PURPOSE: To analyze the reliability level of phonological transcriptions of speech samples produced by children with Language Impairment (LI), and to verify whether there was significant disagreement between the transcriptions, by comparing subjects who were able to produce discourse and those who were not. METHODS: Speech samples of 37 three- to five-year-old subjects with LI, previously collected and analyzed using two tasks (picture naming and repetition of words), were re-transcribed. Subsequently, the researchers accessed the first transcriptions, in order to calculate the agreement level. Transcriptions whose disagreement index was higher than 20% were transcribed for the third time. The ability to produce discourse at the time of data collection was also considered in the analysis. RESULTS: For both tasks, there was a predominance of agreement lower than 80% (p<0.001) when the first two transcriptions were taken into account. Meanwhile, the agreement between the first, the second and the third transcriptions was higher than 80% (picture-naming task: p=0.001; repetition of words: p<0.001). A difference was found between the second and the third transcriptions (p<0.001). There was no difference in the reliability of the analyzed tasks (second transcriptions: p=0.565; third transcriptions: p=0.645). The comparison between the subjects who were able to produce discourse and those who were not showed a difference only in the third transcriptions of the picture-naming task (p=0.033). CONCLUSION: The reliability of phonological transcriptions of children with LI was low, and the picture-naming tasks of children who were able to produce discourse presented higher reliability.
Child language; Speech articulation tests; Reproducibility of result; language disorders; Child
