Analyze and interpret a language sample

Calculate the selected measures

MLU in words divides the number of included words by the number of included utterances. MLU in morphemes uses the morpheme count instead. In either calculation, the method’s rules determine how repetitions, incomplete utterances, and other features enter the counts. Record the unit and counting conventions with the result; the label “MLU” alone leaves both unspecified.

Some results depend on specific codes in the transcript. SALT’s Subordination Index and narrative composite results, for example, require their corresponding coding. Its guide to analysis variables identifies these requirements. SUGAR provides analysis training and sub-analysis forms for its procedures.

Apply the reference comparison

A tutorial comparing CLAN, SALT, and SUGAR found similar MLU values in one case but different distances from the respective reference means. The authors discuss the different reference samples and sampling conditions. A measured value therefore needs its reference source to explain the comparison. Pezold, Imgrund, and Storkel, 2020

Percentiles require the method’s reference distribution or conversion; thresholds and categories come from its interpretation guidance. Record the database or table and the comparison settings used, including the selected age group where applicable.

If the completed sample differs from what the reference procedure requires, see “When an intended analysis is unavailable” below.

Interpret the measures

Each measure summarizes particular features and leaves others to examine:

Definitions and counts vary by method; SALT’s variable guide provides one set.

For example, two retells might receive the same narrative total while differing in their component scores. One speaker establishes the characters and setting clearly but leaves out the resolution. Another describes the outcome but gives the listener little information about who is involved. Retaining the component scores and corresponding passages makes those differences visible.

Describe patterns and select illustrative examples

For a concern about explaining events clearly, examine event sequence, causal connections, and information the listener needs. Other questions call for different areas of description:

ASHA’s spoken-language assessment guidance covers language domains and discourse contexts in more detail.

The surrounding exchange can change what an excerpt tells you. Consider this opening to a picture-based retell in which the listener cannot see the pictures and the student has not yet introduced the characters:

Student: He went back to get it.

Clinician: Who went back?

Student: The boy. He forgot his backpack.

With no shared picture or earlier introduction to identify “he,” the listener needs clarification. The student supplies the character’s identity and a reason for returning after the question. Including the clinician’s turn shows the support under which that information appeared.

To see whether this recurs, locate the other character introductions in the retell. Distinguish clear first mentions, later independent clarifications, and information supplied after questions. That comparison shows whether the excerpt represents a recurring difficulty or an isolated instance, and where the student succeeded.

A retell may contain no explanations of a character’s motives because the story offered little reason to discuss them. If motives are the feature of interest, a task that invites those explanations can help distinguish lack of opportunity from difficulty producing them. The Pezold tutorial discusses opportunities to demonstrate forms when identifying structures for further probing. Clinical tutorial

When an intended analysis is unavailable

A sample may support a measured value even when it does not qualify for a reference comparison. Other limitations affect the calculation itself:

Pezold and colleagues provide a concrete example: their cases did not contain enough qualifying utterances for an overall Developmental Sentence Scoring score. They used the available productions to examine attempted forms and errors and identify potential probes. Clinical tutorial

Compare results across time

Place the same measures from each occasion side by side, with the task, sample length, scoring conventions, and support provided. Differences in these conditions affect what the comparison can tell you.

For example, a later retell may contain longer utterances and more detail about the characters and events. Detail supplied after additional clinician questions shows performance with that support. Detail supplied independently provides evidence of elaboration without those questions. Include the relevant exchanges from both occasions so the comparison captures the difference.

For age-based reference comparisons, track the measured value and relative standing separately. An increase in utterance length may accompany a stable or lower relative standing if the older reference group also produces longer utterances.

Research on sample length shows that reliability findings vary by measure and study design; it also identifies unresolved questions about clinically meaningful differences. Wilder and Redmond, 2022 Where a change threshold has not been established for the intended use, describe the observed difference and supporting sample evidence without assigning one.

Language sample reporting and next steps continues from these results to report examples, intervention targets, and monitoring.