Brown's 14 Morphemes: A Full Clinical Reference for SLPs

Roger Brown’s A First Language (1973) is the closest thing the field has to a founding document for descriptive child-language assessment. From the longitudinal records of three children — Adam, Eve, and Sarah — Brown isolated 14 grammatical morphemes that emerged in a remarkably consistent order across all three participants and that, taken together, account for most of what changes between the telegraphic two-word stage and adult-like English grammar. Six decades later, every clinical Language Sample Analysis report still references the same 14 morphemes by the same names, in the same order.

The morphemes stuck because they have three useful properties at once. They are observable in spontaneous speech rather than requiring an elicited probe, they emerge in a predictable order that maps onto Brown’s five MLU-defined stages, and they each have an obligatory context — the moment in an utterance where an adult English speaker would supply the morpheme without thinking. That third property is what lets clinicians move from "the child does not say -ing" to a hard percentage score: count the obligatory contexts in the sample, count the times the child supplied the morpheme correctly, and the ratio is your accuracy.

The Cazden (1968) doctoral analysis of the same Brown corpus and the de Villiers and de Villiers (1973) cross-sectional replication on 21 unrelated children both confirmed the order. SUGAR (Pavelko & Owens, 2017) and Rice et al. (2010) updated the typical-acquisition windows to a modern American sample. The modern evidence base is good enough that the Brown order is now baked into every commercial transcription package and every clinical scoring rubric SLPs use today.

Brown ordered the 14 morphemes by the median age at which his three children reached 90% accuracy in obligatory contexts. The 90% mastery criterion has stuck — it is the criterion that Pavelko and Owens (2017) still use for the SUGAR norms and that most state IEP rubrics reference. The order below is the consensus Brown order; the developmental ages are pooled from Brown (1973), de Villiers and de Villiers (1973), and Owens (2014, p. 158).

The first four morphemes in Brown’s order are the grammar that shows up first because they extend two-word telegraphic speech without disrupting it. Each one adds a small inflection or function word that makes the utterance closer to adult English without forcing the child to restructure the underlying word order. These four are the morphemes you almost always see emerging together in the late twos.

Present progressive -ing is the first to reach 90% accuracy. It appears on action verbs ("baby running", "daddy eating") and is one of the easiest morphemes to elicit because action pictures and live-action play both create dense obligatory contexts. The preposition pair in and on emerge next, almost in lockstep, because spatial language is one of the first semantic domains a toddler talks about. Regular plural -s rounds out Stage II as the first true productive morpheme: when the child says "doggies" without an adult model, you have evidence of a rule, not just a memorised form.

These four morphemes are the load-bearing grammar of Brown Stage II (MLU 2.0-2.5) and the early part of Stage III. A child who is producing all four in obligatory contexts is on track for a typical MLU climb into the 3.0-4.0 range. A child who is missing two or more of them at age 3 is the kind of profile that warrants a structured language sample and a referral conversation.

The middle band of Brown’s order is the bridge between rote-learned forms and rule-governed grammar. Irregular past tense ("went", "ate", "ran", "fell") shows up before regular past tense -ed, which is initially counterintuitive — but it makes sense once you remember that irregular forms are memorised whole-word lexical items, while regular -ed requires a morphological rule the child has not yet acquired. Children often go through a stage of perfect "went" before regressing to "goed" once the regular rule comes online; this is the famous U-shaped curve and it is a sign of typical, not atypical, development.

Possessive ’s and uncontractible copula (the form of "to be" that cannot contract, as in "There it is" or "Yes, she is.") complete the Stage III-IV cluster. Possessive ’s is one of the cleanest morphemes to score because the obligatory context is so structurally narrow — a noun phrase modifying another noun phrase — and pictures of "[Character’s] [object]" reliably elicit it. Uncontractible copula is harder because the contexts are sparser; clinicians often need to plant a few question-and-answer trades into the sample to get enough obligatory contexts to score it reliably.

The last seven morphemes in Brown’s order are the ones that distinguish Stage V from "still in Stage IV" and they are also the morphemes that drag MLU from 3.5 toward 5.0. Articles a and the come first, partly because they are short and partly because they live inside almost every noun phrase the child produces. Regular past tense -ed comes next, often right around the time the child also produces over-regularised forms like "goed" and "runned" — again, the U-shaped regression is typical, not atypical.

The third-person singular -s morpheme (regular and irregular together) is one of the latest to reach mastery and one of the most diagnostically interesting. It is also the morpheme most often missing in children with Developmental Language Disorder (DLD) and the morpheme that Rice et al. (2010) flagged as a clinical marker. A 5-year-old who consistently omits third-person -s in obligatory contexts ("she walk to school", "he like ice cream") is producing one of the more reliable single-morpheme markers of DLD in monolingual American English.

