Intelligibility is calculated by dividing the number of words (or sounds) correctly understood by the total number of words (or sounds) produced, then multiplying by 100 to get a percentage. That core formula applies whether you’re measuring a child’s speech clarity, evaluating someone with a hearing impairment, or testing how well a public address system transmits spoken messages. The specific method you use depends on what you’re measuring and why.
The Basic Word-Level Formula
The most common approach in speech-language pathology is transcription-based intelligibility. A listener (often someone unfamiliar with the speaker) writes down what they hear, and the score is computed by taking the proportion of words identified correctly divided by the total number of intended words. If a child says 40 words in a speech sample and an unfamiliar listener correctly identifies 30 of them, the intelligibility score is 30 ÷ 40 × 100 = 75%.
This method works best with a “master transcript,” a reference copy of what the speaker was actually trying to say. The listener’s transcription is compared word by word against that reference. When multiple listeners score the same sample, their individual percentages are typically averaged to give a more reliable result.
Percentage of Consonants Correct (PCC)
PCC is a more granular measure that focuses specifically on consonant sounds rather than whole words. It’s widely used in pediatric speech assessments because consonant errors are the primary driver of reduced intelligibility in children. The formula is:
(Total consonants − consonant errors) ÷ Total consonants × 100 = PCC score
A standard imitative sentence protocol uses 36 sentences containing 273 consonants. The child repeats each sentence, and the clinician marks every consonant that was substituted, deleted, distorted, or added. If a child made 40 consonant errors, the calculation would be (273 − 40) ÷ 273 × 100 = 85.3%.
What Counts as an Error
Not every imperfect sound is scored as wrong. Only consonants are evaluated, not vowels. If the child repeats a syllable (saying “ba-balloon”), only the first production of that consonant counts. Sounds that shift due to casual speech patterns, regional dialects, or unstressed syllables are not penalized. Replacing the “ng” sound with “n” at the end of a word (saying “runnin” instead of “running”) is also not scored as an error. If an entire word is unintelligible, none of its consonants are scored at all, since you can’t reliably judge sounds you couldn’t identify.
Age Norms for Children’s Intelligibility
If you’re calculating a child’s intelligibility, the number only means something when you compare it to what’s typical for their age. Research published in the Journal of Speech, Language, and Hearing Research tracked intelligibility growth in typically developing children speaking to unfamiliar listeners and found wide variation, especially in the toddler and preschool years.
At 30 to 35 months, the average child scored about 41% intelligible on multiword utterances, with the top performers reaching around 75% and the lowest typical performers hovering near 12%. By 36 to 41 months, the average climbed to roughly 56% for multiword speech. At 42 to 47 months, the average reached about 69%.
The practical benchmark most clinicians use: by age 4, a child should be at least 50% intelligible to an unfamiliar listener. That threshold represents the floor of typical development (roughly the 5th percentile), not the average. A child scoring well below that range likely warrants further evaluation. Single-word intelligibility tends to run slightly higher than multiword intelligibility at every age, because longer utterances introduce more opportunities for sound errors and coarticulation challenges.
The Speech Intelligibility Index (SII)
In audiology and acoustics, intelligibility is measured differently. The Speech Intelligibility Index, defined by the ANSI S3.5 standard, estimates how much of a speech signal is usable to a listener given the background noise or bandwidth limitations in the environment. Rather than scoring individual words, SII quantifies the proportion of speech information that’s audible and accessible.
SII values range from 0 (no speech information available) to 1 (all speech information available). The calculation divides the speech signal into frequency bands, determines how much of the speech energy in each band exceeds the noise floor, and then weights those contributions based on how important each frequency region is for understanding speech. Low, mid, and high frequency regions are computed separately and combined into a weighted total.
SII is most useful for predicting how well someone with a specific hearing profile will understand speech in a given listening environment, or for evaluating whether a hearing aid is restoring access to enough of the speech signal.
Speech Transmission Index (STI) for Room Acoustics
If you’re calculating intelligibility for a building, public address system, or emergency communication setup, the Speech Transmission Index is the standard metric. STI measures how faithfully a room or sound system preserves the patterns in speech that listeners rely on to distinguish words. It accounts for reverberation, background noise, and distortion.
STI scores also range from 0 to 1, with qualitative ratings assigned to each range:
- Excellent: above 0.75
- Good: 0.60 to 0.75
- Fair: 0.45 to 0.60
- Poor: 0.30 to 0.44
- Bad: below 0.30
STI is typically measured using specialized equipment that plays a modulated test signal through the room or system and analyzes how much of the modulation pattern survives at the listener’s position. You can also convert STI to the Common Intelligibility Scale (CIS) using the formula CIS = 1 + log(STI). This conversion compresses the scale so that differences at the high end (where human perception is less sensitive to changes) appear smaller, better matching how people actually experience improvements in clarity.
Adaptive Testing in Background Noise
Some clinical tests measure intelligibility by finding the noise level at which a person’s understanding breaks down, rather than calculating a flat percentage. The Hearing in Noise Test (HINT) is a widely used example. It presents 250 sentences across 25 lists while background noise plays from surrounding speakers. Speech typically starts at 55 decibels with noise fixed at 65 decibels.
The test adjusts automatically: every time the listener repeats a sentence correctly, the next sentence gets quieter. Every incorrect response makes the next sentence louder. This adaptive process zeroes in on the signal-to-noise ratio where the listener gets about 50% of sentences right. The result isn’t a percentage score but a threshold, the specific noise-to-speech ratio at which intelligibility falls to chance. Lower thresholds indicate better hearing in noise.
Choosing the Right Method
The method you pick depends entirely on what you’re evaluating. For a child’s speech development, word-level transcription intelligibility or PCC gives the most actionable information. For hearing ability in everyday environments, SII or adaptive noise testing reveals how much speech information is actually reaching the listener. For room or system design, STI tells you whether the physical space supports clear communication.
Regardless of the method, two factors dramatically affect the result: the familiarity of the listener with the speaker, and whether the speech sample uses single words or connected sentences. Familiar listeners consistently score speakers higher than unfamiliar ones, and context within sentences helps listeners fill in gaps they’d miss with isolated words. Any intelligibility score should note both of these conditions to be meaningful.