How To Calculate MLU: A Definitive Guide To Mean Length Of Utterance And Linguistic Assessment

To calculate the Mean Length of Utterance (MLU), divide the total number of morphemes by the total number of fully intelligible utterances within a representative speech sample, typically consisting of 50 to 100 entries. This foundational metric serves as a quantitative indicator of a child's syntactic complexity and grammatical development, directly correlating with the established Brown’s Stages of Language Development.

Strategic Planning for Language Sample Collection and Analysis

Calculating a precise Mean Length of Utterance requires rigorous preparation to ensure the validity of the linguistic data. The primary objective is to capture a representative "snapshot" of the subject's spontaneous speech, which means the environment and the examiner’s elicitation techniques must be standardized yet naturalistic. Before beginning the calculation, you must determine whether you are measuring MLU-m (morphemes) or MLU-w (words), though MLU-m remains the gold standard for clinical assessment in early childhood.

• Essential Equipment and Materials: High-fidelity audio or video recording device, a quiet environment free from auditory distractions, a set of age-appropriate toys (e.g., a farm set, a dollhouse, or a garage) to stimulate spontaneous conversation, and transcription software or standardized forms.
• Mandatory Knowledge Standards: Proficient understanding of Brown’s Rules for Morphemic Segmenting (1973), familiarity with standard orthographic transcription, and awareness of the subject's primary dialectal influences to avoid penalizing linguistic variations.
• Estimated Duration Benchmarks: Sample collection typically requires 15 to 30 minutes of interaction. Transcription and segmentation for a 100-utterance sample generally require 60 to 90 minutes for an experienced clinician.
• Sample Size Requirements: A minimum of 50 consecutive, intelligible utterances is required for a basic screening, though 100 utterances provide a statistically significant representation of the child's true grammatical capabilities.

Executing the MLU Calculation Workflow

Step 1: Eliciting the Spontaneous Speech Sample

Initiate the session by engaging the child in play-based interaction. Avoid asking closed-ended "yes/no" questions, as these artificially depress the MLU. Use open-ended prompts such as "Tell me about this" or "I wonder what happens next." The goal is to obtain at least 100 utterances to ensure the sample is not skewed by brief responses or repetitive phrases.

Pro-Tip: If the child is highly reticent, use a "parallel play" approach where you narrate your own actions. This often reduces the pressure on the child and encourages them to offer spontaneous comments that more accurately reflect their syntactic ceiling.

Step 2: Transcribing the Utterances

Listen to the recording and transcribe the child's speech verbatim. An "utterance" is defined as a complete thought, a sentence, or a meaningful unit of speech bounded by a pause or a change in intonation. Follow these specific exclusion rules to maintain data integrity:

1. Exclude fillers such as "um," "uh," and "oh," unless they serve a specific communicative function.
2. Exclude false starts or self-corrections (e.g., if a child says "The dog... the big dog is running," only count "the big dog is running").
3. Exclude identical repetitions of an utterance unless the repetition is for emphasis.
4. Exclude songs, nursery rhymes, or "counting" sequences, as these are rote-memorized rather than syntactically generated.
5. Mark unintelligible words with an "X" and exclude the entire utterance from the final count if the meaning is obscured.

Step 3: Segmenting and Counting Morphemes

This is the most technical phase of the process. You must apply Brown’s Rules to determine the total morpheme count. A morpheme is the smallest unit of meaning.

• Count as One Morpheme: Compound words (birthday, cupcake), proper names (Mickey Mouse), ritualized reduplications (choo-choo, night-night), irregular past tense verbs (went, ate, saw), and auxiliary verbs or catenatives (gonna, wanna, hafta). Diminutives like "doggie" or "mommy" also count as a single morpheme because the child does not perceive the suffix as a separate unit of meaning.
• Count as Two Morphemes: Possessive nouns (daddy’s), plural nouns (dogs), third-person singular present tense verbs (walks, runs), regular past tense verbs (walked, jumped), and present progressive verbs (running, eating).

Warning: Do not count "can't" or "don't" as two morphemes unless there is evidence in the transcript that the child uses "can," "do," "not," or "does" as separate words in other contexts. Usually, these are counted as one morpheme for children in the early stages of development.

Step 4: Applying the Mathematical Formula

Once the transcription is segmented, tally the total number of morphemes across all included utterances. Then, tally the total number of utterances.

