never told in advance that it was coming. They find out the way you find out about a step in the dark: with their foot already in the air.

This is the birthday nobody warns you about. And the thing I most want you to take from this piece is that it isn't really one cliff. It's three, and they all tighten at the same moment: the eligibility category your child qualified under can disappear, the threshold they have to meet can move, and a legal clock runs out on the birthday itself. If you know all three are there, you can walk down instead of falling.

Two systems, one child

In the United States, help for a young child with delays comes from a single law, the Individuals with Disabilities Education Act (IDEA). But that law has two different halves, and your child ages out of one and into the other on their third birthday.

Part C is early intervention, birth to three. It is built around the family. Services usually happen where the child actually lives their life: home, a grandparent's house, childcare. The whole thing is anchored in a document called an IFSP, an Individualized Family Service Plan. In most states it isn't even run by the school district; it's run by a separate state lead agency, which is part of why the handoff feels like moving between two organizations that don't quite speak the same language. They often don't.

Part B, Section 619 is preschool special education, ages three to five. It is built around school, not the family. It's run by your local school district, and it's anchored in an IEP, an Individualized Education Program. The question it is designed to answer is not "is this child a little behind?" but "does this child need specialized instruction to access their education?" That's a different question, and it's asked by different people using different tools.

Hold onto that last row. It's where the ground gives way, and it's the first of the three cliffs.

Why zero to three is different

Before the mechanics, the science, because it's the reason the timing matters at all. If a gap in services opened up at, say, age eight, it would be a problem, but a recoverable one. At age three the stakes are different, and there's a real body of developmental research about why.

The useful distinction here is between two kinds of brain plasticity, described decades ago by William Greenough and colleagues and still holding up well. Experience-expectant plasticity is the brain waiting for the ordinary, species-typical input it "expects" every healthy human to encounter (faces, voices, language, movement) during particular windows when it is unusually ready to be shaped by them. Experience-dependent plasticity is the lifelong kind, the learning you can do at any age. The first kind is the one on a clock. The second isn't.

This is a strong-consensus idea, so I'll state it plainly: some developmental windows really are more sensitive to input than others, and a gap during a sensitive window does not cost the same as a gap later. Where I'll hedge, because the honest reading of the evidence requires it, is on the precise, dramatic version you sometimes see, the "critical window slams shut at month such-and-such" claim. Sensitive periods are real; sharp, universal deadlines with a specific date on them are mostly not. Treat the timing as a gradient of opportunity, not a trapdoor.

The move that matters most, though, is this: development is not one undifferentiated thing. Expressive language, receptive language, gross motor skills, and social communication don't behave the same way when input is interrupted. Lumping them together into a single "they'll catch up" is exactly the error that lets a real risk pass unexamined. Put concretely: a child who misses a few months of physical therapy will usually pick gross motor skills back up without lasting cost, because motor development is robust to pauses. A child who misses a few months of speech or social-communication support is missing input during the exact window when the brain is most ready to use it, and that kind of gap is harder to make up later. That is the uncomfortable through-line of this whole piece: the domains where continuity of input matters most are precisely the ones a service lapse at age three tends to interrupt.

Cliff #1: the category can disappear

Here is the first thing almost no parent sees coming. Under Part C, "developmental delay" is a required way to qualify. Every state has to offer it. A baby who is simply behind, without any specific diagnosis, gets served. That's the design.

Under Part B, the developmental-delay category for young children is optional. Federal law lets states choose whether to use it for ages three through nine, and lets a state that does use it apply it to only part of that range. Most states do use it: as of the most recent federal child-count data compiled by the ECTA Center (2020–21), 47 of the 50 states apply the category to preschoolers. Every state uses it except California, Iowa, and Texas (Puerto Rico opts out too). A local district cannot use the category unless its state has adopted it.

So if you live in one of the states that doesn't use it, the general profile that qualified your child at eighteen months ("delayed across a few areas, no specific label") may suddenly have no box to go in. To get an IEP, your child now has to fit one of the specific federal disability categories: autism, speech or language impairment, intellectual disability, other health impairment, and so on. A child who was served under a broad, forgiving umbrella at eighteen months has to map onto a specific category at thirty-six, and a "generally behind" picture doesn't always map cleanly onto any single one.

One distinction that surprises nearly every parent the first time: the school's eligibility category is an administrative label, not a medical diagnosis, and the two routinely differ. A developmental pediatrician can diagnose your child with something and the school can still determine they don't meet the educational criteria for services, or vice versa. They're answering different questions for different purposes. It's maddening the first time you hit it, and it's worth knowing about before you do.

So here is the first cliff, stated as plainly as I can: your child can lose services at three not because they caught up, but because the category they qualified under stops existing where you live.

Cliff #2: the threshold can move

Even when the category exists, the bar can sit at a different height on the other side of the birthday.

IDEA requires each state to write its own "rigorous definition" of developmental delay. There is no national number. States express the bar as a percentage delay (commonly somewhere in the 20–33% range, in one or two domains), or as standard deviations below the mean (often 1.5 or 2.0), or through informed clinical judgment. The definitions don't have to match each other, and a state's Part B bar doesn't even have to match its own Part C bar. In Illinois, for instance, the line is drawn at a 30% or greater delay; in Montana it's a 25% delay in two or more areas, or a 50% delay in one.

