Inference and sentence comprehension in children with specific or pragmatic language impairments

Children with language impairments have difficulty in reporting verbal inferences, but it is unclear whether the source of this problem lies in limitations of language comprehension, an inability to access world knowledge, or the integration of information in discourse. Children with pragmatic language impairments (CwPLI) are often described as having disproportionate difficulty with verbal inference, although there is little empirical support for this view.

To examine the performance of children with specific language impairment (CwSLI), CwPLI and a control group with typical language development (CwTLD) on a verbal inferential comprehension task; to explore whether CwPLI's verbal inferential comprehension performance are disproportionate to their sentence comprehension difficulties; and to analyse inferential error types made by the three groups.

Sixty-four children with language impairments (CwLI), aged between 6 and 11 years, participated in an inference comprehension task (IC) and a task of sentence comprehension (SC). For the purpose of analysis, the LI group was then split into two subgroups: CwPLI and CwSLI, using a checklist of pragmatic features. A group of CwTLD were matched to LI children on SC ability and gender; a separate group of CwTLD (CA) was matched to CwLI by age and gender.

CwLI had lower raw scores on the IC task than CA-matched CwTLD, but similar scores to SC-matched CwTLD. CwPLI produced lower raw IC scores than CwSLI, and this was the same when age and SC performance were controlled for. CwSLI performed similarly to their SC-matched and CA-matched control subjects on the IC task. CwPLI scored significantly less on the IC task than both their SC-matched and CA-matched control subjects. A regression analysis showed that age and SC ability account for 36% of the variance on the IC task, but group (SLI versus PLI) status accounts for just 5%. A detailed analysis of error types and item performance on the IC task showed a trend for CwPLI to perform more poorly on developmentally more complex inference items, but there was no evidence for a unique pattern of inference errors in CwPLI.

Children with language impairments tend to perform like younger children with matched level of sentence comprehension ability on a verbal inference comprehension task. CwPLI have more difficulty with this task than CwSLI, but the distribution of scores shows considerable overlap exists between clinical groups. Sentence comprehension ability, age and, to a lesser extent, group status contribute to verbal inferencing ability in children with language impairments, but other factors may be important. The implications for assessment and intervention are discussed.