Testing two different models of verb-marking error in children with Developmental Language Disorder and language-matched controls.

List, C., Lieven, E., Ambridge, B. & Pine, J.M., (2017). Testing two different models of verb-marking error in children with Developmental Language Disorder and language-matched controls. Poster presented at the Many Paths to Language (MPaL) workshop, Nijmegen, The Netherlands.


The aim of this study was to test two different models of the pattern of verbmarking error in German-speaking children with Developmental Language Disorder (DLD) and language-matched controls. According to the (Extended) Optional Infinitive ((E)OI) Hypothesis [277] [217] children’s verb-marking errors reflect a stage in which their grammars allow non-finite forms (e.g. ‘build’) in contexts in which finite forms (e.g. ‘builds’) are required. This stage extends further up the MLU range in children with DLD, such that children with DLD produce OI errors at higher rates than both age-matched and language-matched controls. According to the Dual-Factor Model [97] [98], children’s verb-marking errors reflect the learning of infinitives from compound finite structures in the input (which, in German, take the form ‘He can a house build-INF’). Children are expected to produce infinitives at high rates in compound-finite contexts in which they are effectively truncated modals, and at lower rates in simple finite contexts in which they reflect a process of defaulting to the infinitive when it is a particularly high frequency form of the verb — with children with DLD being more likely to default than typically developing children.

In order to test these models, a verb-elicitation experiment was designed and conducted with a group of 50 German-speaking children with DLD (3;0 to 5;6) and a group of 50 language-matched controls (2;0 to 2;11). This study involved eliciting a range of verbs that differ in the relative frequency with which they occur in finite and infinitive form in two conditions: a simple-finite condition (e.g. ‘Lisa builds a tower. Peter ... ’) and a compound-finite condition (e.g. ‘Peter can a house build-INF. Lisa ... ’). An example context from the elicitation task is given in Figure 1. The participants also completed a battery of linguistic and non-verbal IQ tests to establish that they met the criteria for inclusion in the study. The critical predictions of the study were that a) children with DLD would make more OI errors than language-matched controls, particularly in simple-finite contexts EOI Hypothesis) and b) both groups would make more OI errors in compound-finite than in simple-finite contexts (Dual-Factor Model). To verify the models predictions the rates at which the children produced correct responses (as opposed to OI errors) were entered into a 2x2 Mixed ANOVA, where the between-groups factor was Group (DLD, TD) and the within-groups factor was Condition (Simple-Finite, Compound-Finite). The results, which are plotted in Figure 2, show a significant main effect of condition, with higher rates of correct responses in simple-finite contexts and no significant main effect of group. There was also a marginally significant Condition x Group interaction, which reflected the fact that the DLD group  performed better than the TD group in the compound-finite
condition. These results count against the EOI Hypothesis, since they fail to show higher rates of OI errors in DLD children than in language-matched controls. On the other hand, they are broadly consistent with the Dual-Factor Model, since they show higher rates of OIs in the compound-finite than the simple-finite Condition. A stronger test of the Dual-Factor Model is to investigate whether children’s tendency to produce OI errors in simple-finite contexts can be predicted on a verb-by-verb basis in terms of the relative frequency with which verbs occur in infinitive and finite form in German child-directed speech. Further analysis (using mixed effect models) showed a significant effect of relative input frequency in the predicted direction which provides further support for the Dual-Factor Model.