Abstrakt
Sentence imitation task is a recognized marker of language impairment, even though it may appear to be primarily a memory task. The task makes it possible to target specific morphemes or structures in the stimulus sentences.
This is useful in languages with underexplored morphological development (e. g. Smolik, Vávrů, 2014).
Lukács et al. (2009) examined the ability to produce verb morphemes in Hungarian typically developing children and children with SLI, using the sentence imitation task with target morphemes in stimulus sentences replaced by masking sounds. Following Leonard (2014), they found that the if the morpheme completions were ungrammatical, they usually deviated from the targets in one category (tense, person, number, definiteness) only ("near-misses").
The present study used a similar task in Czech, focusing both verbs and nouns. The study also tested whether sentence imitation performance has unique relations with language mastery, after controlling for shortterm memory.
We created a sentence imitation task with masked inflectional endings in nouns (encoding case-number) or verbs (person, number, tense, and sometimes gender). There were total 72 sentences for nouns and 33 for verbs.
They were presented to 17 children with language impairment (LI) aged 5;1 to 7;6, and 17 typically developing (TD) vocabularymatched children (3;8 to 4;11). Children were given a vocabulary comprehension task and the forward and backward digit span task from WISC.
Overall sentence imitation ability was evaluated by scoring sentences with 2 points for exact imitations, 1 for grammatical imitations with 1 or 2 deviations (except masked morphemes), and 0 otherwise. Further analyses focused on the masked morpheme completions.
The overall analysis confirmed that sentence imitation is a sensitive marker of language impairment, with clear differences of the total accuracy between typical and DLD children (Tab. 1). Regression analyses showed significant unique effects of language impairment (β=0.50), vocabulary (β =0.40), and forward digit span (β=0.26), confirming that sentence imitation is a sensitive marker of DLD, and that it is sensitive to language skills even when controlling for memory.
Tab. 2 shows that DLD children are less likely to produce grammatically well-formed contexts for the completions of target morphemes, which is a consequence of lower imitation accuracy. Morpheme completions were then evaluated in the well-formed contexts.
Children with DLD showed significantly lower proportion of grammatical completions of noun targets, but not verb targets (Tab. 3). The accuracy was generally lower in nouns than verbs because the grammatical context in verb-target sentences was less restrictive and there were more possible grammatical completions.
The ungrammatical completions were further analyzed with respect to their deviation from the closest grammatical possibility: about 80% of deviations involved one grammatical category only ("near-miss") in both TD and DLD (Tab. 4). About 75% of the deviations involved forms with higher frequency than the target form, so frequency could explain a majority but not all deviations.
The results lend only limited support to Leonard's (2014) morphological richness account. Ungrammatical morpheme use often but not always consists of near misses, and these are often but not always attributable to higher-frequency forms.