The two auxiliary "be" morphemes — uncontractible auxiliary ("Is she eating? Yes, she is.") and contractible auxiliary ("She’s running.") — are the last to reach mastery and are the morphemes that finally bring the child’s MLU into the 4.5+ range. Until they are productive, an English-speaking child cannot comfortably produce extended progressive aspect, which caps the length of action narratives.

A morpheme is mastered when the child produces it correctly in 90% of its obligatory contexts across at least three consecutive sessions, per Brown’s original criterion. In a clinical setting most of us do not run three sessions — we run one 50-utterance language sample — so the practical version of the rule is that the child must reach 90% in obligatory contexts within that one sample. Below 90% is "emerging"; below 50% is "not yet productive".

Counting obligatory contexts is the part of Brown’s methodology that most often gets dropped. The denominator is not the number of times the morpheme appears — it is the number of times an adult English speaker would have produced it in the same place. If a child says "two doggie" instead of "two doggies", the obligatory context still counts; the missing -s is what brings the accuracy ratio down. Without obligatory context counting you can only report whether the morpheme exists in the sample, not whether the child has acquired it.

Free MLU calculators that implement Brown’s rules in code (including the one on this site) handle compound words, irregular forms, and catenatives the same way Brown specified them in 1973: compounds count as 1 morpheme each, irregular pasts count as 1 morpheme each, and "gonna"/"wanna"/"hafta" each count as 2 morphemes (going-to, want-to, have-to). Hand-counters routinely get all three of those wrong, which is the single biggest source of MLU drift in real clinical reports.

Brown’s 14 morphemes and Brown’s five MLU stages were designed as two views of the same developmental story. As the early morphemes reach mastery, the child’s MLU climbs because every newly-acquired bound morpheme adds to the morpheme count of every utterance it appears in. The relationship is not just correlational — it is mechanical. A child who consistently produces present progressive -ing and regular plural -s adds an average of 0.4-0.6 morphemes per utterance compared with a child who omits them, simply because those morphemes appear in such a high proportion of typical toddler utterances.

In practice this means that the Brown stage and the morpheme inventory are two ways of saying the same thing. Stage II (MLU 2.0-2.5) corresponds roughly to the child producing morphemes 1-4 in obligatory contexts. Stage III (MLU 2.5-3.0) adds morphemes 5-7. Stage IV (MLU 3.0-3.75) adds 8-10. Stage V (MLU 3.75-4.5) adds 11-14. Once all 14 are productive, MLU has typically climbed to the 4.5-5.0 range and the child is past the developmental window in which Brown’s framework was designed to apply.

The clinical workflow is to compute MLU on a 50-utterance sample, look up the matching Brown stage, and then check whether the morphemes that should be productive at that stage are actually appearing in the sample. If MLU is 3.5 (Brown Stage IV) but morphemes 5-7 are still missing in obligatory contexts, the child’s grammar is uneven — their mean utterance length is buoyed by lexical content rather than grammatical structure — and the IEP goal should target the missing morphemes specifically.

A defensible IEP goal written from morpheme data names the morpheme, the structures driving the gap, the target accuracy in obligatory contexts, and the elicitation conditions. The morpheme accuracy on its own is a baseline number; the goal turns it into a measurable instructional plan.

The workflow: identify the morpheme(s) below 90% accuracy in obligatory contexts on the language sample, anchor the gap against Brown stage and the SUGAR or Rice age band, and write a goal that targets the missing morpheme with a 6- to 9-month criterion. Re-run a structured probe at the next IEP review using the same elicitation method — the morpheme accuracy in obligatory contexts is the goal’s own progress monitor.

Three options exist for actually scoring morphemes from a transcript. The browser-based MLU Calculator and Brown’s Stages Lookup tools on this site implement Brown’s 1973 rules in code, accept plain-text utterances, and produce MLU-m, MLU-w, total morphemes, and the matching Brown stage in one pass. They are free, run entirely client-side (the child’s data never leaves the browser), and are the fastest path from a typed transcript to a defensible morpheme count. The 14 morpheme programmatic pages link from this article go deeper on each morpheme individually — elicitation prompts, age bands, and example obligatory contexts.

SALT (Systematic Analysis of Language Transcripts) is the long-standing professional alternative. It implements Brown’s rules in its own scoring engine and ships with a 700-child reference database that lets clinicians compare a child’s morpheme inventory against age-banded percentile data. SALT costs roughly $295 per single-user licence and requires its own transcription syntax. For a clinician running 50+ language samples per year on a research-grade workflow, SALT is still worth the investment. For a school SLP running 10-20 samples per year, the free calculators on this site cover the same Brown rules.

ConductSpeech is the AI alternative for clinicians whose bottleneck is transcription, not scoring. It accepts the raw audio recording, runs the transcription, computes MLU plus morpheme accuracy in obligatory contexts, and drafts a present-levels paragraph that names the morphemes driving the gap. It is the only option in this list that automates the slow step — typing and segmenting the transcript — so it pays for itself on caseloads above 20-30 students.