The production of relative clauses in syntactic SLI: A window to the nature of the impairment

Abstract

The study explored the ability of children with syntactic SLI (S-SLI) to produce relative clauses, using two structured elicitation tasks. A preference task and a picture description task were used to elicit subject and object relative clauses. The participants were 18 Hebrew-speaking children with S-SLI aged 9;3 – 14;6, and the control group included 28 typically developing children aged 7;6 – 11;0. The rate of target responses as well as the types of other responses the S-SLI group produced were analysed and compared to the control group. The results of both tasks indicated that the children with S-SLI had a deficit in the production of object relatives. Their production of subject relatives was better, though below the performance of the control group. Several response types were used exclusively by the S-SLI group: avoidance of object relatives and production of subject relatives and simple sentences instead, thematic role errors and thematic role reduction. Importantly, the S-SLI children did not omit complementizers, nor did they make other structural errors. These results suggest that the deficit is related to thematic role assignment to moved constituents, and not to a structural deficit in embedding.

Keywords:

• Relative clauses
• Hebrew
• specific language impairment
• production
• syntax

Notes

1 According to some linguistic analyses of relative clause derivation, the relative head itself moves from within the embedded clause to CP (Kayne, 1994; Vergnaud, 1974). Other theories take the movement in relative clauses to be a movement of a relative operator (Chomsky, 1986, 1995). According to them an empty operator or the relativizer who or which moves from within the embedded clause to the specifier position of the embedded CP, where it is co-indexed with the head of the relative clause (see Sauerland, 2000, for a discussion of the two analyses).

2 Hebrew also allows topicalization and verb movement within the relative clause, but these occur in a more formal language and none of the children in the control group used these options, so it is not relevant here.

3 All the analyses presented below were done both with t-tests and with nonparametric tests—Wilcoxon signed rank for comparisons within the group of SLI, and Mann-Whitney for the comparison between groups. The results of the t-tests and the nonparametric tests were very similar, every result that was significant with the t-test was also significant with the nonparametric tests, and every non-significant result was non-significant in both tests. In the text we present the results of the t-tests.

4 Note that according to Reinhart and Siloni (2005) some languages express the impersonal meaning by an arbitrary pro, whereas other languages use a lexical operation of saturation or arbitrarization of the agent role. Unimpaired speakers of Hebrew are thought to use the arbitrary pro in this context, but it is possible that the children with S-SLI, who showed mismatches between lexical operations, verbal morphology, and thematic roles, incorrectly reduce the external theta role in this construction too.

5 Another possibility is that there is actually more than one type of S-SLI, one that is related to movement and thematic roles, as described in the current study, and another that involves impairment in the syntactic tree.

6 Why are Hebrew-speaking children treated on passives is yet another question. This might be because treatment and diagnostic tools are many times translated from English without the necessary adaptations to Hebrew.

7 This is also consistent with the errors they made in which the verb morphology, and hence the verb thematic grid—the number of thematic roles it can assign—did not match the number of noun phrases in the sentence. Because it might be that the instances in which they used full passives could also be instances of adjectival passives with two NPs. Interestingly, of the six participants with SLI who used passives in the picture task, four were the four participants who made the errors of morphology not matching the thematic structure.

8 Whereas most of the children with hearing impairment in Friedmann and Szterman (2006) produced doublings in relative clauses like “The girl bought the boots that she wanted the boots”, most of the doubling errors in the S-SLI group came from one participant. Interestingly, his files include recurring hearing infections at the first 2 years of life, possibly pointing to a hearing impairment (and hence insufficient exposure to linguistic input) source for his S-SLI.