Why Did My Child "Pass" Their Speech Test But Still Sound Hard to Understand?
If you've ever sat across from an SLP after an evaluation, looked at a score sheet that said your child did better than expected, and thought — "okay, but have you actually heard him talk?" — you are not imagining things. This is one of the most common points of confusion I run into with parents, and honestly, it's a completely fair thing to question.
Here's what's going on, and what the research actually says about it.
The test isn't lying — it's just only telling you part of the story
Here's a detail worth clarifying first, because I get asked about it a lot: the test I use most often, the Goldman-Fristoe Test of Articulation, actually has two parts. There's the well-known Sounds-in-Words section, where your child names pictures one at a time. But there's also a Sounds-in-Sentences section, which is also standardized and normed — your child hears a sentence and repeats it back, and the SLP scores their sounds in that more connected context.
So this isn't really a "single words vs. sentences" issue. It's something a bit more specific: both of those test sections involve your child producing speech that's been set up for them — a picture to name, a sentence to repeat. What neither one fully captures is spontaneous speech — the kind that happens when your child is genuinely thinking on the fly, choosing their own words, and talking without a script in front of them.
That distinction matters more than it sounds like it should. Repeating a sentence still allows a kind of careful, deliberate focus that isn't there once a child is talking in real, flowing conversation — coming up with what to say and how to say it at the same time. Even sentence-repetition tasks don't fully capture the sound errors that only show up when speech is truly spontaneous and connected.
In other words: testing in a structured format, even a normed sentence-repetition task, can show your child's speech ceiling — what they're capable of under more controlled conditions — not their speech floor, which is what you're actually living with at the dinner table every night.
This isn't just a clinical hunch — it's a documented gap
This gap is well known in the SLP world, not just something individual therapists notice anecdotally. Studies have specifically cautioned that structured testing — even sentence-based tasks — has real limitations when it comes to estimating how a child will actually sound in spontaneous, connected speech, and one investigation concluded plainly that word-based articulation scores don't reliably predict a child's connected speech intelligibility at all. One review of the GFTA-3 itself notes that its structured format doesn't fully capture coarticulatory errors — the way sounds shift and interact with each other — that only appear once speech becomes truly connected and spontaneous.
Why the gap? A few things are happening at once when your child moves from a structured testing task to truly spontaneous conversation:
There's less time to plan. In real conversation, your child is thinking about what to say and how to say it at the same time — there's no script to fall back on the way there is when naming a flashcard or repeating a sentence.
Sentences get longer and more varied. Spontaneous speech isn't limited to a fixed, pre-written set of test sentences — it can include far more complex sound combinations than any single test could sample.
Sound combinations behave differently in real speech. Connected, spontaneous speech can reveal sound errors and patterns that simply don't show up in more structured tasks, because sounds influence each other differently when they're bumping up against whatever words actually come next in a real sentence your child generates on their own.
So which one actually matters more?
Here's where it gets practical for you as a parent: intelligibility — whether a stranger can actually understand your child in a real conversation — is widely considered the most meaningful, real-world measure of how a speech sound disorder is actually impacting your child's life. A standardized score tells an SLP about specific sounds. Intelligibility tells me whether your child can ask for help in the school cafeteria and be understood the first time.
This is also exactly why, if you've ever felt like you understand your child just fine but other people don't, that's not a contradiction — it's a documented pattern. Familiar listeners like parents typically understand a child's speech at a meaningfully higher rate than unfamiliar listeners do, often because we've unconsciously learned our own child's unique speech patterns over time. So "I understand her" and "her teacher doesn't" can both be true, and both are valuable, real information.
What this means for your child's evaluation
This is exactly why I never rely on standardized testing alone, even when the scores look reassuring on paper — including the Sounds-in-Sentences score. A full picture of your child's speech should include:
The Sounds-in-Words section, to identify specific sound errors in a controlled format
The Sounds-in-Sentences section (if your child is 4+), which adds normed data but is still a repeated, structured task
A spontaneous, conversational speech sample — just talking and playing — to see what actually happens when your child is generating their own language in real time
A conversation with the parent, because they’re the one who's heard their child talk in a hundred situations the SLP will never get to observe in a 45-minute evaluation
If your child has ever scored "within normal limits" on a test but you still find yourself translating for them at the grocery store, that gap is worth bringing up directly with your therapist. It doesn't mean the test was wrong — it means the test only ever showed part of the picture, and the rest of that picture matters just as much.
If you've noticed this kind of gap with your own child and want a second look, I'd be happy to talk it through — reach out for a free phone consultation.
References
DuBois, E. M., & Bernthal, J. E. (1978). A comparison of three methods for obtaining articulatory responses. Journal of Speech and Hearing Disorders, 43(3), 295–305. https://doi.org/10.1044/jshd.4303.295
Flipsen, P. (2006). Measuring the intelligibility of conversational speech in children. Clinical Linguistics & Phonetics, 20(4), 303–312. https://doi.org/10.1080/02699200400026095
Johnson, C. J., Winney, B. L., & Pedersen, O. K. (1980). Single word versus connected speech articulation testing. Language, Speech, and Hearing Services in Schools, 11(3), 175–179. https://doi.org/10.1044/0161-1461.1103.175
