Training understanding of reversible sentences: a study comparing language-impaired children with age-matched and grammar-matched controls

Introduction. Many children with specific language impairment (SLI) have problems with language comprehension, and little is known about how to remediate these. We focused here on errors in interpreting sentences such as "the ball is above the cup", where the spatial configuration depends on word order. We asked whether comprehension of such short reversible sentences could be improved by computerized training, and whether learning by children with SLI resembled that of younger, typically-developing children. Methods. We trained 28 children with SLI aged 6-11 years, 28 typically-developing children aged from 4 to 7 years who were matched to the SLI group for raw scores on a test of receptive grammar, and 20 typically-developing children who were matched to the SLI group on chronological age. A further 20 children with SLI were given pre- and post-test assessments, but did not undergo training. Those in the trained groups were given training on four days using a computer game adopting an errorless learning procedure, during which they had to select pictures to correspond to spoken sentences such as "the cup is above the drum" or "the bird is below the hat". Half the trained children heard sentences using above/below and the other half heard sentences using before/after (with a spatial interpretation). A total of 96 sentences was presented over four sessions. Half the sentences were unique, whereas the remainder consisted of 12 repetitions of each of four sentences that became increasingly familiar as training proceeded. Results. Age-matched control children performed near ceiling (≥ 90% correct) in the first session and were excluded from the analysis. Around half the trained SLI children also performed this well. Training effects were examined in 15 SLI and 16 grammar-matched children who scored less than 90% correct on the initial training session. Overall, children's scores improved with training. Memory span was a significant predictor of improvement, even after taking into account performance on training session 1. Unlike the grammar-matched controls, children with SLI showed greater accuracy with repeated sentences compared with unique sentences. Training did not improve children's performance on a standardized test of receptive grammar. Discussion. Overall, these results indicate that a subset of children with SLI perform well below ceiling on reversible sentences with three key words and simple syntactic structure. For these children, weak verbal short-term memory appears to impair comprehension of spoken sentences. In contrast to the general finding that rule-learning benefits from variable input, these children seem to do best if given repeated exposure to the same nouns used with a given sentence frame. Generalisation to other sentences using the same syntactic frame may be more effective if preceded by such item-specific learning.