Investigating the Extended Optional Infinitive Hypothesis in early child German.

Neumann and Charleen gave the talk at the LuCiD Conference in Manchester, UK in 2016.

Abstract:

Understanding how children learn the pattern of verb marking in their language and how this process goes wrong in children with SLI is fundamental to our understanding of language acquisition. One model of this process, which has been particularly influential in the SLI literature is the (Extended) Optional Infinitive ((E)OI) Hypothesis (Wexler, 1994; Rice et al., 1997). According to this view, children’s verb-marking errors reflect a stage in which their grammars allow the use of non-finite forms (e.g. bring) in contexts in which finite forms (e.g. brings, brought) are required, a stage which is particularly protracted in children with SLI.
The main strength of the (E)OI Hypothesis is its ability to make predictions about the cross-linguistic data. For example, Rice et al. (1997) show that German children with SLI produce Optional Infinitive errors at significantly higher rates than MLU-matched controls. Moreover, this is despite the fact that, like typically developing children, they do not make verb-positioning errors in their speech
In this study, we investigate Rice et al.’s EOI account of SLI in German by comparing data from Bastian, a German child with SLI, with control data from a typically developing German child: Leo. The Bastian (BAS) corpus was collected by Bittner (2010) at ZAS Berlin and consists of weekly transcripts from age 2;0 to 4;6. The Leo corpus was collected at the MPI in Leipzig and consists of daily transcripts from age 1;11 to 4;11.
In a first analysis, we compare developmental changes in the rate of OIs in Bastian’s data with control data from Leo at equivalent MLUs. We follow Rice et al. in focusing only on utterances that included an overt subject and at least three constituents. This comparison, which is plotted in Figure 1, reveals a developmental stage from MLU 1.71 to 2.24 in which Bastian produces OIs at significantly higher rates than Leo at equivalent MLUs, and is hence consistent with Rice et al.’s (1997) findings.
In a second analysis, we investigate the rate of verb positioning errors in Bastian’s speech. The results of this analysis, which are presented in Table 1, show that Bastian produces verb positioning errors at much higher rates than the EOI hypothesis predicts.
In a final analysis, we consider the possibility that Bastian’s higher rate of OIs reflects a difference in Bastian and Leo’s MLU distributions. More specifically, we compare the average length of utterances that include a verb (MLUv) in the two children at each MLU point. The results of this analysis, which are presented in Table 2, reveal a systematic advantage for Leo, who shows higher MLUvs at equivalent MLUs at all 8 data-points. This finding suggests that the EOI stage in Bastian’s data may be an artefact of the failure to control for differences in MLUv. It also raises doubts about the practice of matching language impaired and typically developing children on overall MLU.
Understanding how children learn the pattern of verb marking in their language and how this process goes wrong in children with SLI is fundamental to our understanding of language acquisition. One model of this process, which has been particularly influential in the SLI literature is the (Extended) Optional Infinitive ((E)OI) Hypothesis (Wexler, 1994; Rice et al., 1997). According to this view, children’s verb-marking errors reflect a stage in which their grammars allow the use of non-finite forms (e.g. bring) in contexts in which finite forms (e.g. brings, brought) are required, a stage which is particularly protracted in children with SLI.
The main strength of the (E)OI Hypothesis is its ability to make predictions about the cross-linguistic data. For example, Rice et al. (1997) show that German children with SLI produce Optional Infinitive errors at significantly higher rates than MLU-matched controls. Moreover, this is despite the fact that, like typically developing children, they do not make verb-positioning errors in their speech
In this study, we investigate Rice et al.’s EOI account of SLI in German by comparing data from Bastian, a German child with SLI, with control data from a typically developing German child: Leo. The Bastian (BAS) corpus was collected by Bittner (2010) at ZAS Berlin and consists of weekly transcripts from age 2;0 to 4;6. The Leo corpus was collected at the MPI in Leipzig and consists of daily transcripts from age 1;11 to 4;11.
In a first analysis, we compare developmental changes in the rate of OIs in Bastian’s data with control data from Leo at equivalent MLUs. We follow Rice et al. in focusing only on utterances that included an overt subject and at least three constituents. This comparison, which is plotted in Figure 1, reveals a developmental stage from MLU 1.71 to 2.24 in which Bastian produces OIs at significantly higher rates than Leo at equivalent MLUs, and is hence consistent with Rice et al.’s (1997) findings.
In a second analysis, we investigate the rate of verb positioning errors in Bastian’s speech. The results of this analysis, which are presented in Table 1, show that Bastian produces verb positioning errors at much higher rates than the EOI hypothesis predicts.
In a final analysis, we consider the possibility that Bastian’s higher rate of OIs reflects a difference in Bastian and Leo’s MLU distributions. More specifically, we compare the average length of utterances that include a verb (MLUv) in the two children at each MLU point. The results of this analysis, which are presented in Table 2, reveal a systematic advantage for Leo, who shows higher MLUvs at equivalent MLUs at all 8 data-points. This finding suggests that the EOI stage in Bastian’s data may be an artefact of the failure to control for differences in MLUv. It also raises doubts about the practice of matching language impaired and typically developing children on overall MLU.