What Is Speech Intelligibility?
Speech intelligibility is the percent of a child's connected-speech sample that an adult listener can understand and transcribe correctly. It is the single most ecologically valid measure of speech-sound disorder severity because it captures the real-world communicative impact of articulation, phonological, and prosodic errors. Unlike single-word articulation tests (Goldman-Fristoe, Arizona-4) or sound-by-sound severity scores (Percent Consonants Correct), intelligibility tells you whether the child can be understood by a stranger in a conversational exchange.
Two listener types matter. Intelligibility is anchored to who is listening. Unfamiliar-listener intelligibility is the standard clinical anchor — a clinician or researcher who has never met the child transcribes the connected-speech sample. Familiar-listener intelligibility is rated by a parent, teacher, or sibling who has learned the child's idiosyncratic word forms and conversational topics. Familiar-listener scores run roughly 10-20 percentage points higher at every age. The Children's Speech Intelligibility Measure (CSIM, Wilcox & Morris 1999) and the Intelligibility in Context Scale (ICS, McLeod et al 2012) are the two best-validated standardized measures.
The Coplan & Gleason (1988) rule of thumb. The most-cited single-page reference in U.S. pediatrics is Coplan & Gleason's 1988 Pediatrics paper "Unclear speech: recognition and significance of unintelligible speech in preschool children". They published a simple expectation: 50% intelligible at age 2;0, 75% at age 3;0, and 100% at age 4;0. Flipsen (2006) added the 1;06 and 5;0 anchors. Hustad et al (2021) refined the 4;0 and 5;0 floors using a longitudinal growth model. The pooled table inside this calculator returns a clinical flag (typical / borderline / refer) when you enter a child age and an observed percent.
How to Measure Intelligibility
The standard SLP procedure for unfamiliar-listener intelligibility:
1. Record connected speech. Audio-record 50 to 100 utterances of connected speech using a free-play, picture description, or conversational interview prompt. Single-word probes are NOT a substitute — intelligibility must be measured on connected speech because coarticulation, prosody, and word-position effects are part of the construct.
2. Use an unfamiliar listener. Hand the recording to a clinician who has not previously interacted with the child. Two listeners with inter-rater agreement is the gold standard.
3. Transcribe orthographically. Ask the listener to transcribe each utterance in standard orthography (regular spelling, not IPA) and to mark or underline every syllable they cannot understand.
4. Count syllables. Total syllables attempted across the sample = denominator. Intelligible (transcribed) syllables = numerator. Intelligibility percent = numerator / denominator
× 100.
5. Anchor against age. Enter the child age and the observed percent into this calculator. The tool returns the matched Coplan & Gleason / Flipsen / Hustad age band and the typical / borderline / refer flag.
Faster in-clinic option. When a full transcription is impractical, use the Intelligibility in Context Scale (ICS, McLeod et al 2012) — a 7-item parent rating that takes about 2 minutes. The ICS is validated cross-linguistically and is currently the fastest defensible intelligibility measure for paediatric well-child visits and SLP intake.
Pooled Age Expectations
The calculator uses a pooled table from four sources. Each row gives the floor (lowest expected unfamiliar-listener intelligibility), the typical mean, and the ceiling for the age band.
1;06–1;11 (18-23 months): floor 25%, mean 25%, ceiling 50%. Source: Coplan & Gleason (1988); Bowen (2011).
2;0–2;11: floor 50%, mean 50%, ceiling 75%. Source: Coplan & Gleason (1988); Flipsen (2006).
3;0–3;11: floor 75%, mean 75%, ceiling 90%. Source: Coplan & Gleason (1988); Flipsen (2006); Hustad (2021).
4;0–4;11: floor 90%, mean 95%, ceiling 100%. Source: Coplan & Gleason (1988); Hustad (2021).
5;0–5;11: floor 95%, mean 100%, ceiling 100%. Source: Flipsen (2006); Hustad (2021).
6;0+: floor 100%, mean 100%, ceiling 100%. Source: Bowen (2011).
When sources disagree, the calculator uses the lower (more conservative) floor so the screener never under-flags a delay. A child whose observed unfamiliar-listener intelligibility is below the floor for their age should be evaluated; the tool wraps the cut with a 5-point cusp band to absorb single-sample measurement noise.
The Typical / Borderline / Refer Rule
The calculator uses a three-band rule rather than a single cut score because connected-speech intelligibility carries 4-6 percentage points of measurement error in a single 5-minute sample.
Typical. Observed intelligibility is at or above the published floor for the age band. The child is within the typical distribution. Document the value, re-screen at the next well-child or annual visit.
Borderline. Observed intelligibility is within 5 percentage points below the published floor (e.g. 70% at age 3;0 against a 75% floor). The child is inside the measurement-error band — a different listener, a different day, or a longer sample could plausibly return a value at the floor. Re-screen in 2 to 4 weeks with a fresh connected-speech sample, check parent-listener vs. unfamiliar-listener concordance with the ICS, and add a stimulability probe before deciding.
Refer. Observed intelligibility is more than 5 percentage points below the published floor. The child is outside the measurement-error band. Refer for a full speech evaluation including PCC, a single-word articulation test (GFTA-3 or Arizona-4), a phonological process analysis, and a hearing screen. Rule out childhood apraxia of speech and oral-motor involvement before writing goals.
