Mean length of utterance

Also called MLU, MLUm, Utterance length

The number in a therapy report that says how much a person says in one go. Worth understanding, because it is how progress gets measured.

What it is

Mean length of utterance, MLU, is the average number of words, or more precisely morphemes, in what a person says in one go. Take fifty or a hundred utterances from a recording, count the units in each, and average. "Dog running" is two words but three morphemes (dog, run, -ing), and therapists usually count morphemes because that captures grammar as well as length.

It's a rough and widely used measure of expressive language in young children, and you'll see it in reports as "MLU of 2.4" or similar.

Who it can help

Parents and teachers reading therapy reports, and anyone tracking whether a child's language is growing. It's most useful between about eighteen months and five years, when length maps reasonably well onto grammatical development. Above an MLU of around four or five it stops being informative, because older children vary length by choice, not ability.

It's also used, and misused, in AAC. An AAC user's MLU on a device tells you as much about the device and the partners as about the person. Someone whose board only allows single symbols will have an MLU of one for ever, and reporting that number as a description of the person is unfair.

How it works

The classic stages, from Brown's work in the 1970s and still used: MLU under 2 is single words and early combinations (roughly up to two years); 2 to 2.5 is the two-word stage with the first grammatical endings; 2.5 to 3 brings questions and negatives; 3 to 3.75 adds embedded phrases; above 3.75 the child is joining clauses. The ages are approximate, with wide normal variation, and a therapist will treat them as a guide rather than a rule.

Counting is done on a language sample, ideally in play or conversation rather than a test, because children say more when they're not being tested. Imitated utterances and fillers are excluded.

For a person who uses AAC, a fair sample needs the right conditions: a vocabulary that allows combinations, partners who wait, and enough time. If any of those is missing, the number measures the environment.

Getting started

Record five minutes of ordinary play or chat once a month and write down what the child said. Count the words per utterance and average. You don't need the morpheme precision to see the trend, and the trend is what matters.

Aim one step up. If most utterances are one word, model two (expanding and recasting). If they're two, model three. Targeting length directly, by asking for "a full sentence", tends to produce nothing; modelling produces growth.

If a report gives an AAC user a low MLU, ask what vocabulary was available, how long the sample was and how much wait time was given. Those are fair questions and a good therapist will welcome them.

Where to get help

The Royal College of Speech and Language Therapists is the professional body for the people who calculate this, and a therapist should be happy to explain a report line by line; ask. Speech and Language UK gives the same stages in parent language on its ages and stages pages.