The formula is: Total Number of Morphemes divided by Total Number of Utterances equals MLU.

For example, if a child produces 150 morphemes across 50 utterances, the MLU is 3.0. Ensure your math is double-checked, as even a minor deviation of 0.2 can result in a child being categorized into a different developmental stage.

Step 5: Interpreting the Results

Compare the calculated MLU against Brown’s Stages of Language Development. This comparison helps determine if the child's expressive language is age-appropriate or if there is a significant delay necessitating intervention. An MLU that is more than one standard deviation below the mean for the child's chronological age is typically a trigger for more intensive diagnostic testing.

Morpheme Counting Criteria and Brown’s Developmental Stages

The following table outlines the correlation between MLU, the child’s chronological age, and the expected linguistic milestones based on the longitudinal research conducted by Roger Brown.

Rectifying Discrepancies in Language Sample Analysis

Calculating MLU is not always a straightforward process, particularly when dealing with neurodivergent populations or diverse linguistic backgrounds. Identifying the root cause of skewed data is essential for an accurate clinical picture.

• Scenario: Artificially High MLU due to Echolalia

  
  
  
  • Root Cause: The child is repeating long, complex phrases from movies or adult speech (delayed echolalia) without understanding the underlying syntax.
  • Actionable Fix: Exclude all echolalic utterances from the count. MLU is intended to measure productive, generative language, not auditory memory.

• Scenario: Artificially Low MLU in Multilingual Children

  
  
  
  • Root Cause: The child is code-switching or lacks the specific grammatical markers in English that are counted in MLU-m, even though their conceptual complexity is high.
  • Actionable Fix: Calculate MLU in both languages if possible. For the English sample, use MLU-w (words) alongside MLU-m to see if the discrepancy is caused specifically by a lack of bound morphemes (like -ed or -s) rather than a lack of syntactic structure.

• Scenario: High Volume of Unintelligible Utterances

  
  
  
  • Root Cause: Severe phonological impairments or poor recording quality make more than 20% of the sample unreadable.
  • Actionable Fix: The MLU is invalid if unintelligibility is high. You must re-record the sample in a controlled environment or use a different assessment tool that relies less on phonetic precision, such as a parent-report vocabulary checklist.

• Scenario: Skewed Results from "Yes/No" Bias

  
  
  
  • Root Cause: The examiner asked too many closed-ended questions, leading to a sample filled with one-word "yes" or "no" responses.
  • Actionable Fix: Discard the sample and re-collect. Focus on using "Stalled Narratives" where you start a story and stop at a climax, forcing the child to provide a longer, more complex explanation to continue the play.

Frequently Asked Questions

What is the difference between MLU-m and MLU-w?

MLU-m counts individual morphemes (meaningful units), such as the "run" and the "ing" in "running," totaling two. MLU-w counts only whole words, meaning "running" would count as one. MLU-m is generally preferred for children under age five because it captures the emergence of grammatical markers, while MLU-w is often used for older children or in cross-linguistic studies where morphemic structures vary wildly.

At what age does MLU lose its clinical validity?

MLU is a highly sensitive measure of language development up to an MLU of approximately 4.0 to 4.5, which usually corresponds to a chronological age of about five years. Beyond this point, children begin to use more complex, embedded sentences where length no longer directly correlates with sophistication, and other measures like the Index of Productive Syntax (IPSyn) become more useful.

How do I handle contractions like "don't" and "can't"?

According to Brown's original rules, these are counted as one morpheme for young children because they are often learned as unanalyzed wholes. However, if the child demonstrates use of the separate components elsewhere in the transcript—for example, saying "I do not want it" or "I can do it"—then you may count the contractions as two morphemes to reflect their advanced grammatical knowledge.

Should I count "stuttering" or dysfluencies in the MLU?

No, dysfluencies should not be counted as additional morphemes. If a child says "I-I-I want the b-b-ball," it is counted as a four-word/four-morpheme utterance: "I want the ball." Counting repetitions caused by stuttering would lead to a false inflation of the MLU, incorrectly suggesting higher linguistic complexity than the child actually possesses.

Master the Science of Linguistic Assessment

Accurate MLU calculation is a cornerstone of professional speech-language pathology and developmental research. By applying these rigorous counting rules and elicitation techniques, you ensure that every clinical intervention is backed by precise, objective data.