Your fifteen-month-old used to wave when you left for work. Somewhere in the last month, without you noticing exactly when, she stopped. You mention it to your husband over dinner and neither of you can remember the last time she did it.
That's usually where this starts. Not with a doctor pointing at a symptom, but with a parent noticing an absence. Nobody trains you to watch for what a baby stops doing. You're wired to notice what she does.
If you're here because something similar has been nagging at you, this article is for the two questions underneath that feeling: what does the current research actually say counts as an early sign, and what changed recently that's worth knowing before your next pediatrician visit.
Why the timing matters more than you'd think
About 1 in 31 eight-year-olds in the U.S. is now identified with autism, according to the CDC's most recent surveillance report. We've written before about why that number keeps climbing and why it isn't evidence of an "epidemic," it mostly reflects better identification reaching kids who used to be missed entirely.
Here's the part that matters for this article specifically. The same CDC data puts the median age of diagnosis at 47 months, just under four years old, and only about half of children are evaluated by their third birthday. Most of the behaviors this article covers are visible well before then, often in the first eighteen months of life. The gap between when a sign shows up and when a diagnosis happens is, on average, close to two years. That gap is what early screening is trying to close, and it's the whole reason this list exists.
What actually shows up, and when
Autism doesn't announce itself with one clear symptom at one clean age. It shows up as a cluster of small differences that build on each other, and the cluster looks different in every child. Still, research on infant siblings of autistic children (who carry a higher likelihood of being autistic themselves) has given researchers an unusually clear window into what tends to appear first, and when.
By six to nine months, the earliest signals are rarely social. A study out of the University of Missouri's Thompson Center, published in 2025, found that infant temperament at nine months predicted autism risk at age one. The researchers weren't looking at eye contact or babbling. They asked parents about frequent crying, irritability, how hard the baby was to calm down, and how well the baby adjusted to new situations. Infants who struggled more with self-regulation at nine months were more likely to show autism traits a year later. This is a genuinely new finding, and it matters because it means some of the earliest clues aren't in how a baby relates to people. They're in how a baby's nervous system handles everyday stress.
By nine to twelve months, joint attention becomes the thing to watch. This is the skill of sharing focus with another person on the same object, and it's one of the most consistently replicated early predictors in the research. A baby who points at a dog and then looks back at you to check that you saw it too is doing something more sophisticated than it looks. A baby who doesn't respond to their name by twelve months, who rarely makes eye contact during ordinary moments like feeding, or who doesn't seem to try to share excitement with you, is showing an early version of the same pattern.
By twelve to eighteen months, communication differences usually become harder to miss. Most babies wave, point, and reach to be picked up well before their first birthday. An autistic toddler may skip these gestures almost entirely, sometimes using an adult's hand as a tool instead, physically pulling you toward what they want rather than pointing or asking. Some children develop echolalia, repeating phrases they've heard without using them to communicate a need of their own. It's worth saying clearly that some echoing is completely typical at this age. The distinction researchers care about is whether a child has any other flexible way to communicate underneath the echoing, not whether echoing happens at all.
By eighteen to twenty-four months, play and language patterns start to separate more clearly from typical development. Pretend play, feeding a doll, talking on a toy phone, usually shows up by now, and its absence is one of the more reliable flags. This is also the window where regression, the loss of skills a child already had, most often appears.
The regression window, and what recent research says about it
Regression deserves its own section because it scares parents more than almost anything else on this list, and because the research on it has gotten more precise.
Documented loss of skills, most often language, shows up in somewhere between one in five and one in three children later diagnosed with autism, depending on how the study defines regression and whether it relies on parent recall or prospective video review. It typically happens between fifteen and thirty months. The newer and more interesting finding is that regression usually isn't as sudden as it looks from the outside. Prospective studies that film high-likelihood infants every few months, rather than asking parents to remember months later, keep finding a subtler dip in development that precedes the obvious loss, sometimes by weeks. What looks like a switch flipping is often closer to a slope that was already tilting.
One thing worth saying plainly, because it still comes up: there is no credible evidence linking vaccines to autism or to regression. Multiple large, independent studies have looked for that connection and found nothing. If regression happens around the same time as a vaccination, it's a coincidence of timing, not cause and effect, and it's worth not letting that particular worry crowd out the more useful question of when to get evaluated.
The screening tools got sharper, and one got genuinely new
The American Academy of Pediatrics recommends general developmental screening at 9, 18, and 30 months, with autism-specific screening built into the 18 and 24 month visits. Most pediatricians use a short parent questionnaire called the M-CHAT-R/F. It asks about the behaviors described above: pointing, response to name, interest in other children, pretend play.
Here's the part that doesn't get mentioned enough. The tool is only as good as how it's used. When a child screens positive, the protocol calls for a follow-up phone interview before referring out, and that follow-up step roughly triples the tool's accuracy. Skip it, which happens often in busy practices, and a meaningful share of positive screens turn out to be false alarms, while some children who screen negative are later diagnosed anyway. None of this means the tool is useless. It means a single screening result, in either direction, is a starting point for a conversation, not a verdict.
