Comparison of Grammar in Neurodevelopmental Disorders: The Case of Binding in Williams Syndrome and Autism With and Without Language Impairment

Abstract

This study investigates whether distinct neurodevelopmental disorders show distinct patterns of impairments in particular grammatical abilities and the relation of those grammatical patterns to general language delays and intellectual disabilities. We studied two disorders (autism and Williams syndrome [WS]) and two distinct properties (Principle A that governs reflexives and Principle B that, together with its associated pragmatic rule, governs pronouns) of the binding module of grammar. These properties are known to have markedly different courses of acquisition in typical development. We compare the knowledge of binding in children with autism with language impairment (ALI) and those with normal language (ALN) to that of children with WS, matched on age to the ALN group, and on age and nonverbal mental age (MA) to the ALI group, as well as to two groups of typically developing (TD) controls, matched on nonverbal MA to ALI and ALN groups. Our results reveal a remarkably different pattern of comprehension of personal pronouns and reflexives in ALI as opposed to ALN, WS, and two groups of TD controls. All five groups demonstrated an equal delay in their comprehension of personal pronouns, in line with widely reported delays in TD literature, argued to be due to delayed pragmatic abilities. However, and most strikingly, the ALI group also showed a pronounced difficulty in comprehension of reflexive pronouns, and particularly of the knowledge that the antecedent of a reflexive must c-command it. The revealed pattern confirms the existence of a particular impairment concerning Principle A in this module of grammar, unrelated to general language delays or cognitive deficits generally present in a large portion of individuals with autism as well as WS, or to general pragmatic deficits, known to be particularly prevalent in the population with autism.

Notes

¹Node A in a phrase-marker c-commands node B if the lowest node that dominates A also dominates B.

²That children are sensitive to the distinction between binding and co-reference is confirmed by their better performance on pronouns when bound by a quantified antecedent, which automatically excludes coreferential interpretation, as opposed to the referential antecedent (e.g., Chien & Wexler 1990). See Elbourne (2005) for the claim that some unpublished results (two sources) show that pronouns are difficult to interpret when bound by either referential or quantified antecedent. One of these publications, a dissertation, contains contradictory results with small sample sizes, and the other one we have not been able to obtain. The field standardly accepts the general result, in many replications.

³Eleven more children with ASD were recruited but were excluded from the study: five children aged between 7 and 12 couldn't complete the task, five children younger than 6, and one 21 years old were excluded in order to keep the age range as close possible to that of the children in the other clinical group.

⁴Note that all 14 participants recruited in Perovic et al. (2012) and included here are now classified as ALI; see below for details of reclassification of children with ASD according to general language abilities.

⁵The diagnosis of AS was clinically confirmed for 15 of the children, with all but one functioning at the very high end of the spectrum. In view of the known difficulties between distinguishing high-functioning autism (HFA) and AS (Howlin 2003), the imminent elimination of the AS diagnosis from the DSM–V (for discussion see Worley & Matson 2012), and following the advice of two anonymous reviewers, we included both HFA and AS diagnoses in the current ALN sample.

⁶There were seven children with scores in the impaired range only on one language measure, TROG-2. Four of these children (three with AS) had scores on the two vocabulary measures clearly in the average and even higher than average range: between 104 and 154, with TROG scores above 70: 74, 76, 76 and 78 (scores defined as “borderline” in Kjelgaard & Tager-Flusberg 2001). It was decided to classify these children as ALN. The remaining three children were classified as ALI, as their TROG SSs were particularly poor: 55 (below the 1st percentile), 67 (1st percentile), and 69 (in the 2nd percentile; scores defined as “impaired” in Kjelgaard & Tager-Flusberg 2001), and just about average on the other measures (SSs between 87 and 93, with one exception of a PPVT SS of 114).

⁷The number of TD controls is lower than in the ALI group because we could not find matches for two ALI boys whose raw score on the KBIT Matrices was too low (0 and 3), and for two ALI boys who had not been administered the test.

⁸As a result of their general functioning levels being in the unimpaired range on all available measures, the ALN group did not differ significantly to their controls on CA, KBIT SS, PPVT 3 raw scores, PPVT SS, and TROG 2 raw scores.

⁹See the discussion soon to follow. We know of no evidence, for example, to suggest that TD children or children with ASD (or any other type of impairment) perform passive or raising movements to positions that don't c-command the trace of movement.

¹⁰We are grateful to an anonymous reviewer for this suggestion.

¹¹We say might because it depends on how it is worked out that reflexives must be part of an A-chain. It really would not involve the A-chain Condition as stated, but rather the predictions that reflexives are part of a chain and that a chain involves c-commanding links. So perhaps an incorrectly represented A-chain Condition would mean that children don't know that the links in an A-chain c-command each other in order.

¹²We do not make this argument for the ALN group, as this group consisted of children diagnosed with both HFA and AS, and the latter is argued not to involve language delays (Howlin 2003).