Structural Analysis of a Language Sample
- Created: Monday, 28 November 2011 20:22
- Updated on Friday, 10 June 2016 14:01
When Speech-Language Pathologists (SLPs) / Speech and Language Therapists (SLTs) make an assessment of a child's progress in acquiring spoken language, a procedure they often use is a structural analysis of a language sample.
Cite this article as:
Bowen, C. (2011). Structural analysis of a language sample. Retrieved from http://www.speech-language-therapy.com/ on [insert the date that you retrieved the file here].
First, the SLP/SLT samples the child's language in the form of a written-down list of separate "utterances". The words in the utterances are recorded verbatim. For example, if a child says, "It two mouse" (meaning "It's two mice") - the SLP/SLT records "It two mouse". It takes practice to learn to record language samples accurately as there is a natural tendency to perceive and write down a "corrected" version of what was said.
BUT - if a child mispronounces words (including those with word-endings that serve a grammatical function such as marking plurals and verb tenses) the SLP/SLT writes down the sentence as though all the words were pronounced correctly.
For example, some children produce a "stop" consonant in place of the 's' sound, saying things like "One bee, two beed" instead of "One bee, two bees". In this instance the SLP/SLT would record "One bee, two bees" as though all the words had been pronounced correctly. This provides a record of the child's way of "marking" -s plurals even though it may not conform to the adult standard.
Some children omit final consonants (as in final consonant deletion or deletion of coda) and may say for example "One mou two mie" in place of "One mouse, two mice". Here, the SLP/SLT would record, "One mouse, two mice". Similarly, "One mout, two mite" would be recorded as "One mouse, two mice".
It can be very difficult, even impossible, to collect language samples from children with many speech sound errors. For example, if a child says "It hih book", are they saying "It his book" or "It him book" or "It is book" or "It's his book"?
The SLP/SLT usually makes a note of the linguistic context in which the utterances were spoken. Ideally, a language sample comprises at least 200 utterances.
Many speech professionals like to take samples in different settings (e.g., in the child's home, at pre-school, and in the clinic) and with different communicative partners (e.g., the child talking to his/her parent(s), sibling(s), peer(s), teacher or SLP). Language samples are often, but not always, audio or video recorded.
An excerpt from Kim's language sample
The resultant language sample looks something like the following 30-utterance excerpt from Kim's 230-utterance language sample. The numbers in parentheses represent the number of units of meaning, or morphemes in each of Kim's utterances. Kim was aged 3;8 (44 months) when the sample was taken as part of his initial speech and language assessment.
The context of Kim's sample
What was the linguistic context? Kim was having a conversation with his mother in my room while I looked on (audio-taping and transcribing the conversation). They were finding things in a toy box.
The child's written-down (transcribed) utterances are checked for accuracy against the audio recording. The speech-language pathologist then adds up the number of utterances, and the number of morphemes in each utterance. To determine the mean length of utterance in morphemes (MLUm) the number of morphemes is divided by the number of utterances.
In Kim's mini-sample there were 30 utterances, and a total of 56 morphemes. So his MLUm was: 56 ÷ 30 = 1.86 morphemes.
When the same calculation was done with Kim's full sample of 230 utterances and 573 morphemes his MLUm was 2.49 morphemes (which just goes to show that too small a language sample can be, and usually is, misleading).
Interpreting the MLUm
An MLUm of 2.49 is appropriate for a child in Brown's Stage 2 of language development. Stage II language structures are usually being used by children in the 27 to 30 months age range. As Kim was aged 44 months when the sample was taken, his MLUm was below age-expectations (SLP/SLT-speak for "low for his age"!).
Structural analysis ("Brown's Stages")
Looking in detail at Kim's mini-sample it can be seen that all the language structures that emerge in Stage II ("-ing", "in", "on" and "-s plurals") are present. None of the Stage III or Stage IV structures expected at his age can be seen. This was reflected in the full sample also.
Kim's MLUm, and his structural analysis, both placed him in Brown's Stage II (27 to 30 months level) at 44 months of age.
