Examining the Impact of Sample Length on the Reliability of Narrative Retell Language Samples

Clinicians routinely cite a lack of time available to elicit, transcribe, segment, and analyze language samples. Therefore, the purpose of this study was to determine if a shorter narrative language sample maintained acceptable levels of reliability as a more traditional, longer narrative sample.

The participants included 190 school-age children in the first to sixth grades. Participants completed two, brief narrative retells using the Narrative Language Measures (NLM) Listening subtest of the CUBED assessment and one longer, narrative retell using the wordless picture book, Frog, Where Are You? (FWAY). These language samples were then analyzed for language productivity, complexity, and story grammar elements using the Systematic Analysis of Language Transcripts software program and the NLM Flow Chart.

Our hypothesis was that there would be evidence of relative reliability between the NLM and FWAY, but the correlations would not be as strong as data reported in previous studies that used segments from within the same language sample; this hypothesis was confirmed. Regarding absolute reliability, we hypothesized that the NLM would yield greater language complexity due to the model stories having a greater density of language complexity features. This hypothesis was confirmed. The NLM elicited significantly higher values for subordination index, moving average type-token ratio, and words per minute. Only percent opportunities of narrative structure was significantly higher for FWAY. Mean length of utterance was not statistically different.

It appears that shorter narrative retells elicited from a complex model have the possible advantage of taxing a student's language ability, potentially yielding more complex language from a child. Academically relevant information can be gathered from a brief narrative retell elicited from a complex model. Consequently, the CUBED NLM subtest, which can be administered and scored in approximately 2 min, provides a clinically efficient, less cumbersome, and academically relevant language sample for clinicians.