The scale of the variation is genuinely the story here. Nationally, about 7% of children get early-intervention services in the birth-to-two window, but the U.S. Government Accountability Office (GAO), the nonpartisan federal agency that audits how government programs actually perform, analyzed 2021 data and found that figure ran from roughly 2% of children in Arkansas to about 20% in Massachusetts. That is a tenfold spread, driven substantially by how broadly or narrowly each state writes its rules. The same report found states listing, on average, 48 different qualifying conditions, ranging from none at all to 167. Where the bar sits is close to arbitrary from a family's point of view, and it's set by geography.

The trap at the transition is subtle: a child can clear the Part C bar and miss the Part B bar in the very same state, because the two definitions can differ and because Part B is also asking an additional question: not just "is there a delay?" but "does this child need specialized instruction?" I want to be careful how I frame this, because it's easy to make it sound like a conspiracy and it isn't. Nobody is moving a bar to keep your specific child out. It's that two separately-written bars happen to sit on either side of one birthday, and your child has to clear both.

The second cliff, stated plainly: the bar your child has to clear at three can sit higher than the one they already cleared, and how high it sits depends on where you live.

Cliff #3: the calendar

The third cliff is a clock, and it's the most concrete of the three. The federal timeline, in plain terms:

• A transition conference, the meeting where the early-intervention team, the family, and the school district sit down together, must be held at least 90 days before the third birthday for a child who might be eligible for Part B.
• The transition plan has to be written into the IFSP somewhere between nine months and 90 days before the birthday. It's not a new document; it's a section of the plan you already have.
• The one that catches families: a child is entitled to a free appropriate public education (FAPE) starting the day they turn three. Which means that if the child is eligible, the IEP has to be in place by the third birthday. Which means the evaluation has to happen before it. The birthday isn't the start of the process. It's the deadline.

Two quieter calendar traps deserve their own mention, because they're where the gaps actually open:

The single-evaluation gap. Part C typically runs one formal evaluation. Part B requires "multiple measures," meaning more than one source of assessment data, so the district usually has to do additional testing before it can decide eligibility. That takes time to schedule and complete, and the clock is still the birthday. If nobody starts early, the assessment and the deadline collide.

The summer gap. A spring or summer birthday can land right in the seam of the school calendar. Early-intervention services end at three, but the preschool program may not actually start serving the child until the school year begins, and a family can experience weeks with nothing in between. And this is not a rare edge case. Births in the United States peak in late summer: in CDC birth data, August is reliably the biggest month of the year, and roughly one in four American children is born in June, July, or August. A meaningful share of all families making this transition will hit the seam. It's a known friction point, and it's one of the first things to ask about, because it's very solvable if raised early and very frustrating if discovered late.

One more wrinkle, briefly: if a child is referred to early intervention close to their third birthday, different rules apply, and if the referral comes less than 45 days before the birthday, Part C is not required to evaluate at all. The practical takeaway is simple. If you suspect a delay in a child who is already around two and a half, do not wait. Waiting can cost you the entire early-intervention on-ramp.

The third cliff, stated plainly: the law's deadline is the birthday itself, and the process only finishes on time if someone starts it months earlier. Nobody is more motivated to be that someone than you.

"Will they catch up?" The honest answer

This is the question every parent is actually asking underneath the paperwork, and a credible piece has to answer it honestly rather than reassuringly. The honest answer is: sometimes yes, sometimes no, and it depends heavily on the domain and the child.

A meaningful share of children really do age out of services because they no longer need them, and that is a success, not a loss. Here's the figure, stated precisely: in the most recent national data compiled by the Congressional Research Service, about 30% of children exiting Part C no longer required IDEA services by the time they turned three, either because they had completed what their plan set out to do or because they didn't qualify under Part B. That number blends two very different stories. Some of that 30% is a child who genuinely caught up. Some of it is a child with real, ongoing needs who happened to fall below a differently-drawn bar. The number alone can't tell you which one you're looking at.

So how do you tell? Come back to the science. The children for whom "wait and see" carries the most risk are the ones whose difficulty sits in exactly those input-sensitive domains, language and social communication, where an interruption is hardest to recover from. If your child's delays are in a domain that a pause tends to shrug off, the reassuring reading is more likely to be the true one. If they're in the domains where early continuity matters most, "they'll catch up" is a hypothesis to be tested, not an assumption to rest on. The 30% is real. So is the risk of being told you're in it when you aren't.

What to actually do: a timeline working back from the birthday

This is the part to screenshot. None of it requires you to become an expert or a lawyer. It's a handful of specific questions, asked early, in writing where you can. Working backward from the third birthday:

The bigger picture

It's worth zooming out for a second, because it takes some of the sting out of the whole thing. This gap is structural. It is not the fault of your coordinator, who is often the one working hardest to get you across, and it's not the fault of the district. It sits at the seam between two federal programs that were written separately, defined independently, and handed to different agencies to run. The seam is the problem, not the people.

And because it's structural, it's also fixable by design. Some states have started fixing it. Federal law offers a "Part C extension option" that lets a state keep providing early-intervention services past the third birthday, so a child doesn't fall between programs; a handful of states have adopted some version of it, including Maryland, Missouri, the District of Columbia, Connecticut, Colorado, and Tennessee. The details and age ranges vary, and a child in an extension always keeps the right to switch to Part B services instead. I mention it not to send you state-shopping, but to make a simple point: the cliff is a choice, not a law of nature. Where the bridge has been built, children walk across.

Which brings us back to that family on the living-room floor. The reason to understand all of this six months early is not to spend those six months afraid. It's the opposite. A family that knows the three cliffs are coming can ask the six questions, get the answers in writing, and turn the third birthday into what it should have been all along: a planned handoff, not a step in the dark. The birthday is going to come either way. The only thing you get to decide is whether it's a surprise.