The MLUm is not without its critics. Miller and Chapman (1981) showed a strong positive correlation between MLUm and age which has proved difficult to replicate, though it has since been done by some investigators. MLUm calculations above 4.0 must interpreted cautiously, particularly considering the reservations of both Crystal (1974), and Bennett-Kastor (1988). Both authors had concerns about the disadvantages of MLUm’s in terms of definition, application and interpretation.
While it is readily acknowledged that the MLUm is a weak measure of language complexity above 4.0 morphemes, it is widely applied as a clinical measure for children with MLUm's below 4.0 morphemes. It does not stand alone, and should be used in conjunction with other tests of language production.
MLUm in a speech and language research setting
At the 16th Symposium on Research in Child Language Disorders, University of Wisconsin-Madison, June 2-3, 1995, Pamela Rosenthal Rollins of the University of Texas at Dallas presented a paper entitled: MLU AS A MATCHING VARIABLE: UNDERSTANDING ITS LIMITATIONS.
The abstract speaks for itself:
"Despite much criticism MLU is often used to "language-age- match" children with language disorders to younger children with typical language. This study explores the fallacious nature of MLU as a variable for language-age-matching children. The linguistic performances of five children with SLI were each compared with an MLU matched child, a lexically matched child, and a morphologically matched child. The findings demonstrated that MLU did not ensure that the children were matched on linguistic performances. When taken as a group, the children with SLI were not comparable to the control group with respect to morphological or lexical skill. Thus, the language skills were not equivalent between the groups. The findings suggest that we must go beyond global indicators such as MLU to the notion that a child's language level should be represented as a profile of scores on a variety of component skills".
MLU and MLUm mean the same thing. "Brown's Stages" were identified by Roger Brown in his classic book: Brown, R. (1973). A first language: The early stages. London: George Allen & Unwin Ltd.
2002 Article in JSLHR
Grammatical Morpheme Production in 4-Year-Old Children
Journal of Speech, Language, and Hearing Research
Volume 45, Issue 5. Pages 961 - 969. October 2002
Denise Balason and Christine Dollaghan have written a terrific article which caused many to stop and take stock of what we regard as "normal" with regard to 4 year olds' grammatical morphemes, and the way in which we apply the MLUm metric.
"Grammatical Morpheme Production in 4-Year-Old Children. Despite the importance of grammatical morpheme (GM) production for both clinical decision-making and theoretical accounts of child language impairment, evidence concerning developmental expectations for GM use is inadequate.We studied grammatical morpheme production in 15-minute spontaneous language samples from a large (N = 100), sociodemographically diverse group of 4-year-olds.
Substantial variability was observed in both the frequency of obligatory context(OCs) and in the percentage of correct usage of GMs. For only one morpheme did all 100 samples contain the minimum number of 3 OCs; for only 7 of the 14 GMs was an adequate number of OCs found in at least half of the 100 samples.
Although mean percentages of production from samples with 3 or more OCs were high (> 85%), fewer than 25% of participants contributed to the "group" means for 6 of the 14 GMs.
Results from the present investigation indicate a need for caution in interpreting information on GM production derived from samples of this nature from children at this age; the validity of using such data to identify deficits in inflectional morphology for either clinical or research purposes appears questionable." Results from this study were presented at the Symposium for Research on Child Language Disorders (SRCLD) in June 2001.
Balason, D.V. and Dollaghan, C.A. (2002). Grammatical Morpheme Production in 4-Year-Old Children. Journal of Speech, Language, and Hearing Research, 45(5). 961 - 969.
Bennett-Kastor, T. (1988). Analyzing Children's Language. Blackwell Publishers.
Brown, R. (1973). A first language: The early stages. London: George Allen & Unwin Ltd.
Crystal, D. (1974). Review of R. Brown, A first language, Journal of Child Language, 1, 289-307.
Miller, J. F. (1981). Assessing Language Production in Children: Experimental Procedures, Baltimore, MD: University Park Press.
Miller, J. F. & Chapman, R. S. (1981). The relation between age and mean length of utterance on morphemes. Journal of Speech and Hearing Research, 24, 154-161.