The genuinely new development is objective measurement. In 2023, the FDA cleared EarliPoint Evaluation, an eye-tracking device that measures how a toddler's gaze moves during short video scenes of social interaction, comparing it against what's typical for their age. A trained clinician's full behavioral assessment can take three to four hours. EarliPoint compresses the same underlying information into about ten minutes of watching videos. It's currently used in specialty centers for children aged sixteen to thirty months, not yet a standard part of a routine pediatric visit, but it's the first time autism assessment has had a genuinely objective, machine-measured component alongside clinical judgment. Where this technology lands in five years is an open question. Where it stands right now is promising and still limited to specialists.
Girls don't always fit the pattern everyone's watching for
Most of the checklist above was built from research samples that were mostly boys, and that has consequences worth naming.
A large Swedish study published in BMJ in early 2026, tracking more than 2.7 million people born between 1985 and 2020, found that the gap between how often boys and girls are diagnosed narrows sharply with age. By early adulthood, men and women are being diagnosed at similar rates. The researchers' interpretation is straightforward: autism likely occurs close to equally in both sexes, and girls are disproportionately missed or diagnosed years later, not less often affected in the first place.
At toddler age, this can look like better eye contact, more socially acceptable interests such as animals or dolls instead of mechanical objects, and a stronger early ability to copy the social behavior of other children, which can mask underlying differences from a checklist built around how boys tend to present. None of this means girls are unaffected. It means a toddler girl who seems unusually socially observant, rather than withdrawn, is not automatically ruled out, and it's worth saying that directly to a pediatrician if something still feels off despite a clean-looking screen.
Help doesn't have to wait for a diagnosis
This is the part parents are least likely to hear from anyone but a specialist, so it's worth stating clearly. A randomized clinical trial published in JAMA Pediatrics gave parent-mediated intervention to infants showing early signs, before any formal diagnosis had been made, and found measurable improvement in outcomes by the time those children reached diagnostic age. Since then, several smaller trials have tested similar caregiver-coaching programs with infants as young as six months who show elevated likelihood, with consistent results pointing the same direction: age at the start of support and how consistently it happens matter more than which specific program a family uses.
In practical terms, this means you don't need a diagnosis in hand to start getting help. In the U.S., every state runs an Early Intervention program for children under three, and in most states a parent can request a free evaluation directly, without a pediatrician's referral or a formal autism diagnosis. If your child qualifies for speech, occupational, or developmental therapy, that support can begin while the longer diagnostic process is still underway.
What to actually do if you're noticing these things
Write down what you're seeing, with dates if you can manage it, and take short videos on your phone when the behavior happens naturally. This turns a vague feeling into something specific a clinician can actually evaluate, rather than a memory you're reconstructing under pressure in a fifteen-minute appointment.
Bring it up at the next well-child visit, or sooner if you don't want to wait. You can request a developmental screening at any pediatric visit, not only the ones officially scheduled for it. If your pediatrician doesn't share your concern and you can't shake it, a second opinion or a direct referral to a developmental pediatrician is a reasonable thing to ask for. You know your child in a way a fifteen-minute appointment can't fully capture.
Contact your local Early Intervention program directly if your child is under three. You don't need to wait for anyone's permission to make that call.
And hold the whole list loosely. One item on this page, on its own, usually means nothing. A cluster of them, persisting over weeks rather than days, across more than one category, is what actually moves the needle for an evaluation. Toddlers are inconsistent by nature. The pattern matters more than any single afternoon.
If you're just starting to sit with the possibility of a diagnosis, we've also written a plain answer to what autism spectrum disorder actually is, for the moment right after this one.
You don't have to figure out tonight whether this is autism. That's not your job. Your job, if you're noticing something, is just to say it out loud to the right person and let them help you find out. Most parents who end up here already know something, and what they need isn't more certainty about the checklist. It's permission to trust what they've already seen.
This article is for general understanding and isn't a substitute for an evaluation from a pediatrician, developmental pediatrician, or psychologist. If you have concerns about your child's development, they're worth raising with one directly.
Sources:
- CDC Autism and Developmental Disabilities Monitoring (ADDM) Network, 2025 surveillance report (MMWR Surveillance Summaries)
- Andres, E. and Sheinkopf, S., University of Missouri Thompson Center for Autism and Neurodevelopment, 2025
- Wieckowski et al., "Sensitivity and Specificity of the Modified Checklist for Autism in Toddlers (Original and Revised): A Systematic Review and Meta-Analysis," JAMA Pediatrics, 2023
- Whitehouse et al., "Effect of Preemptive Intervention on Developmental Outcomes Among Infants Showing Early Signs of Autism: A Randomized Clinical Trial of Outcomes to Diagnosis," JAMA Pediatrics, 2021
- EarliTec Diagnostics, FDA 510(k) clearance for EarliPoint Evaluation, 2023
- Fyfe et al., "Time Trends in the Male to Female Ratio for Autism Incidence," BMJ, 2026
This is educational content, not clinical advice, and does not establish a clinical relationship.
